| Dokumendiregister | Riigikogu |
| Viit | 1-2/26-353/1 |
| Registreeritud | 05.06.2026 |
| Sünkroonitud | 05.06.2026 |
| Liik | EL dokument |
| Funktsioon | |
| Sari | |
| Toimik | Ettepanek - SEC(2026) 502, SWD(2026) 502, SWD(2026) 503, COM(2026) 502 |
| Juurdepääsupiirang | Avalik |
| Adressaat | |
| Saabumis/saatmisviis | |
| Vastutaja | |
| Originaal | Ava uues aknas |
EN EN
EUROPEAN COMMISSION
Brussels, 3.6.2026
COM(2026) 502 final
2026/0138 (COD)
Proposal for a
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strengthening Europe’s cloud and AI
ecosystem (Cloud and AI Development Act)
{SEC(2026) 502 final} - {SWD(2026) 502 final} - {SWD(2026) 503 final}
(Text with EEA relevance)
EN 1 EN
EXPLANATORY MEMORANDUM
1. CONTEXT OF THE PROPOSAL
• Reasons for and objectives of the proposal
In recent years, cloud and AI technologies, services and applications have evolved from new
concepts to indispensable pillars underpinning the functioning of our economy and society. AI
unlocks unprecedented opportunities from automation and data-driven decision-making to
personalised services, while cloud computing provides the computational resource, software
building blocks and interfaces necessary for efficient AI development and deployment. The
rapid proliferation of AI has resulted in an unprecedented and growing demand for
computational capabilities. Consequently, computing infrastructures are no longer mere
technical assets but have become strategic resources critical to the Union’s economic security,
sovereignty, resilience, and competitiveness.
As Mario Draghi’s report ‘The future of European competitiveness’ states, the EU must
maintain a foothold in areas where technological sovereignty is required, such as security and
encryption (“sovereign cloud” solutions) and thus reduce critical external dependencies by
strengthening homegrown cloud and AI capabilities and infrastructure. To this end, the
Draghi report calls on the European Commission to take targeted actions aimed at regaining
and retaining control over data and cloud computing services, expanding domestic
computational capacity and establishing a robust financial and talent flywheel to drive
innovation.
These objectives were later enshrined in the European AI Continent Action Plan, which
presented a strategic roadmap to ensure European AI leadership. Central to achieving this is a
nexus between five key domains: computing infrastructures, data, skills, development and
adoption of AI algorithms, and regulatory simplification. The ongoing deployment of AI
factories and AI gigafactories aims to provide broad access to high-capacity, next-generation
computational resources for European businesses and researchers requiring AI capabilities. To
complement this, the EU needs to expand its cloud and data centre capacity to support the
wider deployment and diffusion of AI.
The Union’s limited data centre capacity poses a significant threat to its ability to benefit
from the digital transformation and adopt AI-driven solutions, notably those requiring low-
latency compute capacity. In particular, several obstacles hinder the rapid deployment of data
centres in the EU. As data storage and processing demands continue to rise - particularly due
to the surge in AI workloads - the lack of data centre capacity in the EU forces European
enterprises to route critical workloads through foreign hyperscaler infrastructure. This makes
the EU a less attractive destination for tech investment than regions with more abundant,
lower-cost compute resources.
The current landscape of cloud and AI is characterised by a pronounced dependence on a
limited pool of third-country providers. While the EU market for cloud computing services
market is growing significantly, the market share of EU providers decreased from 29% in
2017 to 15% in 2022 and has remained stagnant since then.
Currently, three non-EU hyperscalers control over 70% of the European cloud market. Large
market incumbents are subject to third-country jurisdictions where laws with an
extraterritorial effect apply, including laws mandating data access and transfer that may
conflict with EU fundamental rights and data protection frameworks. This dependence also
exposes European users to the risks related to operational discontinuity, particularly in
scenarios where unilateral decisions by third-country actors could disrupt service provision.
EN 2 EN
Against this background, Europe has world-class research and development capabilities,
vibrant open-source communities and a strong industrial base in cloud and AI, which however
remain largely untapped.
Against this background, the Commission has prepared the proposed Cloud and AI
Development Act (‘the proposal’) which aims to address the limited and geographically
concentrated availability of computing capacity in the EU and the risks associated with
dependence on cloud and AI supplied by non-European providers. The proposal aims to
(1) increase computing capacity and AI developed and deployed in the EU through
innovative and sustainable Cloud and AI technologies;
(2) ensure attractive conditions for the deployment of sustainable and innovative
computing capacity across the Union;
(3) address concerns regarding data sovereignty and operational continuity of cloud and
AI;
(4) help protect public order by making the supply of cloud computing services more
resilient, in particular in the public sector.
The proposal responds to the need for a coordinated ‘ecosystem approach’ to make the EU
more competitive and resilient in the cloud and AI area. It combines supply-side measures to
boost domestic capabilities, demand-side measures to drive adoption, and enablers to foster
innovation and investment into cloud and AI. It places a specific focus on open source as a
lever to boost technological sovereignty, in line with the EU Open Source Strategywhich
aims to promote open European alternatives across the European technology stack.
First, the proposal supports projects that are the outcome of the research and innovation
initiative launched under this framework and implemented jointly with Member States. This
initiative will integrate networks, cloud, AI and software into coherent ecosystems to address
the following:
(1) future challenges across energy-efficient compute infrastructure;
(2) autonomy across the cloud stack;
(3) advanced EU capabilities in advanced AI technologies such as frontier AI, physical
AI and industrial AI;
(4) adoption of cloud and AI across the public and private sectors.
The proposal places a particular emphasis on large-scale, cross-sectoral initiatives addressing
the most strategic technological and industrial challenges (‘grand challenges’). It will
demonstrate the feasibility of this effort and pave the way for similar initiatives in the future,
creating the conditions for investment in next-generation infrastructure and technologies.
Second, the proposal responds to the growing gap in data centre capacity with a framework
that simplifies and harmonises the deployment of data centres EU-wide, while ensuring their
sustainability. It aims to triple EU capacity in the next five-to-seven years and reach the
needed capacity by 2035, while ensuring balanced geographic deployment across Member
States. To support this, the proposal presents a mechanism to identify and support data centre
strategic projects with significant built-in innovation and sustainability or that contribute to
the balanced distribution of computing capacity.
Third, the proposal aims to mitigate the risks stemming from the EU’s reliance on third
countries for cloud computing services via a single EU-wide sovereignty framework. It
provides a harmonised and auditable set of criteria at different levels of sovereignty of cloud
EN 3 EN
computing services. It also provides a framework for services to be assessed and formally
recognised at a particular level of sovereignty. Finally, it obliges the Member States to
undertake sovereignty risk assessments to determine which sub-sectors and use cases should
be served by services aligned with the respective sovereignty levels to ensure appropriate
protection in terms of data confidentiality, to ensure operational autonomy and to prevent
harm that could undermine public order.
Complementing the Cybersecurity Act (CSA2) revision, which addresses supply chain risks,
the proposal ensures that contracting authorities can use sovereign cloud computing services.
Together, the proposal and the CSA2 fill long-standing gaps in sovereignty and non-technical
risks. At the same time, cloud computing services in Europe must meet high cybersecurity
standards, which calls for a robust cybersecurity framework that provides a comprehensive
response to today’s geopolitical security challenges. In this legislative context, work will
resume on the European Cybersecurity Certification Scheme for Cloud Services (EUCS).
The proposal finally provides a framework for contracting authorities to make informed
purchasing decisions and leverage their buying power towards lowering existing
dependencies, including through the use of sector-specific EU-added-value award criteria and
common procurement to drive innovation and growth, with a focus on creating concrete
opportunities for smaller EU-based providers.
Taken together, the measures set out in the proposal establish the foundations for a resilient,
high-performance EU cloud and AI ecosystem. They position Europe not just as a consumer
of advanced digital technologies but as a global hub for trusted, sovereign and scalable digital
infrastructure capable of shaping the standards, capabilities and markets of the next
technological wave.
• Consistency with existing policy provisions in the policy area
The proposal is consistent with the rules on switching between data processing services
introduced by the Data Act. By enabling switching and removing key sources of vendor lock-
in, the Data Act seeks to ensure that cloud computing service providers in the EU compete on
quality, innovation, and price. It seeks to enable cloud users to freely choose the provider that
best meets their needs and combine offers of different providers in a multi-cloud approach.
However, the Data Act does not contain elements to shape up a more competitive offer of
European cloud computing services or encourage the entry into the market of a more diverse
set of cloud computing service providers.
The Data Act opens the path towards a possible reduction of dependencies on non-EU
providers but does not build the road towards a more sovereign and trusted EU cloud
computing sector. Its cloud switching and interoperability provisions, however, make it
possible for users to embrace European cloud computing services more strongly. The Data
Act is thus an enabler for the proposal.
The proposal is also consistent with the Digital Markets Act (DMA). The DMA covers cloud
computing services as a core platform service, meaning that cloud computing service
providers designated as gatekeepers would have to comply with a set of obligations to
increase fairness and market contestability. So far, no cloud computing service provider has
been designated as a gatekeeper for their services. However, on 18 November 2025, the
Commission opened three market investigations on cloud computing services under the
DMA. Two of these market investigations will assess whether two providers should be
designated as gatekeepers for their cloud computing services (in other words whether they act
as important gateways between business users and end users). The third market investigation
EN 4 EN
will assess if the DMA can effectively tackle practices that may limit competitiveness and
fairness in the cloud computing sector in the EU.
While certain providers of cloud computing services could be regulated under both this
proposal and the DMA, the DMA has different objectives and does not contain measures that
would actively promote the uptake of sovereign cloud computing services. The DMA only
aims at maintaining and promoting a fair and contestable cloud market in the Union,
regulating specific behaviours of companies designated as gatekeepers and thus intervenes at
a different level than the proposal, which focuses on the uptake and use of the services
provided.
The proposal also reinforces key objectives of the AI Act. The AI Act harmonises rules for AI
systems and general-purpose AI models to be placed on the EU market, improving the
functioning of the internal market and promoting the uptake of human-centric and trustworthy
AI along the value chain. The AI Act ensures a high level of protection of health, safety and
fundamental rights. It does not cover aspects of sovereignty.
The proposal further complements EU’s broader digital policy framework, including the EU
Open Source Strategy, the Digital Decade Policy Programme and Apply AI Strategy.
The EU Open Source Strategy proposes to foster open source for sovereignty,
competitiveness and security through a series of focused measures, some of which are
proposed in the Cloud and AI Development Act.
The Digital Decade Policy Programme focuses on four cardinal points, under which its
targets fall, namely (i) a digitally skilled population and highly skilled digital professionals;
(ii) secure and sustainable digital infrastructures; (iii) digital transformation of businesses; and
(iv) digitalisation of public services. More specifically, the Digital Decade Policy Programme
sets out a target for monitoring the deployment of edge nodes, but it does not include either a
target for measuring progress in the deployment of compute capacity or data centres in the EU
or concrete support measures for their deployment. This proposal complements the Digital
Decade Policy Programme by leveraging the existing yearly monitoring exercise, thus
creating synergies with the existing framework. It also helps advance all four Digital Decade
policy programme cardinal points, notably by establishing concrete measures centred on
developing innovative AI-enabling technologies, deploying expanded compute capacity, and
creating a trust framework for enhanced use of cloud and AI.
Moreover, the Apply AI Strategy sets out concrete actions to harness AI’s transformative
potential, with a focus on boosting adoption across key industry sectors and the public sector.
It also introduces support measures to strengthen the EU’s technological sovereignty by
tackling cross-cutting challenges in AI development and deployment. The proposal underpins
these objectives and contributes to their implementation by introducing targeted measures
aimed at supporting the development and deployment of cloud and AI, increasing access to
compute capacity and building trust in cloud computing services.
The proposal is fully compatible with the EU’s June 2025 Communication on an
International Digital Strategy. It creates a transparent, non-discriminatory blueprint for
digital autonomy that allows the EU to build resilient, sovereign tech infrastructures at home
while providing a trusted, legally sound model for international partnerships and multilateral
governance abroad. It is fully consistent with the Union’s international commitments and
partnerships and will secure access to the internal market to entities from partner countries
that meet required levels of Union assurance.
EN 5 EN
• Consistency with other Union policies
The proposal complements EU’s broader policy framework on cybersecurity and digital
resilience.
This proposal needs to be read in conjunction with the proposal for the review of the Chips
Act, which includes measures to promote investments in advanced semi conductors, increase
supply chain resilience and demand creation through increased cooperation between the
semiconductor supply chain and end markets.
The Directive on Security of Network and Information Systems (NIS2) improves the
cybersecurity risk management of cloud computing service providers and data centres in the
EU, resulting in greater trust. However, it does not contain measures to boost the uptake and
use of such services and is fully focused on technical cybersecurity as opposed to broader
sovereignty considerations.
As detailed earlier, the proposal further supplements the Cybersecurity Act’s focus on cloud
cybersecurity with sovereignty considerations. Certification under the Cybersecurity Act can
address technical cybersecurity criteria but is not suited for addressing sovereignty concerns
that go beyond these technical elements. Meanwhile, the European Union Agency for
Cybersecurity (ENISA) has been working on developing a European Cybersecurity
Certification Scheme for Cloud Services (EUCS), which has not yet been adopted. When
finalised, it could be leveraged in the framework for sovereign cloud computing services as a
way of ensuring that an audited service meets the highest cybersecurity standards.
Furthemore, the proposed review of the Cybersecurity Act reinforces the trustworthiness of
the hardware and software ICT supply chain.
The proposal also supports the objectives of the Digital Operational Resilience Act
(DORA). The Digital Operational Resilience Act shapes compliance obligations for cloud
computing service providers. It indirectly covers cloud computing service providers if they
provide services to specified financial entities or if their role is significant enough in terms of
operational resilience. It has a sectoral scope and is specific to the financial sector. Under the
Digital Operational Resilience Act, cloud computing service providers must implement ICT
risk management and conduct regular incident response testing to comply with the
requirements for critical third-party service providers. The financial institutions concerned,
which could be public in nature, must carry out due diligence on the cloud computing service
providers they work with.
The proposal also supports the objectives of the Preparedness Union Strategy, which
identifies dependence on critical digital infrastructure as a systemic risk and calls for a whole-
of-government approach to ensuring the continuity of essential services in crisis scenarios.
The sovereignty framework established by this Regulation, and in particular the risk
assessment mechanism in Article 29, contributes directly to the digital preparedness
dimension of that Strategy by ensuring that the cloud and AI services underpinning
emergency management, civil protection coordination and disaster response operations are
provided at the appropriate Union assurance level. The proposal is therefore consistent with,
and supportive of, the Union’s broader resilience and preparedness policy objectives.
The proposal is consistent with existing rules on the processing of personal data, including the
General Data Protection Regulation (GDPR) and the EU-US Data Privacy Framework.
However, while the EU-US Data Privacy Framework addresses transatlantic data transfers, it
does not remove sovereignty concerns about dependence on third-country providers. The
proposal thus complements the EU-US Data Privacy Framework as the notion of sovereignty
goes beyond data transfers and relates to operational autonomy too.
EN 6 EN
The proposal also supplements the Public Procurement Directives. Public authorities in the
EU rely heavily on non-EU cloud computing service providers and associated problem drivers
require a nuanced and targeted sectoral approach, which is not covered by the existing Public
Procurement Directives and would be difficult to account for sufficiently through an
overarching approach. The proposal therefore provides a sector-specific approach to
sovereignty – the many layers of which cannot be addressed in the horizontal acquis that sets
out general principles for the design of procurement procedures. The proposal includes
complementary award criteria that are also tailored to the specificities of cloud computing
services and critical dependencies on third countries in this sector.
The proposal complements the Digital Networks Act, which supports the development of a
robust, fast, secure, cutting-edge digital networks and thus be beneficial to the deployment of
data centres in the EU, for which high-speed, gigabit and beyond connectivity is a
prerequisite. The proposal leverages the Digital Networks Act’s advancements for
connectivity and thus stays focused on the deployment of data centres capacities, not the prior
or parallel build-out of the necessary connectivity infrastructure. The Digital Networks Act
also addresses the convergence of networks infrastructure, including scenarios where a cloud
computing service provider operates an electronic communications network and has so far not
been subject to obligations under the European Electronic Communications Code although
falling into its scope. Thus, the Digital Networks Act will clarify the procedures for
connectivity between providers of various networks and other market participants within a
broader ecosystem cooperation.
The proposal contributes to the objectives of the Energy Efficiency Directive on guiding the
data centre industry towards greater energy efficiency. However, beyond transparency and
reputational considerations, it does not set incentives for data centre operators to improve
their sustainability performance and does not contain measures for accelerating the roll-out of
sustainable data centres across the EU or increasing related investment.
Similarly, the EU Code of Conduct on Data Centre Energy Efficiency does not concern
measures to deploy a data centre, instead focusing fully on practices for operating one. Thus,
the proposal provides measures to incentivise the roll-out of energy-efficient data centres,
aligning with the Strategic roadmap for digitalisation and AI in energy, which seeks to
optimise energy consumption in digital technologies while accelerating the EU’s twin green
and digital transition. To identify which data centres are sustainable, the proposal refers to the
rating scheme developed under the Energy Efficiency Directive.
The proposal also complements the European Grids Package, which aims to ensure grids
are in place and ready to uptake future loads in a horizontal manner. Here, the proposal
focuses on data centres as an ultimate client of grid capacity and ensures data centres location
considers grid availability, information is exchanged sufficiently in advance to feed into grid
planning and hence ensure timely connection of data centres.
The European Grids Package sets out ways in which Member States can accelerate grid
connections, including a more efficient structure of their grid connection queue, for example
by considering a project’s readiness and grid-friendly uses as opposed to a pure first-come-
first-served approach. The proposal leverages these considerations.
The proposal also complements the Renewable Energy Directive and can leverage it. Data
centres may benefit from the increased availability of renewable energy and storage even if
they are not themselves covered by the Directive. Proximity to acceleration areas for
renewables could also be a relevant factor in designating sites for faster data centre
deployment.
EN 7 EN
The Industrial Accelerator Act puts forward measures to accelerate permitting, including
setting out the principle of ‘one project, one procedure’ and establishing a single digital
procedure to cover the entire permit-granting process. The Industrial Accelerator Act also
requires Member States to designate at least one manufacturing industrial acceleration area or
cluster where further business facilitation measures apply, including prioritised access to
materials, and access to EU finance. However, as the proposed industrial acceleration
measures only account for manufacturing facilities and do not account for the specific needs
of data centres as well as their benefits, they have been left out of scope of the Industrial
Accelerator Act. Addressing these differences, the proposal complements the general
industrial acceleration measures with measures that are tailored to the accelerated deployment
of data centres.
2. LEGAL BASIS, SUBSIDIARITY AND PROPORTIONALITY
• Legal basis
The legal basis for this proposal is Article 114 of the Treaty on the Functioning of the
European Union (TFEU), which empowers the EU to adopt measures for improving the
functioning of the internal market through the harmonisation of national provisions.
The current fragmentation in data centre deployment is driven by divergent national
approaches to capacity expansion, sustainability requirements, and permitting procedures.
These risks creating regulatory disparities that could undermine the internal market.
Variations in public procurement practices for cloud computing services, as well as
inconsistent sovereignty criteria, may hinder providers’ ability to operate seamlessly across
Member States, leading to market inefficiencies and unequal competitive conditions.
Given these challenges, EU-level intervention in the form of a legislative proposal is justified
under Article 114 TFEU, as it seeks to eliminate barriers to the internal market and levels the
playing field for cloud computing service providers.
Additionally, the proposal draws on Article 173(3) TFEU, which provides the legal basis for
measures aimed at enhancing the EU’s industrial competitiveness and innovation capacity.
The current shortage of computing capacity in the EU constrains the ability of European
industries to fully benefit from the use of cloud and AI technologies, especially those
requiring low-latency and high-performance computing. By increasing the availability of
energy-efficient compute capacity and the development and deployment of cutting-edge cloud
and AI technologies, this initiative directly helps strengthen Europe’s industrial
competitiveness and technological leadership, in line with the objectives of Article 173(3)
TFEU.
Given the proposal’s dual objective of remedying internal market distortions and bolstering
the EU’s industrial competitiveness, it will be adopted as a single legislative instrument under
the cumulative legal basis of Articles 114 and 173(3) TFEU. This joint legal foundation
provides a unified regulatory response to the interlinked challenges of market fragmentation
and strategic industrial capacity-building within the EU’s cloud and AI ecosystem.
• Subsidiarity (for non-exclusive competence)
EU action has a clear added value in addressing the problem of limited and geographically
concentrated availability of computing capacity. By providing a common approach to
accelerating data centre deployment, it enables coherent planning and deployment of
computing capacity in a geographically balanced way, while avoiding a race to the bottom
and reducing regulatory complexity for investors and data centre operators.
EN 8 EN
The EU is uniquely positioned to ensure that investment and acceleration policies reflect
collective priorities and avoid fragmentation. EU-level action would ensure that all businesses
and public administrations can access sufficient compute capacity to meet their needs and is a
prerequisite for Europe to become an AI continent.
In addressing the dependence on cloud computing services supplied by non-European
providers, EU action delivers benefits that exceed what Member States could achieve
individually, especially in addressing the underlying market failures of imperfect information.
This will improve the functioning of the internal market and enable cloud computing
service providers to grow beyond their national markets.
• Proportionality
The proposal adopts a targeted and proportionate approach to address the critical bottlenecks
in the single market, specifically the compute capacity deficit and overreliance on third-
country providers. The proposed solutions are confined to measures essential for achieving
core objectives of the proposal notably the strengthening of the EU technological sovereignty,
fostering a competitive EU cloud and AI market, and supporting an enhanced availability of
sustainable computing resources in the EU.
By focusing on harmonised standards for sovereign cloud computing services, streamlined
data centre deployment, and EU-wide cooperation mechanisms, the proposal avoids excessive
regulation while directly tackling the structural barriers that impede the digital advancements
across Europe. The chosen measures represent the most suitable and least intrusive means of
addressing the identified market and regulatory failures, as they leverage existing EU
instruments while introducing only those new provisions necessary for cohesive and efficient
implementation.
The proposal will give rise to direct compliance costs, covering both administrative and
adjustment expenditures. These are to be borne mainly by national and local public
authorities, and businesses such as data centre operators, cloud-computing service providers
as well as their subcontractors and private-sector entities operating in sectors listed under
Annex I to the NIS2 Directive, so that they comply with the obligations set out in this
Regulation.
However, the exploration of different options and their expected costs and benefits has
resulted in a balanced design of the instrument. The costs to public authorities will be
counterbalanced by the value of reduction of total cost of ownership for IT systems and faster
and more reliable procurement processes, while the costs to businesses will be
counterbalanced by the value of improved predictability of regulatory direction and reduced
fragmentation of national approaches together with administrative simplification.
• Choice of the instrument
The Commission proposes a Regulation as the optimal legal instrument for the
accomplishment of the proposal’s objectives, given its capacity to ensure uniform application
and immediate effect across all Member States. A Regulation provides the necessary legal
certainty and consistency to fully harmonise the regulatory framework for cloud computing
services. This approach is essential to remove single market barriers, particularly in areas such
as sovereignty and sustainability standards, where divergent national rules could otherwise
undermine the EU’s technological sovereignty and competitiveness. Given the geopolitical
urgency, including the need to reduce strategic dependencies on third-country providers, a
Regulation ensures a rapid, coordinated, and EU-wide response, preventing delays or
inconsistencies that could arise from Member States acting greater independence.
EN 9 EN
3. RESULTS OF EX-POST EVALUATIONS, STAKEHOLDER
CONSULTATIONS AND IMPACT ASSESSMENTS
• Stakeholder consultations
Between October 2024 and November 2025, an extensive stakeholder consultation was
carried out comprising the following consultation activities.
Between April and July 2025, the Commission organised a Public Consultation consisting of
an online questionnaire and a call for evidence for stakeholders to submit detailed position
papers outlining their views and recommendations on the objectives and proposed actions
envisaged by the proposal, while also giving direct evidence to inform the design of policy
options. A total of 436 responses were received: 243 for the consultation survey and 193 for
the call for evidence.
A series of workshops, seminars and roundtables complemented the consultation. These
included: (i) the 6th General Assembly and Alliance Forum of the European Alliance for
Industrial Data, Edge and Cloud; (ii) roundtables, namely on investment in cloud compute
with financial investors, on policy measures to facilitate data centre integration to the EU grid,
with European Cloud Service Providers’ CEOs on developing sovereign cloud offerings in the
EU, and with American Chamber of Commerce; (iii) seminars with industry, academia and
public authorities on the economic dynamics of the AI stack; and (iv) presentations on future
EU cloud and AI policy with industry.
The Commission also maintained a continuous dialogue with Member States’ relevant
authorities gathered in the informal Member States Cloud Cooperation Group under the
European Alliance for Industrial Data, Edge and Cloud, complemented by targeted bilateral
meetings.
In addition, the Commission engaged in further bilateral discussions with third countries,
including Japan, Switzerland and the United Kingdom, to present and discuss the considered
policy options while gathering insight on best practices and assess the external effect of
measures considered.
Finally, the Commission held over 100 bilateral meetings with a diverse array of stakeholders,
including industry representatives, academic institutions, think tanks and civil society
organisations supported the formulation of the proposed policy options.
• Collection and use of expertise
The Commission contracted a consortium led by Technopolis Group to conduct a study to
support the evidence collection and analysis stage of the impact assessment preceding the
proposal. The study incorporated multiple stakeholder engagement activities, including over
60 targeted interviews, surveys (over 250 replies), and workshops (over 100 participants).
• Impact assessment
The proposal is accompanied by a comprehensive impact assessment the final version of
which was submitted to the Regulatory Scrutiny Board on 30 April 2026. On 8 May 2026, the
Board issued a positive opinion accompanied by a request for further improvements.
• Fundamental rights
The proposal has been subject to a comprehensive assessment of its implications for
fundamental rights, with particular emphasis on the protection of personal data as guaranteed
under Article 8 of the Charter of Fundamental Rights of the European Union. The proposal
introduces a framework for sovereign cloud computing services, which establishes stringent
EN 10 EN
safeguards to ensure that the processing of personal data involving EU citizens complies with
EU data protection standards, including the GDPR. By setting these standards, the proposal
minimises risks associated with non-compliant data handling while reinforcing trust in digital
infrastructure.
One of the core objectives of the proposal is to reduce dependence on third-country providers,
whose operations may be subject to non-EU jurisdictions that can permit data access and
transfer. Consequently, the proposal enhances the protection of personal data by ensuring that
such data remains under the effective supervision of EU authorities, including Member State
data protection agencies and the European Data Protection Board. This approach strengthens
legal certainty and upholds the right to privacy as enshrined in the Charter.
Furthermore, the proposal fosters a competitive single market for cloud computing services,
where providers are fully bound by EU legislation and subject to applicable obligations. This
not only improves data availability and continuity but also ensures that EU legal obligations,
such as those related to data subject rights and regulatory oversight, are consistently met. By
increasing the market presence of EU-based providers, the proposal further embeds
fundamental rights protections into the digital ecosystem, offering stronger guarantees against
unauthorised access or misuse of personal data. The proportionality of these measures has
been carefully considered to balance innovation with the imperative of rights protection, in
line with the Charter’s principles.
4. BUDGETARY IMPLICATIONS
The budgetary implications of the proposal are described in detail in the accompanying
financial statement. In order to effectively implement the initiative, 25 full-time equivalents
(FTEs) are required, comprised of 9 establishment plan posts and 16 contract agent posts.
This staffing configuration will be achieved through a combination of redeployment and
additional recruitment. Specifically, 15 existing staff members will be reassigned from within
the Commission to support the initiative, leveraging their expertise and experience in
managing similar actions. The necessary staff will be drawn from DG CNECT and DG
DIGIT, where they are currently assigned to relevant units or will be redeployed from within
the Commission services. To supplement these existing resources, the initiative will require
an additional 10 FTEs: of 6 FTEs for DG CNECT and 4 FTEs for DG DIGIT. These
additional staff members will be requested to augment the current staffing levels to ensure the
initiative’s successful implementation and the effective performance of new activities and
tasks.
The additional administrative expenditure associated with new tasks described in the proposal
will be mostly financed through fee-based revenue streams that fall under internal assigned
revenues, thus ensuring a sustainable and viable funding model. Specifically, fee-based
streams will be used to support tasks related to joint procurement and to the administration of
the EuroCloud initiative intended to make it easier for interested Member States to share idle
cloud and data centre capacity.
5. OTHER ELEMENTS
• Implementation plans and monitoring, evaluation and reporting arrangements
The Commission should monitor the application of the proposed Regulation and evaluate its
effectiveness over time. Five years after its entry into force, the Commission should review
the functioning of this proposed Regulation and submit a report to the European Parliament
EN 11 EN
and to the Council. These reports will be public and detail the effective application and
enforcement of the proposed Regulation.
• Detailed explanation of the specific provisions of the proposed Regulation
Title I of the proposed Regulation contains the general provisions, including the subject
matter (Article 1) and the definitions (Article 2).
Title II contains the provisions that establish the framework for the implementation of
development and deployment activities across the cloud and AI ecosystem. The framework is
organised into the Cloud and AI Leadership Initiatives and is reinforced by further supporting
measures. Under Chapter I, Article 3 sets out general objective of the Cloud and AI
Leadership Initiatives aiming to support research and innovation activities and achieve large-
scale capacity throughout the Union’s cloud and AI ecosystem. Article 4 establishes the
operational objectives of the Cloud and AI Leadership Initiatives. Article 5 introduces a
network of Experience and Acceleration Centres for AI built on the European Digital
Innovation Hubs and aiming to support the achievement of the Cloud and AI Leadership
Initiatives objectives. Article 6 sets out the implementation mechanisms for the Cloud and AI
Leadership Initiatives. Article 7 requires Member States to adopt a national cloud and AI
strategy coherent with the Regulation’s objectives within one year of its entry into force.
Article 8 establishes criteria for projects to be recognised by the Commission as a frontier AI
priority project. Article 9 supports the allocation of AI computing resources tofrontier AI
priority projects, while also supporting industrial and physical AI projects and the
development and deployment of AI models for the public sector.
Title III sets out the provision on data centre capacity. Under Chapter I, Article 10 sets out the
obligations for Member States to designate data centre acceleration zones within their
territory, where they are deploying data centre capacity. Article 11 prescribes the conditions
within data centre acceleration zones. Article 12 establishes the obligation for Member States
to designate single information points, or where possible, upgrade or integrate with the
already existing single information points, for data centre operators of data centre projects in
data centre acceleration zones and further principles to be satisfied with respect to the
responsibilities of the information point. Article 13 addresses the facilitated administrative
and permit granting processes for data centre projects deployed in data centre acceleration
zones. Under Chapter II, Article 14, lays down the mechanism for expression of interest and
conditions and designation by the Commission of data centre strategic projects. Under
Chapter III, Article 15 establishes a mechanism for the Commission to monitor the Union
progress in increasing the compute capacity available, the volume of demand for data centre
capacity, and the size of the capacity gap.
Title IV contains the provision relating to autonomy and adoption. Under Chapter I, Section 1,
Article 16 sets out a Union cloud computing sovereignty framework consisting of four
assurance levels and introduces the requirements established in Annex II to the Regulation for
cloud computing services to be considered as providing Union assurance across level 1 to
level 4. Article 17 established a mechanism for cloud computing service providers to be
recognised as providing a Union assurance level 1, 2, 3, or 4, where they must submit an
application for recognition to the national competent authority of establishment. Article 18
sets out conditions and a mechanism for a possible recognition of third-countries as providing
sufficient assurances to allow for cloud computing services controlled from that third country
to become eligible to qualify under Union assurance level 3. Section 2, Article 19 sets out the
conditions for the conformity self-assessment documenting the alignment against the
requirements of the Union assurance level 1. Section 3, Article 20 sets out the framework of
EN 12 EN
assessment performed by auditing organisations against the requirements of the Union
assurance levels 2-4. Article 21 sets out the requirements for audit evidence to be established
as part of the third-party assessment of requirements under Union assurance levels 2-4.
Article 22 sets out the obligation for the Commission to establish and maintain a central
repository of services recognised as offering Union assurance levels 1-4. Article 23 sets out
transparency obligations for cloud computing service providers’, to report any material
changes which may substantiate a resulting change in their recognition as offering Union
assurance levels 1-4. Article 24 sets out the penalties and compensation rules applicable to
infringement by cloud computing service providers. Section 4, Article 25 sets out the Member
States obligations to designate a national competent authority. Article 26 sets out the powers
of national competent authorities designated by the Member States. Section 5, Article 27 sets
out principles of mutual assistance between Member States national competent authorities in
the context of information sharing and investigation for the purpose of this Regulation. Article
28 sets out the principles of cross-border cooperation between Member States national
competent authorities in the context of enforcement actions. Under Chapter II, Section 1,
Article 29 sets out the obligations for Member States and Union entities to conduct risk
assessments to determine the required level of conformity against the Union assurance levels
2-4 for different public sector activities. Article 30 sets out obligations for contracting
authorities that procure cloud computing services to procure, as a minimum requirement,
Union assurance level 1. Where a risk assessment determines that the activities of such
contracting authorities have public order relevance, they must only procure and use services
that have been recognised as offering Union assurance levels 2, 3, or 4. Section 2, Article 31
allows for private sector entities within the meaning of the NIS2 Directive to conduct impact
assessments with a similar purpose to the ones conducted by Union entities and public sector
bodies. Section 3, Article 32 sets out obligations for the Member States contracting authorities
to apply Union added value criteria in public procurement of cloud computing services and AI
systems within the scope of the Regulation. Article 33 sets out obligations for Member States
to monitor their procurement of innovation of cloud computing services and AI systems,
introduces a lean reporting framework and creates links to national cloud and AI strategies
where Member States must create plans for the achievement of the objectives of this Article.
Under Chapter III, Article 34 establishes the European public sector cloud federation
(‘EuroCloud Federation’) and sets out its scope and purpose. Article 35 sets out the conditions
applicable to the sharing of data centre services and cloud computing services within the
EuroCloud Federation. Article 36 sets out how the Commission can recover the costs incurred
in relation to the EuroCloud Federation, including the establishment of the EuroCloud
Federation and of the platform facilitating the sharing of services among its members. Chapter
IV, Article 37 sets out the conditions for the Commission to carry out procurement activities
for Union entities,for contracting authorities from Member States, and for partner
organisations selected by the Commission, including by establishing the necessary
derogations to the Financial Regulation. Article 38 sets out the necessary arrangements for
implementing the common procurement framework, including the governance mechanism.
Article 39 sets out the applicable public procurement rules. Article 40 sets out the mechanism
for the compensation of costs incurred by the Commission for the operation of the common
procurement framework via fees levied on participating contracting authorities. Chapter VI,
Article 41 sets out the obligations of the Union entities and public sector bodies to encourager
and facilitate thier use of opensource solutions over proprietary ones. Article 42 sets out the
requirements on sharing and reuse of software developed by or for Union entities and public
sector bodies. Article 43 sets out the obligation for the Commission to provide and maintain
an Open Source Solutions Catalogue. Article 44 sets out the rules for the establishment of a
network of Member States Open Source Programme Offices.
EN 13 EN
Title V sets out the final provisions, including the power for the Commission to adopt
delegated acts (Article 45) and implementing acts (Article 46), the review clause (Article 47),
and the specification of the entry into force and dates of application of the Regulation (Article
48).
EN 14 EN
2026/0138 (COD)
Proposal for a
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strengthening Europe’s cloud and AI
ecosystem (Cloud and AI Development Act)
(Text with EEA relevance)
THE EUROPEAN PARLIAMENT AND THE COUNCIL OF THE EUROPEAN UNION,
Having regard to the Treaty on the Functioning of the European Union, and in particular
Article 114 and Article 173(3) thereof,
Having regard to the proposal from the European Commission,
After transmission of the draft legislative act to the national parliaments,
Having regard to the opinion of the European Economic and Social Committee (1),
Having regard to the opinion of the Committee of the Regions (2),
Acting in accordance with the ordinary legislative procedure,
Whereas:
(1) The Union is committed to become a global leader in artificial intelligence (AI). By
enabling faster innovation, greater efficiency and smarter decision-making, AI
contributes to substantial economic, environmental and societal gains and is a
fundamental driver of competitiveness. In order to fully harness the benefits of AI
across the key sectors of the economy, the Union should act with ambition and
foresight, advancing its innovation capabilities, strengthening its competitiveness,
security of supply, while reinforcing its technological sovereignty and strategic
autonomy in cutting-edge digital technologies.
(2) The Union has already laid strong foundations to position Europe as a continent that
leads AI development and uptake, in particular through the AI continent action plan
communication (3) and Apply AI Strategy (4). With Regulation (EU) 2023/2854 (5) of
the European Parliament and of the Council on harmonised rules on fair access to and
1 OJ C , , p. . 2 OJ C , , p. . 3 Communication from the Commission to the European Parliament, the Council, the European
Economic and Social Committee and the Committee of the Regions, ‘AI Continent Action Plant’,
9.4.2025, COM(2025) 165 final. 4 Communication from the Commission to the European Parliament and the Council, ‘Apply AI
strategy’, 8.10.2025, COM(2025) 723 final. 5 Regulation (EU) 2023/2854 of the European Parliament and of the Council of 13 December 2023 on
harmonised rules on fair access to and use of data and amending Regulation (EU) 2017/2394 and
Directive (EU) 2020/1828 (Data Act) (OJ L 2023/2854, 22.12.2023, ELI:
http://data.europa.eu/eli/reg/2023/2854/oj).
EN 15 EN
use of data (‘the Data Act’) and Data Union Strategy (6), the Union has also set up the
necessary frameowrk to create a secure and interoperable single market for data, which
underpins the development of AI.
(3) The European Parliament, the Council, the Commission and the Member States have
committed themselves to cooperate on delivering the Union’s technological
sovereignty in an open manner, in particular by secure and accessible digital and data
infrastructures that enable other technological developsments, supporting the
competitiveness and sustainability of the Union’s industry and economy, in particular
of small and medium-sized enterprises (‘SMEs’) (7) and small mid-caps (‘SMCs’) (8),
and the resilience of the Union’s value chains, as well as fostering the start-up
ecosystem, including through the smooth functioning of the former European digital
innovation hubs (‘EDIHs’), which have been refocused as the experience and
acceleration centres for AI (‘Centres for AI’).
(4) The Union has become increasingly dependent on a limited number of cloud
computing service providers from third countries. Reinforcing the Union’s capacity to
develop and deploy cloud and AI technologies within its territory has become a
strategic priority for the Union’s competitiveness, security of supply and technological
sovereignty, as highlighted in the report by Mario Draghi on the future of European
competitiveness (9) and in line with the Strengthening EU economic security
Communication (10). Those challenges to the Union’s cloud and AI ecosystem call for
the achievement of large-scale technological capacity building and support related to
research and innovation activities and require collective effort by Member States, with
the Union supporting the development and deployment of cloud and AI technologies
and of large-scale cloud computing capacity.
(5) Those dependencies translate not only into limited market shares for the European
cloud computing service providers, but also into significant risks for the Union’s
operational autonomy, resilience and security. The Council has called for a Cloud and
AI Development Act to include common criteria for sovereign cloud computing
services, allowing market transparency risks and risks associated with dependencies,
including extraterritorial effects of legislation adopted by third countries, to be
addressed (11).
(6) A framework for increasing the Union’s resilience and security in the field of cloud
and AI technologies should be established, reinforcing the Union’s cloud and AI
ecosystem by reducing dependencies, enhancing technological sovereignty,
stimulating investment and strengthening the capabilities, security, adaptability and
resilience of the Union’s cloud and AI supply chain.
6 Communication from the Commission to the European Parliament and the Council, ‘Data Union
Strategy’, 19.11.2025, COM(2025) 835 final. 7 As defined in Title I, Article 2, of the Annex to Commission Recommendation 2003/361/EC of 6 May
2003 concerning the definition of micro, small and medium-sized enterprises (OJ L 124, 20.5.2003,
p. 36, ELI: http://data.europa.eu/eli/reco/2003/361/oj). 8 As defined in point 2 of the Annex to Commission Recommendation (EU) 2025/1099 of 21 May 2025
on the definition of small mid-cap enterprises (OJ L 2025/1099, 28.5.2025, ELI:
http://data.europa.eu/eli/reco/2025/1099/oj). 9 The future of European competitiveness. Part A, A competitiveness strategy for Europe (‘the Draghi
report’), September 2024, https://data.europa.eu/doi/10.2872/9356120. 10 Joint Communication from the Commission to the European Parliament and the Council,
‘Strengthening EU economic security’, JOIN(2025) 977 final. 11 Council Conclusions on European Competitiveness in the Digital Decade, 5.12.2025,
https://data.consilium.europa.eu/doc/document/ST-16430-2025-INIT/en/pdf.
EN 16 EN
(7) The framework pursues separate objectives, relying on two distinct legal bases.
(8) First, it is necessary to strengthen the competitiveness, capacity and resilience of the
cloud and AI technological and industrial base of the Union in accordance with
Article 173(3) of the Treaty on the Functioning of the European Union (TFEU). Such
measures should not entail the harmonisation of national laws or regulations. To that
end, this Regulation establishes the Cloud Leadership Initiative and the AI Leadership
Initiative (the ‘Cloud and AI Leadership Initiatives’) to foster the development of
cutting-edge cloud and AI technologies and facilities and ensure their widespread
deployment, bridging the gap between the Union’s advanced research and innovation
capabilities and their sustainable exploitation.
(9) Second, the available compute capacity and resilience of the cloud and AI ecosystem
can best be addressed through Union harmonisation measures on the basis of Article
114 TFEU. A single coherent regulatory framework harmonising certain conditions
for service providers and deployers of cloud computing services, including capacity
building and provision of cloud computing services, is necessary to ensure the
functioning of the internal market.
(10) The definition of ‘cloud computing service’ in this Regulation should be the same as
that in Article 6, point (30), of Directive (EU) 2022/2555 of the European Parliament
and of the Council (12), which defines a ‘cloud computing service’ as a digital service
that enables on-demand administration and broad remote access to a scalable and
elastic pool of shareable computing resources, including where such resources are
distributed across several locations. This definition of ‘cloud computing service’
encompasses on-demand access to AI systems as defined in Article 3, point (1), of
Regulation (EU) 2024/1689 (‘Artificial Intelligence Act’) (13) of the European
Parliament and of the Council, hosted and operated remotely. Only the delivery and
making available of an AI system forms part of the service. The AI system itself and
its underlying model are excluded from the scope of this definition.
(11) The Cloud and AI Leadership Initiatives should reinforce the competitiveness and
resilience of the cloud and AI technological and industrial base of the Union, while
strengthening the innovation capacity of its cloud and AI ecosystem and achieving the
deployment of large-scale digital infrastructures, in line with the objectives set out in
the AI continent action plan and the Apply AI Strategy. In particular, the Cloud and AI
Leadership Initiatives should increase the cloud and data centre capacity of the Union,
while advancing cutting-edge cloud and AI technologies together with broad cloud
and AI adoption across the Union’s economy. It should also bring advanced research
in sector-specific deployment of frontier, physical and industrial AI. The Apply AI
Strategy should extend the ambitions and objectives set out in the AI continent action
plan and should be updated where necessary to take account of the latest technological
developments and progress made in the Cloud and AI Leadership Initiatives.
12 Directive (EU) 2022/2555 of the European Parliament and of the Council of 14 December 2022 on
measures for a high common level of cybersecurity across the Union, amending Regulation (EU)
No 910/2014 and Directive (EU) 2018/1972, and repealing Directive (EU) 2016/1148 (NIS 2 Directive)
(OJ L 333, 27.12.2022, p. 80, ELI: http://data.europa.eu/eli/dir/2022/2555/oj). 13 Regulation (EU) 2024/1689 of the European Parliament and of the Council of 13 June 2024 laying
down harmonised rules on artificial intelligence and amending Regulations (EC) No 300/2008, (EU)
No 167/2013, (EU) No 168/2013, (EU) 2018/858, (EU) 2018/1139 and (EU) 2019/2144 and Directives
2014/90/EU, (EU) 2016/797 and (EU) 2020/1828 (Artificial Intelligence Act) (OJ L, 2024/1689,
12.7.2024, ELI: http://data.europa.eu/eli/reg/2024/1689/oj).
EN 17 EN
(12) In order to pursue the achievement of those general objectives, the Cloud and AI
Leadership Initiatives should support complementary operational objectives designed
as actionable initiatives to be implemented at Union level.
(13) To enable data centres to act as key enablers of a sustainable digital transition, the
Cloud and AI Leadership Initiatives should support research and innovation capacities
for the development of data centre technologies incorporating principles of energy and
resource efficiency by design and throughout operations, with a view to achieving
large-scale sustainability. This includes advancing resource and utilisation-efficient
computing technologies, such as the optimisation of energy and water efficiency, the
use of emerging quantum computing technologies, the development of AI-powered
technologies for server efficiency, the integration of computing infrastructure with
energy grids, the promotion of clean energy adoption and on-site energy generation in
data centres. The Cloud and AI Leadership Initiatives should also provide access to
pilot lines for data centre technologies to steer and accelerate innovation uptake in the
market. Such pilots should function as demonstration facilities for cutting-edge data
centre and edge semiconductor technologies, quantum computing prototypes and AI-
powered data centre operation tools.
(14) The Cloud and AI Leadership Initiatives should also support research, development
activities and the uptake of cloud stack technologies with a view to closing the
capacity gap and strengthening the technological autonomy of the Union. In particular,
the Cloud and AI Leadership Initiatives should foster the development of cloud
computing stacks alternatives for strategic sectors. It should also facilitate the
development of AI-optimised servers and software including processors and
accelerators manufactured and designed in the Union, such as those developed under
Regulation (EU) 2026/XXX (14) of the European Parliament and of the Council. In
addition, both the Cloud and AI Leadership Initiatives under this Regulation and the
Chips for Europe Initiative 2.0 supported by that Regulation should foster the co-
design and cross-optimisation of hardware and software development and the
integration of AI computing infrastructures. Furthermore, the Cloud and AI
Leadership Initiatives should support the complementary development and
deployment of a smart and secure middleware cloud platform for common European
data spaces, in accordance with the Data Union Strategy.
(15) The Cloud and AI Leadership Initiatives should also promote the development of
technologies relying on open standards, open specifications and open source and foster
the development of innovative, competitive and resilient cloud and AI technologies. It
should foster the work on open standards and specifications and the creation of open-
source software foundations supporting the design, development and maintenance of
open-source components, in particular by providing governance and coordination
mechanisms and facilitating the pooling of resources. The Cloud and AI Leadership
Initiatives should also create a catalogue of software tools including open source in
order to enable federation with existing catalogues for the private and public sectors
and to develop a one-stop-shop for open-source resources in the Union.
(16) Frontier AI technologies are advancing rapidly and are expected to have a profound
impact on the Union’s economy and society. As those technologies have become
critical strategic assets, strengthening the Union’s capacity to develop and govern
them is essential to ensure that the AI transition is aligned with Union values, safety
14 References to Chips Act 2.0 to be added once adopted. See Commission proposal [add reference once
published]
EN 18 EN
standards and long-term economic interests. By supporting pioneering projects, the
Union should scale up essential breakthroughs to maintain a competitive edge in the
global digital economy. Fostering the development of frontier AI technologies as
strategic assets should also reduce current dependencies on third-country technologies
and strengthen the Union’s AI ecosystem.
(17) The emergence of physical AI, which refers to AI systems and models capable of
perceiving the physical environment and executing complex actions within that
environment, represents a promising frontier where advanced digital intelligence is
integrated into tangible systems, such as robotics, autonomous drones and self-driving
vehicles. Physical AI is essential to mitigate external dependencies and foster
industrial competitiveness and strategic autonomy. It requires a dedicated approach to
data and computing infrastructure. It is therefore necessary to facilitate the collection
and preparation of high-quality data and access to computing resources. Furthermore,
targeted support for the testing and validation of physical AI models and systems in
diverse real-world environments is necessary to ensure their robustness and reliability.
(18) The digital transformation of the Union’s key industries is a central pillar of the apply
AI strategy. Accelerating the uptake of AI across those strategic sectors is essential to
maintaining global competitiveness and increasing societal benefits. The Cloud and AI
Leadership Initiatives should accelerate the development and uptake of industrial AI
across the Union’s strategic industrial and service sectors, while fostering the
technological development of highly capable sector-specific AI models designed to
meet the operational requirements of the industries prioritised under the Apply AI
Strategy, such as healthcare, transport, including aerospace, automotive,
manufacturing, defence and space, climate and environment and agri-food. The Cloud
and AI Leadership Initiatives should also accelerate the development of service sectors
prioritised in the Apply AI Strategy and scientific discovery as priorities in the
European strategy for AI in science, in line with COM(2025) 724 final.
(19) In healthcare, those advancements should improve the accuracy of clinical decisions
and transform the pharmaceutical sector. In the automotive sector, they should support
the development, testing and deployment of innovative software platforms
contributing to the Union industrial leadership in software defined vehicles and
autonomous driving. The Cloud and AI Leadership Initiatives should also reduce
obstacles to test and deploy AI models, in particular within cities and regions
contributing to the development of Union leadership in software defined vehicles and
autonomous driving. Furthermore, Member States should facilitate the development,
testing and deployment of AI systems for autonomous driving, including through
cooperation with the Centres for AI, the automotive industry, suppliers, cities and
regions, with a view to enabling the safe and trustworthy deployment of AI-enabled
connected and autonomous mobility solutions across diverse European environments.
In manufacturing, the Commission should facilitate data pooling across industrial
sectors through trusted third parties to train specialised AI models, ensuring a
sufficient volume of training data, while strictly preserving intellectual property rights.
Secure and verifiable compute approaches should be explored to enable the use of AI
in sensitive contexts. In the defence sector, where AI has emerged as a disruptive
technology with significant impact on security and defence, the Cloud and AI
Leadership Initiatives could support the development of advanced capabilities in full
complementarity with, and without prejudice to, dedicated Union instruments in
support of the defence industry, including the European defence fund (‘EDF’) and the
European defence industry programme (‘EDIP’). In the space sector, digital advances
EN 19 EN
should transform the way space assets, services and data flows are operated and used,
and in the field of transport, digital advancements in aviation should transform the way
operations are managed, with AI harnessing decades of available mission, operational
and observation data. In the field of climate and environment, geospatial AI should be
developed, in particular by leveraging Earth observation data from the Copernicus
programme and the capabilities of the Union Space and connectivity programmes in
general (15). In agriculture, AI can turn data from sensors, satellites and farm
machinery into actionable insights for farmers. It can strengthen competitiveness and
resilience, for instance by improving yield forecasting, enabling early pest and disease
detection, optimising irrigation and fertiliser use, and supporting more sustainable
food production. As the deployment of AI in industrial contexts requires rigorous
validation in real-world environments, the Union should provide industrial actors with
cloud-based AI tools and testing environments.
(20) The Union should also foster the availability of highly secured computing
infrastructures for the training, testing and deployment of defence-related AI models
and systems.
(21) As AI agents have become increasingly capable and AI applications have become
more deeply embedded in real-world business scenarios, industry is rapidly evolving
towards a new paradigm that equips such systems with autonomous execution
capabilities. This transition to a new paradigm requires a robust technical framework
to ensure the safety, accuracy and legal compliance of those systems, given the
stringent engineering requirements pertaining on AI platforms. Accordingly, the Cloud
and AI Leadership Initiatives should aim to establish sovereign and secure AI
platforms dedicated to the large-scale deployment and orchestration of advanced AI
agents. Those platforms should be supported by innovative orchestration frameworks
that ensure transparency and accountability in multi-agent interactions. It is also
necessary to facilitate the development of rigorous testing and experimentation
methodologies of AI agents and orchestration to minimise unintended autonomous
behaviour.
(22) The Cloud and AI Leadership Initiatives should increase the development and
adoption of AI models and systems across the Union’s public sector. In particular, AI
models and systems should be used to support better decision-making, simplify
administrative procedures and reduce unnecessary burdens, in particular for critical
public domains such as healthcare where data reuse for AI models and tools should be
facilitated while ensuring security and data protection. To that end, it is appropriate to
take measures to enhance the quality of public sector data and promote the sharing and
reuse of training data and AI models across the Union public sector, thereby avoiding
fragmentation and enabling the scaling-up of successful, user-oriented solutions.
(23) The Cloud and AI Leadership Initiative should help accelerate the adoption of AI and
cloud computing on a large scale, including at regional and local level. Broad adoption
of AI in private and public sectors should be promoted through the network of Centres
for AI. Harnessing this network, complementary support measures should be
developed, to help achieve the target of digital transformation of businesses set in
15 Regulation (EU) 2021/696 of the European Parliament and of the Council of 28 April 2021 establishing
the Union Space Programme and the European Union Agency for the Space Programme and repealing
Regulations (EU) No 912/2010, (EU) No 1285/2013 and (EU) No 377/2014 and Decision No
541/2014/EU (OJ L 170, 12.5.2021, p. 69, ELI: http://data.europa.eu/eli/reg/2021/696/oj).
EN 20 EN
Decision (EU) 2022/2481 of the European Parliament and of the Council (16) (’the
Digital Decade Policy Programme 2030’). Furthermore, a dedicated curriculum on
cloud computing and AI skills should be developed to equip workers in both the public
and private sectors with advanced competencies to reduce dependence on non-EU
providers and develop next-generation capabilities, in line with the Union policies in
the field of education, training and skills including the Union of Skills communication
(17).Where relevant, the curriculum should be built on relevant European initiatives,
including the AI Skills Academy, the Centres for AI and the Interoperable Europe
Academy, and include the participation of stakeholders with the necessary expertise.
(24) The Cloud and AI Leadership Initiatives should also ensure the uptake of cloud
computing services provided by European cloud computing service providers across
the public and private sectors to ensure that cloud adoption is consistent with the
objective of strengthening the Union’s technological autonomy, particularly in sectors
such as healthcare and education which involve the processing of critical data. This
objective could leverage the outcomes of relevant European digital infrastructure
consortiums (‘EDICs’), including shared infrastructure, common standards and best
practices. The Cloud and AI Leadership Initiatives should support the establishment of
the European public-sector cloud federation (‘the EuroCloud Federation’) under this
Regulation to facilitate the sharing of secure and resilient public-sector data centre
services and cloud computing services. Moreover, the Cloud and AI Leadership
Initiatives should support procurement activities carried out by the Commission for
Union institutions, agencies and bodies (‘Union entities’), but also national contracting
authorities across the Union, as the procurement of digital services should not only
advance the digitalisation of public-sector bodies, but also enable them to utilise their
purchasing power and accelerate the adoption of resilient and secure digital solutions.
(25) Member States should establish Experience and acceleration centres for AI (‘Centres
for AI’) with a view to ensuring an appropriate coverage of their territory. Centres for
AI need to act as regional and local accelerators for the uptake and deployment of AI,
cloud computing and other advanced technologies across the Union, supporting SMEs,
SMCs and public sector bodies in their digital transformation. In order to facilitate the
integration of AI into strategic industrial sectors, Centres for AI should also establish
synergies with initiatives launched under the Data Union Strategy. By providing
expertise, testing, skills and innovation support, Centres for AI should reinforce the
competitiveness and resilience of the Union’s AI industrial base while strengthening
the innovation capacity and widespread deployment of AI and other advanced
technologies across the Union. The network should build on the network of European
digital innovation hubs. The network will collaborate closely with other initiatives
supporting the uptake of AI and other advanced technologies such as testing and
experimentation facilities and AI factories. It will be built on skills-related initiatives,
including the Digital large scale skills partnership under the Pact for skills and the
skills academies, including the AI Skills Academy.
(26) The implementation of the Cloud and AI Leadership Initiatives should be entrusted to
the Commission and Member States. The implementation of the Cloud and AI
16 Decision (EU) 2022/2481 of the European Parliament and of the Council of 14 December 2022
establishing the Digital Decade Policy Programme 2030, (OJ L 323, 19.12.2022, p. 4, ELI:
http://data.europa.eu/eli/dec/2022/2481/oj). 17 Communication from the Commission to the European Parliament, the Council, the European
Economic and Social Committee and the Committee of the Regions, ‘The Union of Skills’, 5.3.2025,
COM(2025) 90 final.
EN 21 EN
Leadership Initiatives should, where relevant, also be entrusted to any other structure
with the appropriate expertise and resources. In particular, without prejudice to the
next (2028-2034) multiannual financial framework, it should be possible to entrust the
implementation of the Cloud and AI Leadership Initiatives to joint undertakings, such
as the Smart Networks and Services Joint Undertaking (‘the JU’) established by
Council Regulation (EU) 2021/2085 (18) or the European High Performance
Computing Joint Undertaking (‘the EuroHPC JU’) established by Council Regulation
(EU) 2021/1173 (19) and, where relevant, their successor.
(27) The Cloud and AI Leadership Initiatives’ operational objectives should, in particular,
be implemented by setting ambitious, forward-looking objectives that aim to go
beyond the current state of the art in infrastructure development, cloud computing and
AI. The Cloud and AI Leadership Initiatives should therefore support major strategic
‘grand challenges’ focusing on the development and deployment of cutting-edge cloud
and AI technologies and infrastructure of key importance for the Union. Those grand
challenges should build on those established in Regulation (EU) 2026/XXX [Chips
Act 2.0] on a framework of measures for strengthening Europe’s semiconductor
ecosystem, aimed at enabling semiconductor technologies underpinning AI, cloud
computing, data centres and edge infrastructures.
(28) The Cloud and AI Leadership Initiatives may be supported by funding from Union
programmes and other instruments, in particular from Horizon Europe and the digital
Europe programme, as well as the InvestEU programme, in accordance with
Regulation (EU) 2021/694 (20), Regulation (EU) 2021/695 (21) and Regulation (EU)
2021/523 (22).Under the 2028-2034 multiannual financial framework, the Cloud and
AI Leadership Initiatives could continue receiving support under successive Union
programmes, subject to their adoption and in accordance with their respective legal
bases.
(29) In addition to receiving funding under Union programmes, the Cloud and AI
Leadership Initiatives may be supported by Member States through research,
development an innovation measures, in line with the applicable State aid rules,
ensuring that national policies and Union policy are mutually consistent, as well as
through private-sector investments. In particular, private-sector stakeholders should be
encouraged to take into consideration the Cloud and AI Leadership Initiatives when
developing their investment strategies for cloud computing and AI. In so doing,
private investments can help provide a coherent and coordinated investment pathway
18 Council Regulation (EU) 2021/2085 of 19 November 2021 establishing the Joint Undertakings under
Horizon Europe and repealing Regulations (EC) No 219/2007, (EU) No 557/2014, (EU) No 558/2014,
(EU) No 559/2014, (EU) No 560/2014, (EU) No 561/2014 and (EU) No 642/2014 (OJ L 427,
30.11.2021, p. 17, ELI: http://data.europa.eu/eli/reg/2021/2085/oj). 19 Council Regulation (EU) 2021/1173 of 13 July 2021 on establishing the European High Performance
Computing Joint Undertaking and repealing Regulation (EU) 2018/1488 (OJ L 256, 19.7.2021, p. 3,
ELI: http://data.europa.eu/eli/reg/2021/1173/oj). 20 Regulation (EU) 2021/694 of the European Parliament and of the Council of 29 April 2021 establishing
the Digital Europe Programme and repealing Decision (EU) 2015/2240 (OJ L 166, 11.5.2021, p. 1, ELI:
http://data.europa.eu/eli/reg/2021/694/oj). 21 Regulation (EU) 2021/695 of the European Parliament and of the Council of 28 April 2021 establishing
Horizon Europe – the Framework Programme for Research and Innovation, laying down its rules for
participation and dissemination, and repealing Regulations (EU) No 1290/2013 and (EU) No 1291/2013
(OJ L 170, 12.5.2021, p. 1, ELI: http://data.europa.eu/eli/reg/2021/695/oj). 22 Regulation (EU) 2021/523 of the European Parliament and of the Council of 24 March 2021
establishing the InvestEU Programme and amending Regulation (EU) 2015/1017 (OJ L 107, 26.3.2021,
p. 30, ELI: http://data.europa.eu/eli/reg/2021/523/oj).
EN 22 EN
aligned with the broader policy objectives and long-term implementation goals set out
by the Cloud and AI Leadership Initiatives.
(30) The Commission should receive advice from stakeholders with appropriate expertise
on the implementation of the Cloud and AI Leadership Initiatives. The Commission
should, in particular, foster cooperation with existing expert and advisory forums, such
as the Alliance for Industrial Data, Edge and Cloud, the Apply AI Alliance and the
Industrial Alliance for Semiconductors. The Alliance for Industrial Data, Edge and
Cloud should act as an exchange forum for various stakeholders in next-generation
edge and cloud technologies, including businesses, Member States’ representatives
and other relevant experts. It should help the Commission to design strategic
investment road maps to enable the next generation of highly secure, distributed,
interoperable and resource-efficient computing technologies. The Apply AI Alliance
should act as a coordination forum for AI stakeholders and policymakers to advance
the discussion on the potential of AI in strategic Union sectors. The Apply AI Alliance
should continue to help implement the objectives of the Apply AI Strategy, in
particular by organising industrial workshops, identify and assess new AI use cases in
strategic sectors and call for specific supporting policy actions. Alongside the Apply
AI Alliance, the AI Observatory should provide robust indicators to assess the impact
of AI in strategic sectors, monitor technological developments and trends, as well as
the changes it may bring to the labour market.
(31) The Cloud and AI Leadership Initiatives should enhance synergies with actions
currently supported by the Union and Member States, including under Horizon Europe
and the Digital Europe programme, as well as Council Regulation (EU) 2021/1173
and Regulation (EU) 2026/XXX [Chips Act 2.0] on a framework of measures for
strengthening Europe’s semiconductor ecosystem. The Commission and the Member
States should ensure consistency, complementarity and synergies between the Cloud
and AI Leadership Initiatives, and relevant national and regional strategies,
programmes and investment plans, including those implemented under national reform
programmes, smart specialisation strategies, recovery and resilience plans and other
national or regional funding instruments supporting the objectives of the Cloud and AI
Leadership Initiatives. Such coordination should aim to maximise the impact of public
investments, avoid duplication of funding, promote alignment of priorities across
governance levels, and facilitate the scaling-up and deployment of results across the
Union.
(32) In order to ensure that the policies of the Union and the Member States are mutually
consistent, Member States should adopt national strategies to help achieve the Union’s
objectives on the development of cloud and AI, in line with the AI continent action
plan and the Apply AI Strategy. The national strategies should notably include the ‘AI
first’ principle defined in the Apply AI Strategy, urging organisations to reflect on
their business processes, considering the needs of an opportunities offered by AI,
while taking into consideration the potential risks. The national cloud and AI strategies
should also be aligned with the associated digital targets set under Decision (EU)
2022/2481 (23), in particular on the adoption of cloud computing services, big data and
AI by at least 75% of Union enterprises for their business operations, and the
deployment of at least 10 000 climate-neutral highly secure edge nodes in the Union,
23 Decision (EU) 2022/2481 of the European Parliament and of the Council of 14 December 2022
establishing the Digital Decade Policy Programme 2030 (OJ L 323, 19.12.2022, p. 4, ELI
http://data.europa.eu/eli/dec/2022/2481/oj).
EN 23 EN
while ensuring low latency. The measures adopted under the national strategies could
inform the national digital decade strategic road maps (the ‘national road maps’).
(33) Where a Member State has already adopted a national strategy that adequately covers
the objectives set out in this Regulation, it should not be required to adopt another
strategy. However, if a Member State identifies gaps in its existing strategy in light of
those objectives, it should update it accordingly. The European Artificial Intelligence
Board established by the Artificial Intelligence Act (‘the AI Board’) plays a central
role in ensuring the consistent and effective implementation of Union AI policy. As
cloud computing underpins and enables AI, the AI Board should serve as a platform to
facilitate cooperation and coordination of AI adoption-related activities between the
Union and the Member States. In order to perform this task, the AI Board should have
the necessary expertise and appropriate participation.
(34) Given the unprecedented scale of resources required for frontier AI development, it is
necessary to set criteria for the designation of a project as a frontier AI priority project.
Such projects should support the development and scaling-up of frontier AI
technologies, notably in the sector of cybersecurity. In view of their technical
complexity and capital-intensive nature, the projects require a collaborative approach
at Union level. It is therefore appropriate to require them to involve broad participation
from entities across the Union, in particular through EDICs established pursuant
Decision (EU) 2022/2481 or any other legal structure capable of representing a
meaningful share of the Union’s interest.
(35) The allocation of sufficient AI computing resources to frontier AI priority projects
should be of strategic importance to the Union and the Member States. The Union
should match, on a proportional basis and within the limits of available European
high-performance computing (‘EuroHPC’) capacity, the AI computing resources
contributed or committed by the Member States to the designated frontier
AI priority projects. The Union and the Member States should also provide sufficient
compute time for AI industrial innovation, physical AI and public sector AI projects.
This is without prejudice to the rules and procedures laid down in Council Regulation
(EU) 2021/1173. The EuroHPC JU access policy should be accommodated to reflect
the allocation of such computing resources in an efficient, transparent and timely
manner without prejudice to the continuity of ongoing operations and the rights of
projects already benefiting from allocated EuroHPC AI computing resources.
(36) To achieve the Union’s AI ambitions, it is necessary to strengthen and invest in digital
infrastructures, including cloud and edge capacity enabling training, fine-tuning,
deployment and real-time operation. The deployment of data centres across the Union
is lagging and remains concentrated in a limited number of established hubs, creating
structural imbalances between Member States and inefficiencies such as higher costs,
increased latency for peripheral regions, unequal opportunities for businesses and
slower digital transformation. Increasing and geographically balancing data centre
capacities across the Union is also key to reducing dependencies on external
infrastructures, thus mitigating economic security risks and supporting the
competitiveness, resilience and sovereignty of the Union.
(37) Data centres are critical infrastructure for the Union, and can create substantial
economic value, including valuable investments and jobs, and may support innovation
ecosystems – especially if they are integrated with local needs and follow best
practices. If properly managed, the expansion of data centre capacity in the Union can
bring significant economic and strategic benefits, help modernise the energy system,
EN 24 EN
support clean energy growth and the sustainable use of energy. This depends on the
implementation of an adequate framework preventing any negative impacts, such as
energy supply stress, adverse environmental impacts and lost opportunities. Data
centre acceleration zones (‘acceleration zones’) should contribute to this objective by
enabling infrastructure deployment at scale and speed within a clear and streamlined
regulatory framework.
(38) Where capacity is being deployed on the territory of Member States, acceleration
zones should be designated where the development, expansion and modernisation of
data centres may be facilitated . The designation of such zones should help address the
Union capacity gap and increase the Union’s competitiveness, autonomy and
technological resilience, while ensuring compliance with applicable Union law,
including requirements relating to energy efficiency and environmental protection.
Sufficient and timely energy supply to the acceleration zones constitutes a
fundamental enabling condition for their effective deployment and for the
development of data centre capacity across the Union. Reliable and accurate
information on future energy demand contributes to cost-effective grid development.
Member States should therefore prepare an analysis for each acceleration zone,
identifying its current and future energy needs. Such analysis should serve the purpose
of providing information for the national grid planning thereby contributing to
purposeful anticipatory grid investments and faster energy connections for the
acceleration zone. When defining the scope, Member States should take into account
the availability of relevant transport and network infrastructure. The results of these
assessments should be reflected in national network development plans to adequately
capture future points of energy demand in upcoming grid planning. To ensure that
acceleration zones enable the right conditions for the deployment of capacity, Member
States should facilitate clear and efficient procedures for grid connection and flexible
connection agreements and should clarify the conditions for grid connection pursuant
to Directive (EU) 2019/944 to data centre operators. An upcoming legal proposal to
future-proof electricity bills in the EU will provide incentives to make an optimal and
cost-effective use of the grid infrastructure and incentivise system-friendly
consumption. Power purchasing agreements (‘PPAs’) are important instruments for
data centres as they provide long-term price stability, while enabling data centre
operators to procure clean electricity at scale, thereby supporting reliable operations
and the transition to a clean energy system. Member States should therefore promote
the uptake of PPAs, also in acceleration zones, by removing unjustified barriers and
disproportionate or discriminatory procedures or charges, with a view to providing
price predictability.
(39) When setting sustainability requirements for data centres deployed in data centre
acceleration zones, Member States should ensure that the key performance indicators
as defined in Commission Delegated Regulation (EU) 2024/1364 (24) in accordance
with Directive (EU) 2023/1791 of the European Parliament and of the Council (25) are
used. The objective is to ensure consistent environmental standards, increase energy
efficiency and support the Union’s broader climate, environmental and sustainability
goals in acceleration zones. Furthermore, to prevent speculative reservation of
24 Commission Delegated Regulation (EU) 2024/1364 of 14 March 2024 on the first phase of the
establishment of a common Union rating scheme for data centres (OJ L 2024/1364, 17.5.2024, ELI:
http://data.europa.eu/eli/reg_del/2024/1364/oj). 25 Directive (EU) 2023/1791 of the European Parliament and of the Council of 13 September 2023 on
energy efficiency and amending Regulation (EU) 2023/955 (recast) (OJ L 231, 20.9.2023, p. 1, ELI:
http://data.europa.eu/eli/dir/2023/1791/oj).
EN 25 EN
resources in acceleration zones, ensuring fair, reasonable and non-discriminatory
access that preserves effective competition and supports the timely and efficient
development of acceleration zones, Member States should ensure that the allocation
and use of resources within those zones takes place on fair, reasonable and non-
discriminatory terms and does not give rise to any speculative reservation or
foreclosure practices capable of impeding effective competition or the effective
development or use of those zones.
(40) To facilitate and accelerate the deployment of data centre projects in acceleration
zones, Member States should designate single information points or where possible,
upgrade or integrate with those designated pursuant to Regulation (EU) 2024/1309 of
the European Parliament and of the Council (26).
(41) Regulation (EU) 202X/XXX [on speeding-up environmental assessments] (27)
establishes a common acceleration framework for environmental assessments to boost
the Union’s roll-out of key technologies, reduce dependencies and increase
competitiveness. Procedures linked to environmental assessments should be
accelerated and streamlined for plans, programmes and projects across all sectors of
the economy while maintaining high levels of protection of human health and the
environment. Some sectors may, however, require faster environmental assessments.
To ensure the coherence of the legal framework for environmental assessments, while
accommodating the need for accelerated deployment in certain strategic sectors,
Regulation (EU) 202X/XXX [on speeding-up environmental assessments] establishes
a dedicated toolbox. Given their role in ensuring the achievement of the Union’s
climate and environmental objectives through their contribution to improving energy
efficiency, enabling clean energy integration, and providing the infrastructure needed
for smarter grids, transport systems, and low-carbon technologies, and their
contribution to the Union’s resilience and economic security by ensuring reliable
infrastructure in the Union to protect critical services and strengthening the Union’s
capacity to operate independently, data centre projects deployed in acceleration zones
should be considered strategic projects within the meaning of Regulation (EU)
202X/XXX [on speeding-up environmental assessments] and therefore benefit from
the dedicated toolbox established under that Regulation. Member States should
establish an aggregated baseline permit reflecting the specific characteristics of each
identified acceleration zone. That aggregated baseline permit issued by public
authorities should cover the permits commonly required for such activities within the
area, excluding the grid connection permits.
(42) It should be possible for the Commission to designate as strategic projects data centre
projects that significantly contribute to the Union’s digital and energy sectors and that
meet clear criteria. Considering the importance of the data centre strategic projects,
Member States may, without prejudice to Articles 107 and 108 TFEU, apply support
measures in a proportionate manner to those projects. When planning such support
measures, Member States should, where applicable, make use of the relevant
frameworks for providing public support. Strategic projects should address a market
26 Regulation (EU) 2024/1309 of the European Parliament and of the Council of 29 April 2024 on
measures to reduce the cost of deploying gigabit electronic communications networks, amending
Regulation (EU) 2015/2120 and repealing Directive 2014/61/EU (Gigabit Infrastructure Act) (OJ
L 2024/1309, 8.5.2024, ELI: http://data.europa.eu/eli/reg/2024/1309/oj). 27 Add reference to the Regulation on speeding-up environmental assessments once adopted. See
Commission proposal COM(2025) 984 final, https://eur-lex.europa.eu/legal-
content/EN/TXT/?uri=celex:52025PC0984.
EN 26 EN
failure in a proportionate manner, without duplicating or crowding out private
financing, while ensuring clear Union added value.
(43) Data centre strategic projects should be granted support from Union programmes,
funds and financial instruments, in accordance with the objectives set out in the
regulation establishing those funds and programmes and without prejudice to the next
(2028-2034) multiannual financial framework. In particular, those strategic projects
should be granted the competitiveness seal where they fulfil the conditions set out in
Regulation (EU) 2026/XXX [on establishing the European Competitiveness Fund]
(ECF’) (28), as high-quality projects that contribute to the objective of the European
Competitiveness Fund.
(44) To foster the strategic deployment of data centre capacity across the Union, the
Commission should monitor the available compute capacity and the volume of
demand for data centre capacity and identify the size of the capacity gap across the
Union. Such monitoring may be used by the Commission to inform its possible
recommendations. To guide Member States in accelerating the deployment of data
centre capacity, the Commission may recommend, where appropriate, measures to
address the identified Union capacity gap. In accordance with the Digital Decade
Policy Programme 2030, the Commission should also review the digital decade targets
to reflect the technical, economic or societal developments and the evolution of the
Union’s priorities in that regard.
(45) This Regulation should apply to Union institutions, bodies, offices and agencies
(‘Union entities’) when carrying out procedures for the procurement of cloud
computing services and AI systems falling within the scope of this Regulation.
(46) The Union still remains critically dependent on a limited number of cloud computing
service providers subject to the control of third countries or legal entities established in
third-countries. This exposes the Union to critical strategic dependencies and
concentration risks, including vulnerabilities arising from the extraterritorial
application of third-country laws, potential disruptions affecting the continuity, quality
and resilience of cloud computing services, reduced control and oversight over
personal and non-personal data and infrastructure, and the risk of undue economic or
political influence being exercised through the control by third countries or legal
entities established int third-countries of cloud computing services. Against this
background, the ability of the Union and its Member States to retain control over
infrastructure, data, assets and technology systems under Union and national
jurisdiction has become an imperative policy objective.
(47) Existing Union law addresses cybersecurity, data protection, interoperability and data
portability requirements which cloud computing services are subject to. However,
there is no cross-cutting Union regulatory framework establishing a harmonised
understanding of what constitutes a trusted cloud computing service for mitigating
such risks. Some Member States have developed or are in the process of developing
national approaches to identifying national sovereign services. However, national
measures do not adequately address the cross-border issues related to the Union’s lack
of sovereignty in the cloud computing ecosystem and risk fragmenting the Union
internal market and undermining common goals of autonomy and sovereignty.
28 Add reference to Regulation on establishing the European Competitiveness Fund once adopted. See
Commission proposal, COM(2025) 555 final/2, https://eur-lex.europa.eu/legal-
content/EN/TXT/?uri=celex:52025PC0555R%2801%29.
EN 27 EN
(48) Cloud computing service providers have launched tailored versions of their service
offerings in response to the Union’s growing concerns over sovereignty. However,
those versions do not address the core sovereignty issues allowing for the
extraterritorial reach of third-country laws and the possible degradation or disruption
of the service. Consequently, the Union will not ensure autonomy or control over its
data, assets and digital infrastructure. The current framework therefore needs to be
complemented by targeted actions at Union level by introducing a harmonised
mechanism that can strengthen the Union’s long-term strategy for technological
autonomy, control and resilience in the cloud and AI ecosystem.
(49) Against this background, the significant increase in public order concerns – including,
for example, economic security risks – requires effective and coherent implementation
of safeguards for activities supported by the Union budget. In the context of Union
entities, Article 136 of Regulation (EU, Euratom) 2024/2509 (29), sets out the scope,
rules and procedures for identifying and implementing sensitive public procurement
procedures.
(50) To protect public order, it is therefore necessary to specify the conditions that Union
and Member States’ contracting authorities should use in public procurement
procedures of cloud computing services. The consideration of possible exposure to
risk is fundamental when selecting appropriate mitigation measures to preserve the
public order of the Union and Member States. The Union and Member States being
critically dependent on a limited number of cloud computing service providers subject
to the control of a third country or a legal entity established in a third-country may
lead to risks such as misuse (i.e. manipulation, remote access and control, sabotage,
weaponisation), access to information (i.e. access to sensitive information,
unauthorised communication, technology leakage, data manipulation or exfiltration,
espionage) and dependency vulnerabilities (i.e. political and/or economic coercion, for
example by using vendor or technology lock-ins, embargos or sanctions, monopoly
pricing damaging the financial interest of the Union and Member States).
(51) To address those risks and provide for the appropriate mitigation measures, it is
necessary to establish a Union cloud computing sovereignty framework determining
criteria for trusted cloud computing services. To cater for the nuanced and layered
nature of sovereignty, the framework should provide for four different levels of trusted
offers (‘Union assurance levels’).
(52) The Union assurance levels should provide for a proportionate framework to ensure
that public order is preserved by maintaining control and agency by public-sector
bodies. Most public services would not require the highest levels of assurance. In
some specific cases Union assurance levels 3 or 4 may be considered necessary and
proportionate in preserving public order. The risk assessment to be performed by
Member States and Union entities ensures that the principles of proportionality and
subsidiarity are complied with, by assessing the specific cases in which protection of
public order requires the highest level of assurance.
(53) It is important that national competent authorities of establishment of the cloud
computing service provider can assess whether cloud computing service providers
aiming to provide their cloud computing services to Union entities and public sector
bodies offer the appropriate assurance level. A mechanism for recognition of cloud
29 Regulation (EU, Euratom) 2024/2509 of the European Parliament and of the Council of 23 September
2024 on the financial rules applicable to the general budget of the Union (recast) (OJ L, 2024/2509,
26.9.2024, ELI: http://data.europa.eu/eli/reg/2024/2509/oj).
EN 28 EN
computing services that either have an EU statement of conformity against Union
assurance level 1 or have received a ‘positive’ audit opinion and audit report by an
auditing organisation has been established. Once the mechanism of recognition has
been positively concluded, the cloud computing service is recognised across the Union
as offering the appropriate Union assurance level. In the interest of clarity, simplicity
and effectiveness, the powers to supervise and enforce the obligations relating to the
cloud computing sovereignty framework should be conferred to the competent
authorities in the Member State where the main establishment of the cloud computing
service provider is located.
(54) In order to demonstrate compliance with Union assurance level 1, cloud computing
service providers should have sole responsibility for carrying out conformity self-
assessments by applying the relevant criteria for that Union assurance level. Such self-
assessments should be based on documented evidence, internal control procedures and
continuous monitoring that are sufficient to demonstrate that the applicable criteria
have been fulfilled.
(55) Independent audits are an important tool for monitoring the compliance of cloud
compuring services provided by cloud computing service. Given the need to ensure
that the applicable criteria for Union assurance levels 2, 3 or 4 are verified by third-
party independent experts, cloud computing service providers should be accountable,
through independent auditing, for their compliance with the criteria set out by this
Regulation. Cloud computing service providers should be free to select the auditing
organisation of their choice as long as the auditing organisation demonstrates the
necessary independence and compliance with the requirements in this Regulation. To
ensure that audits are carried out in an effective, efficient and timely manner, cloud
computing service providers should provide the necessary cooperation and assistance
to the organisations carrying out the audits, including by giving the auditing
organisations access to all relevant data and premises necessary to perform the audit
properly and answering oral or written questions. Auditing organisations should also
be able to make use of other sources of objective information. Cloud computing
service providers should not undermine the performance of the audit. Audits should be
performed in accordance with best industry practices and high professional ethics and
objectivity, with due regard for auditing standards and codes of practice. Auditing
organisations should guarantee the confidentiality, security and integrity of the
information, such as trade secrets, that they obtain when performing their tasks. That
guarantee should not be a means to circumvent the applicability of audit obligations in
this Regulation. Auditing organisations should have the necessary expertise in risk
management and technical competence to audit cloud computing services. They
should comply with core independence requirements for prohibited non-auditing
services, firm rotation and non-contingent fees. If their independence or technical
competence of auditing organisations is not beyond doubt, they should abstain or
resign from the audit engagement.
(56) The audit report should be substantiated to give a meaningful account of the activities
undertaken and the conclusions reached during the audit. It should help provide
information for, and where appropriate suggest improvements to, the measures taken
by the cloud computing service providers to comply with the applicable criteria and
their obligations under this Regulation. The audit report should include an audit
opinion based on the conclusions drawn from the audit evidence obtained. A ‘positive
opinion’ should be given where all evidence shows that the provider complies with the
audit criteria and obligations set out by this Regulation. A ‘negative opinion’ should
EN 29 EN
be given where the auditing organisations considers that the provider does not comply
with the criteria set out in this Regulation. Where the audit opinion cannot reach a
conclusion on specific aspects that fall within the scope of the audit, an explanation of
the reasons why this was not possible should be included in the audit opinion. Where
applicable, the report should include a description of specific points that could not be
audited, and an explanation as to why they could not.
(57) The establishment of a central repository of recognised Union-assured cloud
computing services is necessary to facilitate the secure and efficient storage, access
and exchange of relevant information between public sector customers of services
offering Union assurance levels, auditing organisations, competent authorities and the
Commission.
(58) To ensure the continued accuracy and reliability of the status of cloud computing
services as offering Union assurance levels pursuant to the cloud sovereignty
framework, providers should be required to promptly report any relevant information
or material changes in circumstances to the auditing organisation and the competent
authorities of establishment. That information should enable the auditing organisation
to reassess, amend or withdraw the audit report and opinion where necessary, with
notifications subsequently being sent to the relevant competent authority of
establishment for it to review its recognition of the cloud computing service as
offering a certain Union assurance level. The information should also enable the same
for the relevant competent authority regarding its recognition of the cloud computing
service.
(59) To ensure effective and consistent application of the cloud sovereignty framework,
Member States should designate one or more competent authorities responsible for
recognising the auditing procedure and framework and the supervision of recognised
cloud computing service providers. Those authorities should be granted the necessary
powers, resources, expertise and technical means to carry out their tasks in an
effective, impartial and independent manner. Member States should ensure that the
responsibilities of those authorities are clearly set out and that they cooperate closely
with each other, with other relevant national authorities, including Data Protection
Authorities and Cybersecurity Authorities, and with the Commission, where
appropriate, to ensure consistent supervision and facilitate the exchange of relevant
information and best practices across the Union. It should be possible for Member
States to designate one or more existing authorities as competent authorities.
(60) The provision of mutual assistance between competent authorities is essential to
ensure the effective supervision and enforcement of this Regulation across the Union
borders, including through the timely exchange of information, coordination of
investigative measures and support in the execution of tasks within the Union.
Furthermore, effective enforcement requires robust cross-border cooperation between
competent authorities to ensure the consistent application of this Regulation across
Member States and the timely sharing of relevant information to address systemic
risks within the Union
(61) The Union’s objective of strengthening its autonomy should be pursued in a manner
that remains open, cooperative and consistent with the Union’s international
commitments and partnerships. The policy objectives pursued through Union
assurance levels 1, 2, and 3 should therefore be understood as the Union’s capacity to
act autonomously where necessary, while remaining engaged with its international
partners and fostering mutually beneficial cooperation. Against this background, the
EN 30 EN
Commission may decide, for Union assurance level 3, that a cloud computing service
subject to the control of a third country or a legal entity established in a third-country
can still be audited against the audit criteria where the third country has implemented
specific safeguards that ensure that there is no risk of unauthorised access to Union
data or possible disruption of service quality or continuity. The Commission should
assess whether the third country is covered by an adequacy decision adopted pursuant
to Article 45 of Regulation (EU) 2016/679. In particular, it should be determined
whether the adequacy decision applies generally to the third country as a whole or is
limited to specific sectors or certified organisations. It should be further assessed
whether the scope of the adequacy decision extends to the specific processing
activities that are carried out in the context of the service provision, or whether
transfers remain subject to the requirements to implement appropriate safeguards.
(62) To ensure a coherent and risk-based approach to the autonomy of the Union, Member
States and the Union entities should carry out one or more risk assessments to
determine public-sector activities that concerns public order. The risk assessment
should determine which Union assurance level is appropriate for the activities, due to
their importance in preserving public order in sectors falling under Directive (EU)
2022/2555 and in the areas of national security, internal security, external border
management, defence, justice or law enforcement, including the prevention,
investigation, detection and prosecution of criminal offence. To ensure consistent
application of this Regulation and preserve the integrity of the digital single market,
the Commission will provide guidance to assist Member States in carrying out their
risk assessments. Whereas the determination of the level of sensitivity of information
that may be hosted in a cloud computing service that offers a Union assurance level
lies within the competence and discretion of the Member States, to provide
consistency across the Union, Union assurance levels 3 and 4 should allow for the
secure hosting of EU classified information.
(63) In their risk assessments, Union entities and Member State shall assess the sensitivity,
criticality and magnitude of personal and non-personal data processed in cloud
environment. Such processing may include ordinary business information,
commercially sensitive information, operationally critical data, personal data within
the meaning of Regulation (EU) 2016/679, and data that is subject to sector-specific
obligations under Union law, including Directive (EU) 2022/2555 and Regulation
(EU) 2022/2554. The guidance by the Commission allows for a degree of flexibility to
Union entities and Member States in determining the appropriate Union assurance
levels and the categories of information and users for which such levels are
appropriate. At the same time, divergent national approaches to the classification and
mapping of data sensitivity and assurance requirements may undermine the consistent
application of the sovereignty framework across the Union. To ensure harmonised
implementation across the Union, the Commission should, in cooperation with
relevant authorities, provide centrally coordinated guidance on the mapping between
Union assurance levels and categories of information, taking into account the
sensitivity, criticality and magnitude of the data processed by the cloud environment,
the systematic importance of the activities of the contracting authorities, and the
applicable obligations arising from Union law. Furthermore, the criteria under the
Union assurance levels should not affect obligations of cross-border cooperation
provided by Union law. Where cloud computing services are used to process personal
data, Regulation (EU) 2016/679 provides for an obligation to agree on organisational
and technical measures to comply with that Regulation. Where the cloud computing
service provider relies on subcontractors in the provision of the services, the same
EN 31 EN
agreements apply to the subcontractors. Where specific technical and organisational
measures should be implemented pursuant to this Regulation to ensure that personal
data are processed in line with this Regulation, such specific measures could be
foreseen in the mandatory agreements pursuant to Regulation (EU) 2016/679 and
could be relied on to demonstrate that the necessary Union assurance levels are met.
(64) The free flow of data within the Union is an essential condition for the proper
functioning of the internal market. To promote the free flow of data within the Union
and to support the functioning of the internal market, it is appropriate that Member
States ensure that data is not confined to the territory of a single Member State and
may be stored and processed across the Union without unjustified restrictions.The
Union maintains an open and non-discriminatory framework for market access, in
accordance with the TFEU and subject to international commitments. Those include
commitments under the World Trade Organization (WTO) Agreement on Government
Procurement (GPA), as well as bilateral trade agreements. Nevertheless, where
necessary and in duly justified circumstances, the Union retains the right, in
accordance with Article III:2(a) of the WTO GPA, to adopt or maintain measures
necessary to protect public morals, order or safety, allowing for necessary and
proportionate restrictions on access to public procurement procedures. Indeed,
identifying and addressing risks such as critical dependencies, unauthorised access to
Union data, technology leakage, sabotage and espionage by third-country actors is
fundamental for preserving Union public order. Preserving the protection of public
order of the Union and its Member States requires a prudent but firm political, legal
and operational response for both national and Union-level award procedures, in full
respect of international commitments. To protect and preserve the public order of the
Union and its Member States, contracting authorities whose activities have been
identified on the basis of the Member State risk assessment should therefore procure
only the cloud computing service providing the appropriate level of assurance between
levels 2 and 4. A minimum assurance level, by mandating Union assurance level 1
across the Union, is necessary to establish a consistent baseline of safeguards for the
public sector, thereby reducing vulnerabilities in the public sector to third country
access to Union data and disruption of services.
(65) To enhance resilience and limit dependency on a single cloud computing service
provider, Union entities and Member States should, as part of their public procurement
procedures, consider whether a multi-vendor or multi-cloud strategy may be
appropriate. The decision to adopt and implement a multi-cloud architecture should be
based on a context-specific risk assessment. The assessment should identify any
relevant operational, regulatory or resilience-related circumstances that would support
the adoption of a multi-vendor or multi-cloud strategy.
(66) Public procurement frequently serves as a primary signal of market direction.
Requirements imposed by or on public authorities to adopt specific assurance levels
offered by cloud computing services tend to be mirrored by private-sector entities
operating in regulated industries, with subsequent spillover effects contributing to
broader market realignment over time. Those developments underscore the importance
of private-sector entities operating in the sectors in Annex I to Directive (EU)
2022/2555 to be able to carry out the same assessments as those carried out by Union
entities and Member States.
(67) In public procurement procedures for cloud computing services and AI systems,
contracting authorities should include clear European added value as part of the
quality evaluation of the tender. Such added value should consist in helping reinforce
EN 32 EN
the digital supply chain in the Union; integrating Union technologies; conducting the
innovation required to deliver the service in the Union; and delivering the service
using hardware components designed or manufactured in the Union. The criterion
relating to European added value should not be decisive for award of the contract and
should be applied in a manner that preserves the primacy of technical and financial
criteria directly connected to the performance requirements. For this purpose,
contracting authorities could consider a maximum weighting of 15 out of 120 points to
be allocated to European added value within the overall evaluation methodology,
ensuring that it remains proportionate and subordinate to the core contract award
criteria.
(68) Innovation procurement in cloud computing services and AI systems is essential to
foster technological development, strengthen digital resilience and competitiveness
and enable public authorities to benefit from secure, efficient and trustworthy digital
solutions that evolve with rapidly changing technological and societal needs. Member
States should therefore aspire to award at least 25% of relevant cloud and AI
procurement innovation procedures to SMEs. To that end, Member States should
actively report on their uptake of innovative cloud computing services and AI systems
to the Commission.
(69) In the joint declarations ‘Building the next-generation cloud for businesses and the
public sector in the EU’ of 15 October 2020 (30) and the Berlin Declaration on Digital
Society and Value-Based Digital Government of 8 December 2020 (31), Member
States expressed great interest in determining a common approach to federating cloud
capacities by interconnecting cloud computing infrastructures across the Union and
working towards the deployment of a secure and interoperable European public-sector
cloud federation. In its conclusions on the future of EU digital policy of 21 May 2024
(32), the Council of the European Union confirmed the relevance and importance of
achieving this common approach by inviting the Commission to continue its support
for the development of interoperable public digital services and the cross-border
interconnection of public administrations’ infrastructure, including cloud and edge
infrastructures, to increase their resilience, efficiency and sustainability. To achieve
this common approach, it is necessary to establish the European public-sector cloud
federation (‘the EuroCloud Federation’). The EuroCloud Federation should bring
together national and European cloud initiatives that provide highly trusted and secure
public-sector cloud capabilities and facilitate the sharing of such capabilities between
Union entities and public-sector bodies. When evaluating this Regulation, the
Commission may, at a later stage, assess the possibility for acceding countries,
candidate countries and potential candidates, as well as for international organisations
whose headquarters are in the Union, to participate in the EuroCloud Federation, in
accordance with Union law.
(70) The members of the EuroCloud Federation should comply with specific requirements
to avoid any distortion of competition in relation to private economic operators by
placing a private provider of services in a position of advantage over its competitors.
30 Joint declaration, ‘Building the next generation cloud for businesses and the public sector in the EU’,
15.10.2020, https://digital-strategy.ec.europa.eu/en/news/towards-next-generation-cloud-europe. 31 Berlin Declaration on Digital Society and Value-based Digital Government, 8.12.2020, https://digital-
strategy.ec.europa.eu/en/news/berlin-declaration-digital-society-and-value-based-digital-government. 32 Council conclusions on the Future of EU Digital Policy, 21.5.2024,
https://data.consilium.europa.eu/doc/document/ST-9957-2024-INIT/en/pdf.
EN 33 EN
(71) Participation within the EuroCloud Federation should be limited to public entities,
without direct participation of a private party. In this regard, direct private
participation should be excluded where the sharing entity, either directly or indirectly
through an intermediate legal entity, owns the hardware, as defined in Article 3, point
(5), of Regulation (EU) 2024/2847 of the European Parliament and of the Council (33),
over which the service is made available, and provides that service. The sharing entity
should be deemed to exercise control over that intermediate legal entity where the
following cumulative conditions are fulfilled. First, the sharing entity should exercise
a decisive influence over both strategic objectives and significant decisions of the
intermediate legal entity that owns the hardware and provides the services. Second,
there should not be any direct private capital participation in that intermediate legal
entity. Third, more than 80% of the activities of the intermediate legal entity should be
carried out in the performance of tasks entrusted to it by the sharing entity.
(72) To ensure effective, secure and resilient provision of services, the sharing entity
should put in place appropriate technical, operational and organisational measures.
This should include, in particular, policies on risk analysis and information system
security, including access control policies, policies on incident handling and business
continuity and policies supporting interoperability and connectivity.
(73) Finally, the sharing of data centre services and cloud computing services within the
EuroCloud Federation should be anchored in a public-sector cooperation. Such
cooperation should be governed solely by considerations of public interest, and should
not entail any form of consideration in exchange for another. In particular, the sharing
of services within the EuroCloud Federation should be free of charge, except where
the charges are limited strictly to what is necessary and proportionate to recover the
costs incurred by the sharing entity for the beneficiary using entity. Those costs should
be limited to the additional costs incurred in the sharing of capacity, including for
allocating and isolating resources, managing access, enabling the integration and
interoperability of resources, ensuring compliance with the applicable requirements
under Union law and managing the sharing relationship. The fees levied by the sharing
entity to recover those costs should not be deemed as a consideration for the provision
of a service and should not constitute a pecuniary interest or public contract within the
meaning of Directive 2014/24/EU of the European Parliament and of the Council (34)
and Regulation (EU, Euratom) 2024/2509. Under those conditions, the sharing of
public-sector data centre services and cloud computing services within the EuroCloud
Federation should not fall under Union public procurement rules.
(74) Contracting authorities of Member States frequently encounter significant difficulties
in procuring digital solutions such as data centre services, cloud computing services,
software and AI systems. Limited financial resources, reduced purchasing power,
insufficient technical or procurement expertise prevent public-sector bodies from
effectively accessing such services. Procurement activities conducted by the
Commission in concertation with the contracting authorities of Member States can
play a decisive role in harnessing collective purchasing power and ensuring access to
33 Regulation (EU) 2024/2847 of the European Parliament and of the Council of 23 October 2024 on
horizontal cybersecurity requirements for products with digital elements and amending Regulations
(EU) No 168/2013 and (EU) 2019/1020 and Directive (EU) 2020/1828 (Cyber Resilience Act) (OJ L,
2024/2847, 20.11.2024, ELI: http://data.europa.eu/eli/reg/2024/2847/oj). 34 Directive 2014/24/EU of the European Parliament and of the Council of 26 February 2014 on public
procurement and repealing Directive 2004/18/EC (OJ L 94, 28.3.2014, p. 65, ELI:
http://data.europa.eu/eli/dir/2014/24/oj).
EN 34 EN
those services and supplies on favourable terms. The Commission may already carry
out joint procurement procedures with contracting authorities of Member States
pursuant to Article 168(2) of Regulation (EU, Euratom) 2024/2509. The Commission
may also already act as a wholesaler by buiying, stocking, and reselling or donating
supplies and services to partner organisations selected by it. In order to enable those
entities to fully exploit the potential of the internal market, in particular as regards
economies of scale and benefit sharing, the possibilities for the Commission to act as a
central purchasing body should be extended to contracting authorities of Member
States and to partner organisations selected by the Commission, as a form of
procurement additional to those in Article 168(3) of the Regulation (EU, Euratom)
2024/2509. It should be possible for the Commission to use all the procurement
procedures available for the benefit of contracting authorities of Member States and of
partner organisations selected by the Commission and to include purchasing activities
conducted by the Commission for Union institutions, bodies and agencies. Contracting
authorities of Member States, partner organisations and Union institutions, agencies
and bodies should be considered as ‘participating entities’ in those procurement
procedures.
(75) In order to increase flexibility and administrative efficiency, it is appropriate to
provide, in a derogation from the Regulation (EU, Euratom) 2024/2509, for the
possibility of adding participating entities during the lifespan of a dynamic purchasing
system. Under that derogation, participating entities that have acceded to the
agreement after the establishment of a dynamic purchasing system should be permitted
– subject to the prior approval of the Commission – to join that system at any point
during its period of validity, before any future invitation to tender is issued. That
possibility should be strictly limited to newly acceding participating entities and
should not affect the rights and obligations of entities already participating in the
system or the integrity of procurement procedures already concluded or ongoing.
(76) The Commission should present to the Member States a draft agreement setting out
the practical arrangements governing the procurement activities. Given the potentially
large number of participating entities involved, that draft agreement should be
negotiated and initially concluded between the Commission and the Member States
willing to participate. It should lay down the conditions and procedure by which Union
entities as well as other Member States, contracting authorities of Member States and
partner organisations selected by the Commission may subsequently accede to and
benefit from it. The agreement will enter into force in accordance with its provisions,
subject to the approval of at least two Member States. Although the Commission
should, as far as possible, aim to address the common needs of participating entities,
this should not be construed as an obligation to satisfy needs that are specific to a
limited number of them.
(77) Where participating entities enter into an agreement for the provision of central
purchasing activities, including ancillary purchasing activities, they should not apply
the public procurement procedures provided for in applicable Union law, in
accordance with Directive 2014/24/EU and Regulation (EU, Euratom) 2024/2509.
Any contracting authority from Member States entering into such an agreement for the
purpose of organising central purchasing activities should be deemed to fulfil its
obligations pursuant to the national law transposing Directive 2014/24/EU if it
purchases works, supplies or services from a contracting authority responsible for the
procurement procedure.
EN 35 EN
(78) The agreement should establish a steering committee composed of the Commission
and representatives of Member States. The committee is responsible for strategic
oversight of the procurement activities, including the strategic guidance of the agenda
of public procurement activities and of each procurement procedure. The steering
committee should also oversee the involvement of participating entities. The steering
committee should not be responsible for the operations of procurement activities,
which should remain the responsibility of the Commission, including the setting of
fees. The steering committee should provide for the most adequate method to select
additional representatives from Union entities, from contracting authorities from
Member States and from partner organisations selected by the Commission.
(79) The rules governing responsibility and the applicable public procurement framework
between the Commission, acting as a central purchasing body, and the participating
entities procuring through it should be clarified in the agreement. Where a
participating entity conducts certain parts of the procurement procedure autonomously,
it should remain solely responsible for those stages of autonomous conduct. A
contracting authority which acts as a central purchasing body and has acquired
services or supplies through the Commission should be permitted to offer those
services to other contracting authorities without applying the public procurement
procedures provided for under applicable Union law. In that case, the contracting
authority is bound to comply with the initial contractual provisions in any subsequent
contract.
(80) In order to ensure that the necessary resources remain available, the participating
entities will contribute to the costs incurred in the procurement procedures and any
ancillary activity. To this end, the Commission should be entitled to charge fees to the
participating entities. Those fees should be set at a level sufficient in principle to cover
all the direct and indirect costs incurred by the Commission in connection with the
procurement activities, including any ancillary services. Those fees should be
established in accordance with practices of comparable procurement frameworks.
Initial establishment costs may be borne by the general budget of the Union and
reimbursed by the participating entities over a set period. Revenues generated by the
fees should constitute internal assigned revenues within the meaning of Article
21(3)(a), of the Regulation (EU, Euratom) 2024/2509.
(81) Open source plays an important role in ensuring transparency, security and efficiency
in the use of digital technologies by the public sector. Access to the source code
enables auditability, fosters collaboration and reuse and reduces dependency on a
single vendor, thereby limiting the risk of vendor lock-in. Promoting the use of open
source is therefore essential to support innovation, ensure better value for public
expenditure and strengthen the Union’s digital autonomy. In that context, the choice of
cloud computing services or software has significant implications not only for cost-
efficiency, but also for security, interoperability, accountability and technological
autonomy
(82) To ensure the efficient, transparent and interoperable use of digital technologies across
the Union’s public sector, it is necessary for public administrations to promote open
standards and components released under an open source licence when building their
cloud and AI ecosystem or stack.
(83) An increasing number of Union entities and public-sector bodies are sharing software
developed by or for them and making it available for reuse under an open-source
licence. This may be considered to be in the public interest and may maximise the
EN 36 EN
value of public expenditure, reduce duplication costs and foster innovation across the
Union. However, software is often made available and accessible in different
repositories or catalogues, hampering searchability, discoverability and, ultimately,
reuse. It is therefore necessary to require Union entities and public-sector bodies that
voluntarily decide to make software available for reuse to do so in a catalogue or
repository that is connected to EU Open Source Solutions Catalogue (‘the EU OSS
Catalogue’). The OSS Catalogue should serve as a centralised catalogue for any public
administration to search and access software made available for reuse by Union
entities and public sector bodies. Hosting the EU Open Source Solutions Catalogue on
the Interoperable Europe portal referred to in Article 8 of Regulation (EU) 2024/903
(35) will ensure that solutions can be easily linked to further relevant information and
training.
(84) In order to ensure effective and consistent implementation across the Union of the
obligations to conduct an open-source assessment and to make software available for
reuse, it is necessary to set up a network of open-source programme offices (‘the
OSPO network’) bringing together the relevant structures within Union entities and
Member States. This OSPO network should promote coordination between open-
source programme offices established at local, regional or national level and by Union
entities. The OSPO network should facilitate the exchange of information and best
practices.
(85) In order to take account of technological development and maintain an efficient
framework of measures for strengthening the cloud and AI ecosystem at Union level,
the power to adopt acts in accordance with Article 290 TFEU should be delegated to
the Commission in respect of: amending Annex I to reflect relevant market and
technological developments regarding the Cloud and AI Leadership Initiatives and
amending Annex II to update the criteria for Union assurance levels; supplementing
this Regulation by laying down detailed rules for the performance of audits; amending
Annex III; specifying a Union assurance level for a contracting authority; and
requiring an impact assessment and risk mitigation measures for private companies
operating in sectors of high criticality.
(86) When adopting delegated acts under this Regulation, it is of particular importance that
the Commission carries out appropriate consultations during its preparatory work,
including at expert level, and that those consultations are conducted in accordance
with the principles set out in the Interinstitutional Agreement of 13 April 2016 on
Better Law-Making (36). In particular, to ensure equal participation in the preparation
of delegated acts, the European Parliament and the Council should receive all
documents at the same time as Member States’ experts, and their experts should
always have access to meetings of Commission expert groups dealing with the
preparation of delegated acts.
(87) In order to ensure uniform conditions for the implementation of this Regulation,
implementing powers should be conferred on the Commission. Those powers should
35 Regulation (EU) 2024/903 of the European Parliament and of the Council of 13 March 2024 laying
down measures for a high level of public sector interoperability across the Union (Interoperable Europe
Act) (OJ L, 2024/903, 22.3.2024, ELI: http://data.europa.eu/eli/reg/2024/903/oj). 36 Interinstitutional Agreement between the European Parliament, the Council of the European Union and
the European Commission on Better Law-Making (OJ L 123, 12.5.2016, p. 1, ELI:
http://data.europa.eu/eli/agree_interinstit/2016/512/oj).
EN 37 EN
be exercised in accordance with Regulation (EU) No 182/2011 of the European
Parliament and of the Council (37).
(88) Due to the relevance of this Regulation on the protection of personal data, the
European Data Protection Supervisor should be consulted, where necessary, in
accordance with Article 42(1) of Regulation (EU) 2018/1725 (38).
(89) If any of the measures provided for by this Regulation constitute State aid, the
provisions concerning such measures are without prejudice to the application of
Articles 107 and 108 TFEU.
(90) This Regulation should be without prejudice to the application of Articles 101 and 102
TFEU, and to the enforcement powers of competition authorities.
(91) Since the objectives of this Regulation cannot be sufficiently achieved by the Member
States, but can rather, by reason of the scale or effects of the action, be better achieved
at Union level, the Union may adopt measures, in accordance with the principle of
subsidiarity as set out in Article 5 of the TEU. In accordance with the principle of
proportionality, as set out in that Article, this Regulation does not go beyond what is
necessary in order to achieve those objectives,
HAVE ADOPTED THIS REGULATION:
37 Regulation (EU) No 182/2011 of the European Parliament and of the Council of 16 February 2011
laying down the rules and general principles concerning mechanisms for control by Member States of
the Commission’s exercise of implementing powers, (OJ L 55, 28.2.2011, p. 13, ELI:
http://data.europa.eu/eli/reg/2011/182/oj). 38 Regulation (EU) 2018/1725 of the European Parliament and of the Council of 23 October 2018 on the
protection of natural persons with regard to the processing of personal data by the Union institutions,
bodies, offices and agencies and on the free movement of such data, and repealing Regulation (EC) No
45/2001 and Decision No 1247/2002/EC (OJ L 295, 21.11.2018, p. 39, ELI:
http://data.europa.eu/eli/reg/2018/1725/oj).
EN 38 EN
TITLE I
GENERAL PROVISIONS
Chapter I
Subject matter and definitions
Article 1
Subject matter
1. This Regulation establishes a framework for strengthening the cloud and AI
ecosystem at Union level, in particular through the following measures:
(a) establishing the Cloud Leadership Initiative and the AI Leadership Initiative
(‘the Cloud and AI Leadership Initiatives’);
(b) setting the framework for the accelerated deployment of data centres across the
Union;
(c) enabling the availability of a sovereign cloud and artificial intelligence (AI)
offer to safeguard the Union’s public order;
(d) reducing dependencies on critical technologies;
(e) fostering the adoption of cloud computing services across the public sector.
2. The first general objective of this Regulation is to ensure the conditions necessary for
the competitiveness and innovation capacity of the Union’s cloud and AI ecosystem.
3. The second general objective, separate from and complementary to the first general
objective in paragraph 2, is to improve the functioning of the single market by laying
down a uniform Union legal framework for increasing the Union’s resilience and
strategic autonomy in cloud and AI technologies.
Article 2
Definitions
For the purposes of this Regulation, the following definitions apply:
(1) ‘cloud computing service’ means cloud computing service as defined in Article 6,
point (30), of Directive (EU) 2022/2555;
(2) ‘cloud computing service provider’ means a legal entity which provides a cloud
computing service;
(3) ‘AI system’ means an AI system as defined in Article 3, point (1), of Regulation
(EU) 2024/1689;
(4) ‘frontier AI’ means AI models or AI systems built upon such models that can
perform a wide variety of tasks and that approach, reach or exceed the current state
of the art;
(5) ‘AI agent’ means an AI system or a coordinated set of AI systems, that can perceive
and act upon their environment, with a degree of autonomy, using tools as needed to
achieve specific goals and adapt to changing inputs and contexts;
EN 39 EN
(6) ‘public sector body’ means public sector body as defined in Article 2, point (1), of
Directive (EU) 2019/1024;
(7) ‘Union entities’ means the Union institutions, bodies, offices and agencies set up by
or pursuant to the Treaty on European Union, the Treaty on the Functioning of the
European Union (TFEU) or the Treaty establishing the European Atomic Energy
Community;
(8) ‘small and medium-sized enterprise’ or ‘SME’ means a small or medium-sized
enterprise as defined in Article 2 of Annex I to Commission Recommendation
2003/361/EC;
(9) ‘small mid-cap’ or ‘SMC’ means a small mid-cap enterprise as defined in point 2 of
the Annex to Commission Recommendation (EU) 2025/1099;
(10) ‘data centre’ means data centre as defined in point 2.6.3.1.16 of Annex A to
Regulation (EC) No 1099/2008 of the European Parliament and of the Council;
(11) ‘data centre operator’ means data centre operator as defined in Article 2, point (7), of
Delegated Regulation (EU) 2024/1364;
(12) ‘data centre service’ means data centre service as defined in Article 6, point (31), of
Directive (EU) 2022/2555;
(13) ‘software’ means software as defined in Article 3, point (4), of Regulation (EU)
2024/2847;
(14) ‘hardware’ means hardware as defined in Article 3, point 5, of Regulation (EU)
2024/2847;
(15) ‘component’ means component as defined in Article 3, point (6), of Regulation (EU)
2024/2847;
(16) ‘manufacturer’ means manufacturer as defined in Article 3, point (13), of Regulation
(EU) 2024/2847;
(17) ‘auditing organisation’ means an individual organisation, a consortium or other
combination of organisations, including any subcontractors, that the audited cloud
computing service provider has contracted to perform an independent audit;
(18) ‘audited service’ means a cloud computing service being audited for the purpose of
receiving an audit report and an audit opinion;
(19) ‘audit criteria’ means the criteria, pursuant to Annex II to this Regulation, against
which the auditing organisation assesses whether the audited provider and its audited
service comply with each cumulative criterion to be met for it to be recognised as
offering Union assurance levels 2, 3, or 4;
(20) ‘audit evidence’ means any information used by an auditing organisation to support
the audit findings and conclusions and to issue an audit opinion, including data
collected from documents, databases or IT systems, interviews or testing performed;
(21) ‘control’ means control as defined in Article 2, point (6), of Regulation (EU)
2021/697;
(22) ‘contracting authorities’ means contracting authorities as defined in Article 2(1),
point (1), of Directive 2014/24/EU;
(25) ‘open source licence’ means open source licence as defined in Article 2, point (12),
of Regulation (EU) 2024/903.
EN 40 EN
TITLE II
RESEARCH, DEVELOPMENT AND DEPLOYMENT
ACTIVITIES FOR THE CLOUD AND AI ECOSYSTEM
Chapter I
Cloud and AI Leadership Initiatives
Article 3
General objective of the Cloud and AI Leadership Initiatives
1. The Cloud and AI Leadership Initiatives shall pursue the general objective of
promoting research and innovation activities and achieving large-scale capacity
throughout the Union’s cloud and AI ecosystem, by:
(a) supporting the development and deployment of cutting-edge cloud and AI
technologies, including next-generation resource-efficient data centre
technologies, open cloud computing stack technologies, frontier AI, and
physical and industrial AI;
(b) reinforcing the Union’s data centre and cloud capacity to meet the growing
demands driven by AI, foster innovation and ensure the resilience of the digital
infrastructure;
(c) stimulating the Union’s demand and promoting the deployment and uptake of
cloud and AI technologies across the public sector, and the private sector, in
line with the digital target of digital transformation of businesses, established
by Decision (EU) 2022/2481.
2. The Cloud and AI Leadership Initiatives shall pursue the following operational
objectives:
(a) supporting the development and deployment of advanced data centre
technologies incorporating principles of energy efficiency and resource
efficiency by design and throughout operations (operational objective 1);
(b) supporting the development and deployment of cloud computing stacks
supporting the Union’s technological autonomy (operational objective 2);
(c) advancing Union’s capabilities in frontier AI (operational objective 3);
(d) advancing Union’s capabilities in physical AI models and systems and
fostering their deployment across the Union’s strategic sectors (operational
objective 4);
(e) accelerating the development and uptake of industrial AI across the Union’s
strategic sectors (operational objective 5);
(f) supporting the development of advanced platforms for the large-scale
deployment of AI agents (operational objective 6);
(g) increasing the development and adoption of AI models and systems across the
Union’s public sectors (operational objective 7);
EN 41 EN
(h) increasing the adoption of AI technologies at regional and local level, and the
uptake of cloud computing services provided by European cloud computing
service providers (operational objective 8).
Article 4
Operational objectives of the Cloud and AI Leadership Initiatives
1. Under operational objective 1, the Cloud and AI Leadership Initiatives shall:
(a) advance energy- and water-efficiency technologies for data centres, including
innovative cooling, next-generation direct current data centres, waste heat
utilisation solutions, and energy storage systems;
(b) promote the integration of emerging quantum computing technologies for
cloud and AI computing infrastructure operations;
(c) develop AI-powered technologies for optimising server efficiency, utilisation
rates and computing infrastructure operations;
(d) design and optimise cloud and edge AI infrastructures to ensure effective
integration with energy grids and to increase theirflexibility;
(e) leverage data centres as anchor clients for advanced energy management
systems harnessing diverse energy sources, including small modular reactors
and clean hydrogen, alongside efficient energy storage solutions;
(f) deploy test beds and pilot lines to integrate and test technologies developed
under points (a) to (e), covering energy-efficient semiconductor and quantum
computing prototypes.
2. Under operational objective 2, the Cloud and AI Leadership Initiatives shall:
(a) develop and pilot secure, resilient and performant open cloud computing stacks
covering on-device edge, connectivity, data and AI tools, backend and service
layers for strategic sectors;
(b) develop AI-optimised servers and baseline software based on processors,
accelerators and quantum accelerators designed and manufactured in the
Union, alongside next-generation ultra-high density and long-term data
storage;
(c) boost data availability for AI via open-source middleware platforms
underpinning common European data spaces;
(d) foster the creation of open-source software foundations supporting open-source
components;
(e) establish a catalogue of European open cloud computing solutions developed
under points (a) to (d) of this paragraph.
3. Under operational objective 3, the Cloud and AI Leadership Initiatives shall support
pioneering projects in frontier AI that develop frontier AI models and systems as
strategic assets, including in key sectors such as cybersecurity.
4. Under operational objective 4, the Cloud and AI Leadership Initiatives shall:
(a) accelerate the development of a European physical AI stack, supporting model
training and system development and deployment, in particular for robotics and
autonomous vehicles and drones;
EN 42 EN
(b) facilitate access to, and the collection and preparation of, specific datasets for
physical AI;
(c) support the development, testing and validation in real-world environments of
physical AI models and systems.
5. Under operational objective 5, the Cloud and AI Leadership Initiatives shall:
(a) accelerate the development and uptake of sectoral AI models and systems
across the Union’s strategic industrial sectors;
(b) facilitate access to the necessary computing resources and AI tools required to
develop and operationalise AI models and systems tailored to industrial sector
needs;
(c) enable secure large-scale data pooling for collaborative AI training through
technologies enhancing privacy and preserving confidentiality.
6. Under their operational objective 6, the Cloud and AI Leadership Initiatives shall:
(a) support the development of advanced resilient and secure platforms for the
development, deployment and orchestration of advanced AI agents at scale;
(b) facilitate the development of targeted testing and experimentation
methodologies of advanced AI agents and their orchestration throughout their
lifecycle.
7. Under operational objective 7 the Cloud and AI Leadership Initiatives shall:
(a) accelerate the technological development and uptake of AI models and systems
in critical public sector domains;
(b) develop AI models and systems that increase the effectiveness of public service
delivery and accessibility for the general public, improve decision-making, and
simplify administrative procedures;
(c) promote the sharing and reusing of training data and AI models across the
Union’s public services;
(d) facilitate secure, privacy-enhancing health data reuse for AI models and tools
in healthcare;
(e) facilitate the development, testing and deployment of AI models and tools in
the automotive sector, including for autonomous driving.
8. Under operational objective 8 the Cloud and AI Leadership Initiatives shall:
(a) promote the broad adoption of AI by private and public sector organisations,
including SMEs and SMCs, through the network of Experience and
Acceleration Centres for AI (‘Centres for AI’);
(b) develop a common cloud and AI curriculum, drawing on the network of
Centres for AI and other relevant European initiatives;
(c) promote the sharing of public sector data centre services and cloud computing
services by supporting a European public sector cloud federation (‘EuroCloud
Federation’);
(d) support the procurement of data centre services and cloud computing services
for Union entities and public sector bodies.
EN 43 EN
Article 5
Experience and Acceleration Centres for AI
1. Each Member State shall establish Experience and Acceleration Centres for AI
(‘Centres for AI’). Those Centres for AI shall build on the European digital
innovation hubs established under Article 16 of Regulation (EU) 2021/694 and,
where applicable, any successor entities established under Union law.
2. The objectives of the Centres for AI shall be to:
(a) support the integration and scaling-up of AI use cases in strategic industrial and
public sectors;
(b) accelerate the broad adoption of cloud and AI technologies at regional and
local levels, notably for SMEs, SMCs and public sector bodies, in line with the
‘AI first’ principle;
(c) leverage relevant infrastructure to accelerate the development and fine-tuning
of AI models and systems.
3. The Centres for AI shall be tasked, in particular, with:
(a) helping organisations accelerate their digital transformation through access to
and use of AI technologies, including by connecting organisations with
European providers of cloud and AI technologies;
(b) ensuring or providing access to relevant upskilling and reskilling schemes, in
close collaboration with the AI Skills Academy;
(c) facilitating the transfer of expertise across regions;
(d) supporting the scaling-up of spin-offs and start-ups emerging from universities,
incubators and other accelerators by facilitating access to clients, companies
and organisations seeking specialised AI services.
4. The Commission may adopt implementing acts detailing the procedure for
establishing Centres for AI and further arrangements concerning the participant
organisation profile, selection criteria and details on the implementation of the tasks
and functions. Those implementing acts shall be adopted in accordance with the
examination procedure referred to in Article 46(2).
5. Centres for AI shall have substantial overall autonomy as regards their organisation,
composition and working methods, in compliance with the objectives set out in this
Regulation.
6. A network of Centres for AI shall be established to support collaboration and the
exchange of best practices among Centres for AI, and to provide specialised services
across regions where the required skills or compute capacity are not available locally.
7. Member States and the Commission shall cooperate with existing networks
established under other Union initiatives, including Union initiatives in the field of
semiconductors and data.
Article 6
Implementation of the Cloud and AI Leadership Initiatives
1. The implementation of the Cloud and AI Leadership Initiatives’ operational
objectives shall be entrusted to the Commission and the Member States and, where
EN 44 EN
relevant, to joint undertakings or any other structures capable of achieving those
objectives.
2. The Cloud and AI Leadership Initiatives’ operational objectives shall be
implemented through large-scale, cross-sectoral initiatives addressing major
technological and industrial challenges of strategic relevance for the Union (‘grand
challenges’), as indicated in Annex I.
3. The Cloud and AI Leadership Initiatives may be supported by funding from Union
programmes, including Horizon Europe and the Digital Europe Programme, in
accordance with Regulation (EU) 2021/694 and Regulation (EU) 2021/695.
4. To reflect technological and market developments the Commission is empowered to
adopt delegated acts in accordance with Article 45 to amend Annex I in a manner
consistent with the objectives of the Cloud and AI Leadership Initiatives set out in
Article 4.
Article 7
National cloud and AI strategies
1. By [same day as entry into force plus one year], Member States shall establish
national cloud and AI strategies (the ‘national strategies’).
2. The national strategies shall include at least the following:
(a) key objectives and priorities for cloud and AI adoption, in line with the ‘AI
first’ principle, as well as a governance and monitoring framework to achieve
those objectives and priorities;
(b) measures to accelerate the development and adoption of cloud and AI at
national, regional and local level, particularly among public sector bodies,
SMEs and SMCs, including by supporting the Centres for AI referred to in
Article 5 as entry points to the European AI innovation ecosystem;
(c) measures to support the broad deployment and uptake of AI in strategic
industrial and public sectors, including in healthcare, energy and mobility;
(d) measures to support the deployment of data centre capacity, with a particular
focus on high-value data centres delivering significant economic and societal
benefits while adhering to high environmental and energy-efficiency standards;
(e) measures to invest in high-intensity computing infrastructure, including AI
factories, AI gigafactories and quantum computers as strategic national and
cross-border assets supporting research, development and industrial AI
deployment across strategic sectors;
(f) measures to support the development of cloud and AI capabilities and promote
excellence and innovation, including through public procurement measures,
and public procurement of innovation measures set out in Article 33;
(g) measures to support the development of cloud computing stack technologies
built upon open hardware and software to strengthen technological sovereignty
and enhance the competitiveness of strategic European industries;
(h) measures to ensure the accessibility of high-quality data for AI development,
notably by preventing data bottlenecks encountered by organisations.
3. National strategies shall be consistent with the objectives of this Regulation
EN 45 EN
4. Member States shall ensure that their national strategies are consistent with, and
contribute to, the associated digital targets established under Article 4 of Decision
(EU) 2022/2481.
5. The Member States shall notify the Commission of their national strategies within
three months of their adoption. Member States shall assess their national strategies at
least every three years on the basis of key performance indicators and, where
necessary, update them. The Commission shall monitor the adoption and revision of
the national strategies.
6. The European Artificial Intelligence Board established by Regulation (EU)
2024/1689 (the ‘AI Board’) shall advise and assist the Member States as regards the
coordination of national strategies. The AI Board shall facilitate exchange of best
practices among Member States.
Article 8
Criteria for frontier AI priority projects
The Commission may, by means of a decision, recognise as frontier AI priority
projects, projects selected through open calls for expression of interest that support
grand challenge 3 set out in Annex I, provided that the following criteria are fulfilled:
(a) it is a pioneering project, focused on the support and scaling-up of frontier AI
technologies;
(b) it is undertaken by a European digital infrastructure consortium established
pursuant Decision (EU) 2022/2481 or another legal entity eligible for funding
under Union law and it involves the participation of at least three Member
States;
(c) the participating Member States pool computing time and other relevant
resources to support the implementation of the designated project.
Article 9
Computing support for AI projects
1. The Union and the Member States shall ensure that sufficient AI computing
resources from their compute capacities are allocated to support the development of
frontier AI priority projects that fulfil the criteria set out in Article 8, within the
limits of available capacity.
2. The Union shall at least match the AI computing resources contributed by Member
States to frontier AI priority projects to the extent that sufficient AI computing
capacity is available within the Union’s share of European high performance
computing access time.
3. The Union and the Member States shall endeavour to provide sufficient computing
resource for AI industrial innovation, physical AI and public sector AI projects.
EN 46 EN
TITLE III
DATA CENTRE CAPACITIES
Chapter I
Data centre acceleration zones
Article 10
Designation of data centre acceleration zones
1. Where data centre capacity is being deployed within the territory of a Member State,
that Member State shall designate at least one data centre acceleration zone
(‘acceleration zone’) within its territory by [P.O. insert the date of entry into force of
this Regulation plus 6 months]. Member States shall consider the following aspects
when designating acceleration zones:
(a) the location and dimension of the site or area, and the minimum and maximum
size of the facilities that could be built on that site or area;
(b) the available and future power grid capacity and the possibility and conditions
for on-site storage and clean energy generation;
(c) the available and future network connectivity capacity;
(d) the capacity of the zone to support the phasing out of legacy copper networks;
(e) the available and future facilities that can reuse data centre waste heat;
(f) all the measures taken to accelerate the granting of the necessary permits for
constructing and operating data centres within the given zone;
(g) the preference for reusing brownfield sites over using greenfield sites;
(h) the ability of the site or area to function sustainably, particularly as regards
preventing or minimising environmental impacts and supporting the reduction
of carbon emissions and its climate resilience.
2. Member States, where appropriate to facilitate the development of acceleration
zones, shall:
(a) conduct, and review at least every three years, a comprehensive analysis of the
energy needs and their respective impacts on greenhouse gas emissions, of
current and future acceleration zones and identify the required energy
infrastructure capacity for the proper functioning and development of data
centre projects located in the acceleration zones. Such analysis shall be
conducted, at least, when designating the acceleration zones pursuant to
paragraph 1;
(b) ensure that the network development plans prepared by transmission system
operators pursuant to Article 51 of Directive (EU) 2019/944 of the European
Parliament and of the Council and distribution system operators pursuant to
Article 32 of Directive (EU) 2019/944 take due account of the analysis
prepared pursuant to point (a) of this paragraph, considering the potential of
anticipatory investments to accommodate future system needs.
3. National, regional and local authorities responsible for preparing spatial and
development plans shall consider including, in those plans, provisions for the
EN 47 EN
development of data centre projects deployed in acceleration zones, and of the
necessary infrastructure. Member States shall ensure that all relevant spatial planning
data are available to data centre operators. Where those plans are subject to an
assessment pursuant to Directive 2001/42/EC of the European Parliament and of the
Council and Article 6 of Directive 92/43/EEC, those assessments shall be combined.
Where applicable, the combined assessment shall also address the impact on
potentially affected water bodies referred to in Directive 2000/60/EC of the European
Parliament and of the Council.
4. When designating acceleration zones, Member States shall ensure the involvement of
and coordination among all relevant national, regional and local authorities and
entities, including operators as defined in Article 2, point (29), of Directive (EU)
2018/1972 of the European Parliament and of the Council, transmission system
operators as defined in Article 2, point (35), of Directive (EU) 2019/944 and
distribution system operators as defined in Article 2, point (29), of Directive (EU)
2019/944.
Article 11
Conditions within acceleration zones
1. When setting sustainability requirements for data centres deployed in acceleration
zones, Member States shall use the key performance indicators specified in
Delegated Regulation (EU) 2024/1364 pursuant to Directive (EU) 2023/1791 under
Annex II, from (a) to (n).
2. Member States shall ensure that the allocation and use of resources within
acceleration zones takes place on fair, reasonable and non-discriminatory terms and
does not give rise to speculative reservation or foreclosure practices capable of
impeding effective competition or the effective development or use of those zones.
Article 12
Single information points
1. The data centre operator shall have the right, upon request, to be assisted by a single
information point throughout the entire lifecycle of the data centre project in an
acceleration zone with respect to all authorisations required for the deployment of the
data centre. For that purpose, Member States shall designate one or more single
information points for data centre operators of data centre projects in acceleration
zones. The Member States may designate for this purpose a single information point
established under Regulation (EU) 2024/1309. The functions, procedures and
mechanisms applicable to such single information points under Regulation (EU)
2024/1309, including those relating to digital access, administrative coordination and
dispute settlement, shall also apply.
2. The role of a single information point may include, among other things, coordinating,
facilitating, monitoring and sharing information on the procedure relating to:
(a) spatial planning and building permits;
(b) environmental assessments, in accordance with Regulation (EU) 2026/XXXX
[on speeding-up environmental assessments];
(c) authorisations regarding water abstraction, wastewater discharge, and heat
utilisation and recovery;
(d) compliance with applicable administrative and reporting obligations;
EN 48 EN
(e) information to the public, with the aim of increasing public acceptance of the
data centre project;
(f) applications for connection to the electricity, heat or communications
networks, or to other relevant networks.
3. The single information point shall assist in assessing whether a data centre project
may qualify as a strategic project under Article 14.
4. When providing the administrative support and the assistance referred to in this
Article, the single point of contact shall pay particular attention to SMEs and, where
appropriate, establish a dedicated channel for communication with SMEs to provide
guidance and respond to queries related to the implementation of this Regulation.
Article 13
Facilitating administrative and permit-granting processes
1. Data centre projects deployed in acceleration zones shall be considered as strategic
projects within the meaning of Article 14 of Regulation (EU) 2026/XXX [on
speeding-up environmental assessments] and shall benefit from the toolbox set out in
the Annex to that Regulation.
2. For each designated acceleration zone, Member States shall prepare and issue an
aggregated baseline permit authorising the deployment of data centres in that
acceleration zone. This aggregated baseline permit shall cover the permits and
administrative authorisations required for the data centre projects located within the
acceleration zone, excluding installation-specific permits.
3. Before issuing the aggregated baseline permit referred to in paragraph 2, Member
States shall carry out all necessary procedures and assessments, including any
relevant environmental assessments, planning procedures and evaluations applicable
at the level of the acceleration zone.
4. Data centres deployed in acceleration zones shall be required to obtain additional
permits only for activities falling outside the aggregated baseline permit referred to
in paragraph 2.
5. Member States shall ensure that administrative applications related to the planning,
construction and the operation of data centre deployed in acceleration zones are
processed in an efficient, transparent and timely manner. The permit-granting
procedure for data centre projects deployed in data centre acceleration zones shall not
exceed 12 months, from the moment a comprehensive application has been
submitted. The time limit shall be without prejudice to any shorter time limits set by
Member States. Where such a status exists in national law, data centre projects shall
be allocated the status of highest national significance possible and be treated as such
in permit-granting processes. This paragraph shall apply only where such status
exists in national law and shall not create an obligation for Member States to
introduce such status.
EN 49 EN
Chapter II
Strategic projects
SECTION 1
DESIGNATION OF DATA CENTRE STRATEGIC PROJECTS
Article 14
Designation of data centre strategic projects
1. The Commission may, by means of a decision, designate as strategic projects, data
centre projects selected through open calls for expressions of interest that fulfil at
least two of the following criteria:
(a) the project establishes and operates infrastructure that directly supports and
enhances essential public sector functions, including research and education,
healthcare, public safety and security;
(b) the project includes highly sustainable or innovative features, including
technologies and solutions developed under Title II;
(c) the project contributes to the security, safety, and stability of the electricity grid
and contributes to the electricity system needs as evaluated by the relevant
system operator, in particular for projects involving the colocation of large
clean energy generation and storage facilities;
(d) the project supports the integration of chips, processors and accelerators,
servers or quantum computers designed and/or manufactured in the Union into
data centre systems or data centre facility management, thereby strengthening
the Union semiconductor, quantum and data centre supply chains and
contributing to the objectives of this Regulation and of Regulation (EU)
2023/1781;
(e) the project addresses a major shortage of compute capacity in an area identified
as having such a shortage under Article 15 and contributes significantly to the
growth, development and promotion of the local economy.
2. In its proposal, the applicant shall provide all the necessary and relevant information
to demonstrate that the project fulfils the relevant criteria.
3. The duration of the designation as a strategic project shall be based on the predicted
lifetime of the project. The applicant shall include in the proposal the information
necessary to substantiate the predicted lifetime of the project, on the basis of which
the duration of the designation as strategic project shall be determined.
4. Where the Commission finds that a project designated as a strategic project no longer
fulfils the relevant criteria, or where its designation was based on an application
containing incorrect information affecting compliance with those criteria, it may
withdraw the designation of that project by means of a decision. Projects for which
the designation as a strategic project has been withdrawn shall lose all rights
connected to that status under this Regulation.
EN 50 EN
Chapter III
Monitoring
Article 15
Monitoring the capacity gap
1. For the purpose of monitoring progress in the achievement of the objectives of
Decision (EU) 2022/2481, the Commission shall identify and monitor:
(a) the compute capacity available in the Union, including edge computing
capacity;
(b) the volume of demand for data centre capacity;
(c) the size of the capacity gap and underserved areas that could be identified by
the Commission, in cooperation with the Member States, and subsequently
used as acceleration zones for the deployment of data centre capacity.
EN 51 EN
TITLE IV
AUTONOMY
Chapter I
Cloud computing sovereignty framework
SECTION 1
UNION ASSURANCE LEVELS
Article 16
Scope
1. This Chapter establishes a Union cloud computing sovereignty framework
comprising four Union assurance levels, the criteria for which are set out in Annex
II, that cloud computing service providers shall meet in order to provide their cloud
computing services to Union entities and public sector bodies.
2. The Commission is empowered to adopt delegated acts in accordance with Article 45
to amend the Union assurance levels set out in Annex II and the evidence set out in
Annex III.
3. To ensure Annex II and Annex III remain up to date with new legal or technical
developments, the Commission shall review them at least every 18 months.
Article 17
Recognition of cloud computing service providers
1. A cloud computing service provider that aims to be recognised as offering a Union
assurance level, shall submit an application for recognition to the national competent
authority of establishment. When submitting an application for recognition, the cloud
computing service provider shall include all the relevant evidence required under
paragraphs 3 or 4.
2. The competent authority of establishment shall be the evaluating national competent
authority. An evaluating national competent authority that has received an
application for a candidate recognition, may, where necessary, request one or more
competent authorities of the other Member States to collaborate in the procedure for
a candidate recognition under this Article. Within 15 days of receiving such a
request, the national authority that has received a request for collaboration shall
either provide confirmation that it agrees to collaborate with the evaluating national
competent authority or refuse the request.
3. For Union assurance level 1, the candidate cloud computing service provider shall
submit to the evaluating national competent authority the EU statement of
conformity referred to in Article 19(2) and all the necessary evidence.
By way of derogation from the first subpragraph, the EU statement of conformity
issued under Article 19(2) by cloud computing service providers that are SMEs shall
be directly and automatically recognised in all Member States without the need for
prior recognition by the evaluating national competent authority.
EN 52 EN
4. For Union assurance levels 2, 3 and 4, the candidate cloud computing service
provider shall submit to the evaluating national competent authority the audit report,
the ‘positive’ audit opinion referred to in Article 20 and all the evidence provided to
the auditing organisation during the audit procedure.
5. Within 60 days of accepting an application pursuant to paragraph 1, the evaluating
national competent authority shall assess the evidence submitted pursuant to
paragraphs 3 or 4 and shall either:
(a) prepare a draft recognition decision and notify, as soon as possible, the
competent authorities of the other Member States for a 60-day review period to
confirm its intended recognition of the cloud computing service across the
Union as offering the applicable Union assurance level. The notification to the
competent authorities of the other Member States of the review period shall
include the evidence referred to in paragraphs 3 or 4; or
(b) where the evidence submitted is insufficient to allow the evaluating competent
authority to recognise the cloud computing service, it may request further
information from the applicant and request that the applicant submit such
information within a specified time limit. The period of 60 days referred to in
this paragraph shall be suspended from the date of issue of the request until the
date the information is received. The suspension shall not exceed 30 days in
total unless it is justified by the nature of the information requested or by
exceptional circumstances; or
(c) reject the request for recognition. Prior to rejecting the request for recognition,
the evaluating competent authority shall give the candidate cloud computing
service provider the opportunity to provide written comments on the
conclusions of the evaluation within 30 days. The evaluating competent
authority shall take due account of those comments when finalising its
conclusions.
6. During the review period referred to in paragraph 5, point (a), the national competent
authority of another Member State may submit a reasoned objection or request for
clarification to the evaluating national competent authority, where it considers that
the draft recognition decision does not comply with the applicable Union assurance
level set out in Annex II.
7. Where no reasoned objection or request for clarification is submitted within the
review period referred to in paragraph 5, point (a), the conclusions by the evaluating
national competent authority shall be deemed accepted by all Member States, the
evaluating national competent authority shall adopt the recognition decision and the
audited service shall be recognised throughout the Union at the appropriate Union
assurance level.
8. Where a request for clarification is submitted within the review period referred to in
paragraph 5, point (a), the evaluating national competent authority shall take due
account of such request and, where applicable, request new information from the
applicant as per paragraph 5, point (b) or confirm or modify its original draft
decision. Where the requesting competent authority is not satisfied, it may submit a
reasoned objection.
9. Where a reasoned objection is submitted within the review period referred to in
paragraph 5, point (a), or following the procedure referred to in paragraph 8, the
evaluating national competent authority shall assess the objection and shall either
EN 53 EN
maintain or revoke its original draft decision. The evaluating national competent
authority shall inform the competent authorities of the other Member States within
15 days after the end of the review perioed referred to in paragraph 5, point (a), or
within 15 days after receiving the reasoned objection following the procedure
referred to in paragraph 8, whichever is applicable.
10. In case the evaluating national competent authority intends to maintain its draft
decision, the concerned national competent authority may refer the matter to the
Commission. The Commission shall assess the referral and may request information
from the national competent authorities concerned. The Commission shall adopt a
binding decision determining whether the evaluating national competent authority
may adopt the recognition decision.
11. The evaluating national competent authority may revoke its recognition where it
finds that a cloud computing service provider, whose service was recognised across
the Union as providing a specific Union assurance level, intentionally or negligently,
supplied incorrect or misleading information.
12. The Commission may adopt implementing acts concerning the practical
arrangements for the procedures referred to in this Article. Those implementing acts
shall be adopted in accordance with the examination procedure referred to in Article
46(2).
13. The Commission may, in order to carry out the tasks assigned to it under paragraph
10 , require that national competent authorities of establishment provide, as soon as
possible and within a reasonable period, any relevant information relating to the
concerned cloud computing service provider and the application for recognition.
14. When sending a request for information, the Commission shall state the purpose of
the request, specify what information is required and set the period within which the
information is to be provided.
Article 18
Associated third countries
1. The Commission may adopt decisions, by means of implementing acts, identifying
third countries for which cloud computing service providers subject to the control of
that third country or a legal entity established in that third country may be audited
against the criteria for Union assurance level 3 pursuant to Annex II, provided that
that third country fulfils the following cumulative criteria:
(a) it is subject to a relevant adequacy decision adopted under Article 45 of
Regulation (EU) 2016/679;
(b) it has no measures in place that enable it to exercise control over the cloud
computing service provider in a way that would conflict with the requirements
for lawful access to non-personal data set out in paragraphs 2 and 3 of Article
32 of Regulation (EU) 2023/2854;
(c) it has no measures in place to compel the cloud computing service provider to
degrade or disrupt service continuity or provision. It also has no measures in
place to oblige the cloud computing service provider to implement, enforce,
give effect to, or comply with restrictive measures such as sanction regimes,
embargoes, or any equivalent legal or administrative measures, unless these
EN 54 EN
specific measures are legitimate under the national laws of Member States or
Union law;
(d) it has no measures in place to impede the provision of state-of-the-art
technologies and services provided by the cloud computing service provider;
(e) it maintains an open market to Union cloud computing services;
(f) the third country grants equivalent levels of access to public procurement
procedures of cloud computing services subject to the control of a Union
Member State or entity or a legal entity established in the Union.
Those implementing acts shall be adopted in accordance with the examination
procedure referred to in Article 46(2)
2. Where available information reveals that the third country no longer fulfils the
requirements under paragraph 1, the Commission shall repeal, amend or suspend the
decision referred to in paragraph 1.
3. The Commission shall publish on its website a list of third countries that fulfil the
requirements under paragraph 1 and those that no longer do so.
SECTION 2
CONFORMITY ASSESSMENT PROCEDURES
Article 19
Conformity self-assessment
1. Cloud computing service providers seeking recognition in accordance with Article
17 as offering Union assurance level 1, shall carry out a conformity self-assessment
of compliance with the criteria for Union assurance level 1 set out in Annex II.
2. Following the self-assessment referred to in paragraph 1, the cloud computing
service provider shall issue an EU statement of conformity stating that compliance
with the criteria for Union assurance level 1 have been demonstrated. By issuing
such a statement, the cloud computing service provider shall assume responsibility
for the compliance of the cloud computing service with the criteria for Union
assurance level 1set out in Annex II.
3. The cloud computing service provider shall make the EU statement of conformity
publicly available.
SECTION 3
INDEPENDENT THIRD-PARTY AUDITS
Article 20
Independent audit
1. Cloud computing service providers seeking recognition in accordance with Article
17 as offering Union assurance level 2, 3, or 4, shall undergo at their own expense,
independent third-party audits to obtain an audit report and an audit opinion from an
auditing organisation. An audited provider undergoing an audit procedure at a higher
Union assurance level shall satisfy all the applicable cumulative criteria under Annex
II applicable to the lower Union assurance levels. Failure to meet any requirements
of a lower assurance level shall preclude conformity with the higher Union assurance
levels.
EN 55 EN
2. Audited providers shall cooperate with auditing organisations and provide them
assistance necessary to enable them to conduct those audits in an effective, efficient
and timely manner, including by giving them access to all relevant data and premises
and by answering oral or written questions. Audited providers shall refrain from
hampering, unduly influencing or undermining the performance of the audit.
3. Auditing organisations shall ensure an adequate level of confidentiality and
professional secrecy in respect of the information obtained from the audited
providers and third parties as part of the audits, including after the audits have ended.
That requirement shall not adversely affect the performance of the audits and other
provisions of this Regulation. Under Article 23, the auditing organisation shall only
share information that are necessary for the reporting purposes and do not contain
any information that could reasonably be considered confidential.
4. Audits referred to paragraph 1 shall be performed by auditing organisations that:
(a) are independent from, and do not have any conflicts of interest with, the cloud
computing service provider concerned, and any legal person connected to that
provider, in particular:
i. have not provided non-audit services related to the matters audited to the
cloud computing service provider concerned or to any legal person
connected to that provider in the 12-month period before the beginning of
the audit, and have committed to not providing them with such services
in the 12-month period after the completion of the audit;
ii. have not provided auditing services pursuant to this Article to the cloud
computing service provider concerned or any legal person connected to
that provider in the 10-year period before the beginning of the audit;
iii. are not performing the audit in return for fees that are contingent on the
result of the audit;
(b) have proven expertise, technical competence and capabilities in auditing cloud
computing services;
(c) have proven objectivity and professional ethics, based in particular on
adherence to codes of practice or appropriate standards.
5. Auditing organisations that perform the audits shall prepare an audit report for each
audit. That report shall be substantiated, in writing, and shall include at least the
following:
(a) the name, address and point of contact of the provider subject to the audit, and
the period covered;
(b) the name and address of the auditing organisation or organisations performing
the audit;
(c) a declaration of interests;
(d) a description of the specific aspects audited, and the methodology applied;
(e) a description and a summary of the main findings drawn from the audit;
(f) a list of the third parties consulted as part of the audit;
(g) a ‘positive’ or ‘negative’ audit opinion and any information on whether the
audited service of the audited provider complies with the applicable audit
criteria for Union assurance level 2, 3 or 4 pursuant to Annex II;
EN 56 EN
(h) where the audit opinion is ‘negative’, operational recommendations on specific
measures to achieve compliance and the recommended timeframe to achieve
compliance;
(i) where the audit opinion is ‘positive’, the Union assurance level that needs to be
recognised under Article 17, issued to the audited service of the audited
provider pursuant to the applicable criteria set out in Annex II.
6. Where the auditing organisation was unable to audit certain aspects or to express an
audit opinion based on its investigations, the audit report shall include an explanation
of the circumstances and the reasons why those aspects could not be audited.
7. The auditing organisation may revoke its audit report and audit opinion where the
audited provider, intentionally or negligently, supplied incorrect or misleading audit
evidence.
8. The audited provider shall annually submit for review the audit report and the
associated ‘positive’ audit opinion to the same or a different auditing organisation
which shall assess the continued compliance of the audited service with the
applicable criteria set out in Annex II. On the basis of the annual review, the auditing
organisation may confirm, update, or revoke the initial audit report and audit opinion.
9. The Commission is empowered to adopt delegated acts in accordance with Article 45
to supplement this Regulation by laying down rules on the performance of audits on
the procedural steps, rules for auditing organisations and their technical
competences, auditing methodologies and templates for the audit reports.
Article 21
Content and quality of audit evidence
1. To prepare the audit report and audit opinion, the auditing organisation shall assess
the compliance of the audited service with the criteria set out in Annex II on the basis
of the audit evidence listed in Annex III. The Commission is empowered to adopt
delegated acts in accordance with Article 45 to amend Annex III by laying down the
necessary evidence needed to assess the audit criteria under Annex II.
2. The audit evidence shall be:
(a) relevant and sufficient to enable the auditing organisation to prepare an audit
report and provide an audit opinion; and
(b) reliable, according to the auditing organisation’s professional judgment and
scepticism.
Article 22
Central repository of cloud computing services
1. The Commission shall establish and maintain a dedicated repository of cloud
computing services that have been recognised in accordance with Article 17 (‘central
repository’).
2. The national competent authority of establishment that recognised a cloud computing
service under Article 17 shall register the cloud computing service in the central
repository.
EN 57 EN
3. The revocation of an audit report and audit opinion by an auditing organisation or the
revocation of a recognition by a competent authority shall be published in the central
repository and shall remain available there for five years.
4. The central repository shall be publicly available and regularly updated by the
Commission and the national competent authorities of establishment on a dedicated
and easily accessible website.
Article 23
Transparency obligations
1. On becoming aware of any information or any material change in circumstances that
may affect the audit report and the ‘positive’ opinion under Article 20 or the
recognition under Article 17, the recognised cloud computing service provider shall,
as soon as possible, notify the auditing organisation and the national competent
authority of establishment.
2. On the basis of the notification under paragraph 1, the auditing organisation shall
assess whether the audit report or the audit opinion need to be amended or revoked.
Where the auditing organisation amends or revokes the audit report or the audit
opinion, it shall, as soon as possible, notify the national competent authority of
establishment.
3. On the basis of the notification referred to in paragraph 1 or 2, the national
competent authority of establishment shall assess whether its recognition needs to be
amended or revoked. Where the national competent authority of establishment
amends or revokes it recognition of the cloud computing service, it shall, as soon as
possible, notify the national competent authorities of the other Member States and
the Commission.
Article 24
Penalties and compensation
1. Member States shall lay down the rules on penalties applicable to infringements of
this Chapter by cloud computing service providers within their competence and shall
take all measures necessary to ensure that they are implemented. The penalties
provided for shall be effective, proportionate and dissuasive. Member States shall, as
soon as possible, notify the Commission of those rules and of those measures and
shall notify the Commission of any subsequent amendment affecting them.
2. Member States shall take into account the following non-exhaustive criteria for the
imposition of penalties for infringements of this Regulation:
(a) the nature, gravity, scale and duration of the infringement;
(b) any action taken by the infringing party to mitigate or remedy the damage
caused by the infringement;
(c) any previous infringements by the infringing party;
(d) the financial benefits gained or losses avoided by the infringing party due to
the infringement, insofar as such benefits or losses can be reliably established;
(e) any other aggravating or mitigating factor applicable to the circumstances of
the case;
(f) infringing party’s annual turnover in the preceding financial year in the Union.
EN 58 EN
3. Recipients of the cloud computing services shall have the right to seek, in accordance
with Union and national law, compensation from cloud computing service providers
for any damage or loss suffered due to an infringement by those providers of their
obligations under this Chapter.
SECTION 4
NATIONAL COMPETENT AUTHORITIES
Article 25
National competent authorities
1. By [P.O. insert date of entry into force plus 1 year], Member States shall designate
one or more national competent authorities responsible for enforcing this Chapter. To
that effect, Member States may designate an existing authority or existing authorities
(‘competent authorities’).
2. Member States shall notify the Commission of the names of the competent
authorities and of their tasks and powers. The Commission shall maintain a public
register of those authorities.
3. Member States shall ensure that their competent authorities perform their tasks under
this Regulation in an impartial, transparent and timely manner. Member States shall
ensure that their competent authorities have all necessary resources to carry out their
tasks, including sufficient technical, financial and human resources to adequately
supervise all cloud computing service providers within their competence.
4. The Member State in which the cloud computing service provider has its main
establishment, that is, where the cloud computing service provider has its head office
or registered office from which the principal financial functions and operational
control are exercised, shall have exclusive competence for enforcing this Chapter.
Article 26
Powers of the national competent authorities
1. Where needed to carry out their tasks under Article 17, competent authorities of
establishment shall have the following investigative powers:
(a) the power to require any cloud computing service provider, as well as any other
persons acting for purposes related to their trade, business, craft or profession,
who may reasonably be expected to be aware of information relating to a
suspected infringement of this Regulation, including auditing organisations, to
provide that information as soon as possible;
(b) the power to carry out, or to request a judicial authority in their Member State
to order, inspections of any premises that those providers or those persons
acting for purposes related to their trade, business, craft or profession, use for
purposes related to their trade, business, craft or profession, or to request other
public authorities to do so, in order to examine, seize, take or obtain copies of
information relating to a suspected infringement in any form, irrespective of
the storage medium;
(c) the power to ask any member of staff or representative of those providers or
those persons acting for purposes related to their trade, business, craft or
profession, to give explanations in respect of any information relating to a
EN 59 EN
suspected infringement and, with their consent, to record their answers by any
technical means.
2. Where needed to carry out their tasks under Article 17, national competent
authorities of establishment shall have the following enforcement powers:
(a) the power to order the cessation of infringements and, where appropriate, to
impose remedies proportionate to the infringement and necessary to bring the
infringement effectively to an end, or to request a judicial authority in their
Member State to do so;
(b) the power to impose fines, or to request a judicial authority in their Member
State to do so, for failure to comply with this Regulation, including with any of
the investigative orders issued pursuant to paragraph 1;
(c) the power to impose a periodic penalty payment, or to request a judicial
authority in their Member State to do so, in accordance with Article 24 to
ensure that an infringement is terminated in compliance with an order issued
pursuant to point (a), or for failure to comply with any of the investigative
orders issued pursuant to paragraph 1.
3. Measures taken by national competent authorities of establishment in exercising their
powers listed in paragraphs 1 and 2 shall be effective, dissuasive and proportionate,
having regard, in particular, to the nature, gravity, recurrence and duration of the
infringement or suspected infringement to which those measures relate, and, where
relevant, the economic, technical and operational capacity of the service provider
concerned.
4. Member States shall set out specific rules and procedures for the exercise of the
powers pursuant to paragraphs 1 and 2 and shall ensure that any exercise of those
powers is subject to adequate safeguards under applicable national law in compliance
with the general principles of Union law. Those measures shall be taken only in
accordance with the right to respect for private life and the rights of defence,
including the rights to be heard and to have access to the file, and shall be subject to
the right of all affected parties to an effective judicial remedy.
SECTION 5
MUTUAL ASSISTANCE AND COOPERATION
Article 27
Mutual assistance
1. Competent authorities and the Commission shall cooperate closely and provide each
other with mutual assistance to apply this Chapter in a consistent and efficient
manner. Mutual assistance shall include the exchange of information.
2. A competent authority may request other competent authorities to provide specific
information in their possession relating to a specific cloud computing service
provider to exercise its investigative powers under Article 26 regarding specific
information located in their Member State. Where appropriate, the competent
authority receiving the request may involve other competent authorities or other
public authorities of the Member State in question.
3. The competent authority receiving the request pursuant to paragraph 2 shall comply
with such request and inform the competent authority of establishment about the
EN 60 EN
action taken, as soon as possible and no later than two months after receipt of the
request, unless duly justified.
Article 28
Cross-border cooperation
1. Where a competent authority of destination has reason to suspect that a cloud
computing service provider no longer fulfils the requirement under Annex II to this
Regulation, it may request the competent authority of establishment to assess the
matter and to take the necessary investigatory and enforcement measures to ensure
compliance.
2. The Commission may also request the competent authority referred to in Article 25
to assess the matter and take the necessary investigatory and enforcement measures
to ensure compliance.
3. Requests pursuant to paragraph 1 or 2 shall be duly reasoned and shall be duly taken
into account by the competent authority of establishment. Where the competent
authority of establishment considers that the information provided is insufficient, it
may either request additional information. The period set out in paragraph 4 shall be
suspended until that additional information is provided
4. The competent authority of establishment shall, as soon as possible and in any event
not later than two months after receipt of the request pursuant to paragraph 1 or 2,
communicate to the competent authority that sent the request, and the Commission,
its assessment of the suspected infringement and an explanation of any investigatory
or enforcement measures taken or envisaged in relation to the matter to ensure
compliance with this Regulation.
Chapter II
Demand-side measures
SECTION 1
PUBLIC PROCUREMENT
Article 29
Risk assessments
1. By [date of entry into force plus 1 year], and thereafter every two years, or whenever
necessary, Member States and Union entities shall carry out risk assessments that
shall:
(a) identify the public sector activities that use or will make use of cloud
computing services, that contribute to the preservation of public order in
sectors falling under Annex I or II of Directive (EU) 2022/2555 and in the
areas of national security, internal security, external border management,
defence, justice or law enforcement, including the prevention, investigation,
detection and prosecution of criminal offence;
(b) determine which Union assurance level 2, 3, or 4 set out in Annex II of this
Regulation is appropriate for the identified public sector activities.
EN 61 EN
Where Union entities and Member States share responsibilities in relation to the
public sector activities, they shall, where appropriate, consider carrying out the
relevant risk assessment or assessments jointly.
2. In carrying out their risk assessments, Member States and Union entities shall
consider at least the following aspects:
(a) the sensitivity, criticality, and magnitude of the non-personal data processed,
including the potential impact on public order and the nature, scope, context
and purpose of processing of personal data, as well as the risk of varying
likelihood and severity for the rights and freedoms of data subjects;
(b) the risk and consequent impact on public order of unlawful access under Union
law to such data by a third country or a legal entity established in a third
country;
(c) the risk and consequent impact on public order of possible service disruption;
3. The Commission shall, by means of implementing acts in accordance with Article
46(2), specify the methodology to be applied, the templates to be used and the
elements to be taken into account by the Member States and Union entities for the
purpose of carrying out the risk assessments referred to in paragraph 1. The
methodology shall specify how Member States use the highest level of assurance for
the most critical public sectors activities including, but not limited to, defence.
4. Within three months of carrying out the risk assessments referred to in paragraph 1,
Member States shall provide the Commission with the results of those risk
assessments, indicating where they depart from the implementing acts referred to in
paragraph 3.
5. If the Commission concludes, after reviewing the results of the risk assessment or
assessments of a Member State, that the Union assurance level identified for the
public sector activity in a risk assessment is not appropriate or does not adequately
address the public order concerns, the Commission may adopt implementing acts in
accordance with Article 46(2) specifying the Union assurance levels needed for the
public sector activity.
6. Where the risk assessment requires the migration to another cloud computing
service, the Member State or Union entity shall migrate within a reasonable
transition period that shall not exceed 12 months, taking into account technical
feasibility, continuity of service and data portability requirements applicable to such
migration.
7. Member States shall cooperate with each other and with the Commission through
established consistency mechanisms and promote cooperation and effective
exchange of information and best practices.
8. For the purpose of paragraph 3, the Commission shall be empowered to request
cloud computing service providers to provide all the necessary information.
9. In their risk assessments, Member States and Union entities shall consider whether a
multi-vendor or multi-cloud strategy is appropriate as part of their procurement of
cloud computing services.
EN 62 EN
Article 30
Public procurement
1. This Article applies to contracting authorities that procure cloud computing services
for their exclusive use. Without prejudice to Article 136 of Regulation (EU,
Euratom) 2024/2509, this Article also applies to Union entities that procure cloud
computing services for their exclusive use.
2. Union entities and public sectors bodies whose public sector activities have not been
identified as contributing to the preservation of public order under the risk
assessment referred to in Article 29(1) shall use cloud computing services that have
been recognised under Article 17 as having a Union assurance level 1.
3. Contracting authorities, including the entities acting on their behalf, whose activities
have been identified as contributing to the preservation of public order under Article
29(1) in sectors falling under Annex I or II of Directive (EU) 2022/2555 and in the
areas of national security, internal security, external border management, defence,
justice or law enforcement, including the prevention, investigation, detection and
prosecution of criminal offence, shall only procure cloud computing services that
have been recognised as having a Union assurance level 2, 3 or 4.
4. By derogation from paragraphs 2 or 3, on an exceptional basis and where duly
justified, contracting authorities may decide not to procure cloud computing services
recognised as having a Union assurance level 1, 2, 3, or 4 where one or more of the
following circumstances applies:
(a) the subject matter of the tender cannot be supplied by recognised cloud
computing services available in the central repository referred to in Article 22,
and no adequate or reasonable alternative or comparable cloud computing
service exists, and such absence is not the result of an artificial narrowing
down of the parameters of the public procurement procedure;
(b) the contracting authority has launched a similar procurement process within the
previous year but did not receive any suitable tenders or suitable participants;
(a) applying the requirements of this Regulation would require the contracting
authority to procure services at disproportionate cost.
SECTION 2
PRIVATE SECTOR ENTITIES
Article 31
Impact assessments
1. Entities referred to in Annex I of Directive (EU) 2022/2555 who are not public sector
bodies may carry out similar assessments as those set out in Article 29.
2. The Commission may issue guidance on the methodology for carrying out the impact
assessments under this Article and possible mitigation measures to be adopted by
private sector entities operating in sectors of high criticality.
3. Where, because of specific circumstances, and where duly justified and in
consultation with the Member States, the Commission concludes that entities who
are not public sector bodies operating in sectors of high criticality require an impact
assessment, the Commission may adopt delegated acts to supplement this Regulation
in accordance with Article 45 specifying the need for such impact assessment and the
EN 63 EN
risk mitigation measures that those entities who are not public sector bodies shall
take.
SECTION 3
OTHER PROCUREMENT-RELATED MEASURES
Article 32
Union added value
1. In public procurement procedures for innovative cloud computing services and AI
systems, contracting authorities shall include, as part of the quality evaluation of the
tender, non-price award criteria that allow them to evaluate the tenderer’s
contribution to the development of a European cloud and AI ecosystem.
2. When applying non-price award criteria under paragraph 1, contracting authorities
shall ensure that non-price award criteria are:
(a) linked to the subject matter of the contract;
(b) not conferring unrestricted freedom of choice on the contracting authority;
(c) expressly set out in the procurement documents or in the contract notice;
(d) ancillary and not decisive in the award of the contract.
3. Without affecting contracting authorities’ discretion to apply additional criteria, the
non-price award criteria referred to in paragraph 1 shall enable contracting
authorities to evaluate the extent to which:
(a) the tenderer contributes to strengthening the digital technology supply chain in
the Union, including the use of software or hardware designed or manufactured
in the Union;
(b) the tenderer has integrated technologies developed in the Union, including
research and development results stemming from Union funded research and
development programmes and makes use of tools, such as standards,
specification, software, models or other technology developed in the Union;
(c) the innovation required to deliver the service contributes to strengthening the
security of supply and the development of a European cloud and AI ecosystem;
(d) the service is delivered, to the greatest extent feasible with regard to market
availability and technical requirements, through critical computing, storage and
networking hardware components designed and/or manufactured in the Union,
or, where this is not feasible, through hardware components from a third
country that contributes to strengthening the security of supply and the
development of a European cloud and AI ecosystem.
Article 33
Monitoring of procurement of innovation in cloud and AI
1. Member States shall monitor and report on their use of procurement of innovation in
cloud computing services and AI systems.
2. Member States shall take appropriate measures to ensure that the monitoring and
reporting referred to in paragraph 1 are actively used to identify barriers to SMEs
participation in procurement procedures, to improve access of SMEs to procurement
markets, support the design of simplified, proportionate and SME-friendly
EN 64 EN
procurement strategies, including division into lots, where appropriate, promote the
participation of SMEs in the innovation procedure foreseen under Directive
2014/24/EU and pre-commercial procurement of cloud computing services and AI
systems.
3. Based on the monitoring referred to in paragraph 1, Member States shall inform the
Commission, on a yearly basis, of the following information:
(a) the size of the economic operators participating in such procurement;
(b) SMEs participation trends, including the number of contracts awarded to
SMEs, their share of the total contract value, as a percentage, and, where
available, the share of cross-border SMEs participation;
(c) measures taken to improve SMEs access to public procurement procedures.
4. Member States shall pursue as objective that at least 25% of their procurement for
cloud computing services and AI systems be awarded to innovative SMEs. Member
States shall include, in their national strategies referred to in Article 7, plans on how
they intend to achieve this objective.
5. Union entities and contracting authorities shall promote:
(a) preliminary market consultations;
(b) matchmaking between public buyers and innovative solutions provided by
European SMEs and start-ups;
(c) development of public contract clauses that are favourable for innovative
SMEs.
Chapter III
European public sector cloud federation
Article 34
Establishment of the European public sector cloud federation
1. The European public sector cloud federation (the ‘EuroCloud Federation’) is hereby
established. The EuroCloud Federation shall be open for the participation of Union
entities and public sector bodies on a voluntary basis. Union entities and public
sector bodies may request the Commission to join the EuroCloud Federation.
2. The EuroCloud Federation shall facilitate the sharing of public sector data centre
services and cloud computing services between Union entities and public sector
bodies under the conditions set out in Articles 35 and 36.
3. The Commission shall establish a platform for the EuroCloud Federation providing
at least:
(a) a catalogue providing information on available public sector data centre
services and cloud computing services;
(b) a service platform for the exchange and orchestration of computing, storage
and network resources and services;
4. The Commission is empowered to adopt implementing acts to specify the procedure
to participate in the EuroCloud Federation and template concerning the content and
EN 65 EN
other details of the request for participation. Those implementing acts shall be
adopted in accordance with the examination procedure referred to in Article 46(2).
Article 35
Sharing of public sector data centre services and cloud computing services
1. A member of the EuroCloud Federation (the ‘sharing entity’) may share data centre
services and cloud computing services with another member of the EuroCloud
Federation (the ‘using entity’) where the sharing entity directly, or indirectly through
an intermediate legal entity, owns the hardware through which the service is made
available and provides the service that is made available to the using entity. Where
the sharing entity indirectly owns the hardware and provides the services through an
intermediate legal entity, the sharing entity shall exercise control over that
intermediate legal entity.
2. The sharing entity shall put in place appropriate technical, operational and
organisational measures to ensure an effective, secure and resilient provision of
services.
3. Prior to sharing data centre services and cloud computing services within the
EuroCloud Federation, the sharing entity shall demonstrate to the Commission that it
fulfils the conditions set out in paragraphs 1 and 2.
4. The Commission shall assess the information provided by the sharing entity and
allow the sharing entity to share data centre services and cloud computing services
within the EuroCloud Federation where the conditions laid down in paragraphs 1 and
2 are fulfilled.
5. The sharing entity may charge a fee to the using entity. The amount of the fee shall
be limited to the costs that the sharing entity incurs in relation to the sharing of the
service and shall not constitute a pecuniary interest within the meaning of Article 2
of Directive 2014/24/EU and Regulation (EU, Euratom) 2024/2509.
6. The Commission is empowered to adopt implementing acts to specify the technical,
operational and organisational measures referred to in paragraph 2. Those
implementing acts shall be adopted in accordance with the examination procedure
referred to in Article 46(2).
Article 36
Fees for the administration of the EuroCloud Federation
1. The costs arising from the activities carried out by the Commission pursuant to this
Chapter shall be jointly financed by the members of the EuroCloud Federation
through fees levied by the Commission.
2. If the costs are initially borne by the general budget of the Union, they shall be
reimbursed by the EuroCloud members over a period not exceeding three years from
the date on which the costs were borne by the Union.
3. Revenues generated by the fees shall constitute internal assigned revenues within the
meaning of Article 21(3), point (a), of Regulation (EU, Euratom) 2024/2509. Those
revenues shall be assigned to cover the costs of the activities carried out by the
Commission pursuant to this Chapter, including assessing request to join the
EuroCloud Federation and the establishment of the platform referred to in Article
EN 66 EN
34(3). Any revenue remaining after covering those costs shall be entered into the
general budget of the Union.
4. The Commission shall adopt implementing acts laying down detailed rules for
determining the estimated costs, the individual amount of the fees, and the manner
and conditions under which the fees are to be paid. Those implementing acts shall be
adopted in accordance with the examination procedure referred to in Article 46(2).
Chapter IV
Procurement of data centre services, cloud computing services,
software and AI systems by the Commission
Article 37
Procurement activities of the Commission
1. The Commission may carry out procurement activities to procure data centre
services, cloud computing services, software and AI systems for itself and forUnion
entitiesand for contracting authorities of Member States , in accordance with
Regulation (EU, Euratom) 2024/2509, subject to the exceptions set out in this
Chapter. By way of derogation from Article 168 of Regulation (EU, Euratom)
2024/2509, partner organisations referred to in Article 168(3) of that Regulation
selected by the Commission may also participate in the procurement activities set out
in this Chapter. Contracting authorities of Member States, Union entities, and partner
organisations selected by the Commission, shall be considered as ‘participating
entities’ under this Chapter.
2. Contracting authorities may participate in the procurement procedures on their own
behalf, or as central purchasing bodies within the meaning of Article 2(1), point (16)
of Directive 2014/24/EU when they qualify as such. Specific rules and obligations
governing their participation may be imposed where they participate as central
purchasing bodies in the agreement referred to in Article 38.
3. In addition to the procurement activities provided for in Article 168 of Regulation
(EU, Euratom) 2024/2509, the Commission may act as a central purchasing body for
contracting authorities of Member States and partner organisations selected by the
Commission, by:
(a) procuring data centre services, cloud computing services, software and AI
systems on behalf of, or in the name of, one or more contracting authorities of
Member States and partner organisations selected by the Commission, by
concluding framework contracts or operating dynamic purchasing systems for
services intended for the participating entities;
(b) acting as a wholesaler by acquiring such services and supplies and reselling
them or, in exceptional circumstances, donating them to one or more
contracting authorities of Member States.
4. In carrying out procurement activities, the Commission may provide ancillary
support to participating entities, including:
(a) technical infrastructure enabling participating entities to use awarded contracts
or award contracts, including specific contracts under concluded framework
agreements, for data centre services, cloud computing services, software and
AI systems;
EN 67 EN
(b) advice and support on preparing and implementing procurement procedures;
(c) preparation and conduct of procurement procedures on behalf of, or in the
name of, the entities concerned;
(d) invoicing and other administrative services relating to the contracts awarded.
5. Such ancillary support may be provided directly by the Commission, through a
subcontractor, or by delegation to Union bodies or agencies. Accession of
participating entities to the agreement referred to in Article 38 may be subject to the
acceptance of one or more ancillary support services.
6. The Commission may establish and manage a common procurement platform
including services that may be used to facilitate the procurement activities under this
Chapter.
Article 38
Arrangements for the procurement activities by the Commission
1. Before any procurement activity to be carried out under Article 37, the Commission
and at least two Member States shall enter into an agreement laying down the
practical arrangements for the procurement activities carried out by the Commission
under this Chapter. The agreement shall cover procurement procedures to be carried
out during its period of validity and shall be deemed to satisfy the requirements of
the joint procurement agreement and mandate referred to in Article 168(2) and (3) of
Regulation (EU, Euratom) 2024/2509.
2. The agreement shall constitute a mandate for the Commission to procure on behalf
of, or in the name of, the participating entities within the meaning of Article 168(3),
point (e), of Regulation (EU, Euratom) 2024/2509.
3. The agreement shall include the practical arrangements for the participation of
entities, the decision-making process for the choice of procedure and the applicable
conditions, the evaluation of requests for participation and tenders, the award of
contracts, and the applicable law and competent jurisdiction. The Commission shall
remain responsible for the operation and management of procurement activities,
including for deciding on the launch of a procurement procedure, the type of
procedure and of contract, and the award of contracts.
4. The agreement shall establish a Steering Committee composed of the Commission
and one representative from each participating Member States at national level.
Member States may accede to the agreement at a later stage and shall then be
represented in the Steering Committee. The Steering Committee may appoint
additional representatives of other Union entities, of contracting authorities of
Member States and of partner organisations selected by the Commission.
5. The Steering Committee shall be responsible for the strategic oversight of the
procurement activities, including for proposing the strategic direction of the
procurement agenda for a fixed period, and for approving the strategic direction of
each procurement procedure before it is launched by the Commission, to ensure its
compliance with the framework established by this Regulation.
6. Once the agreement has entered into force, contracting authorities of participating
Member States, Union entities and partner organisations selected by the Commission
may accede to and benefit from it and shall be considered as participating entities in
EN 68 EN
the procedures in which they elect to participate. The Steering Committee may
determine that the agreement shall take the form of a contract of adhesion.
7. The participation of a contracting authority of a Member State shall not be
conditional on that Member State’s participation.
8. The Steering Committee shall set transparent and non-discriminatory conditions for
contracting authorities of Member States to accede to the agreement, in particular as
regards size, minimum amounts and other objective criteria. The Steering Committee
shall also set out the rules and procedures governing the termination of participation
in the agreement of a contracting authority of a Member State that has failed to
comply with its obligations under the agreement.
9. By way of derogation from Article 168(2) of Regulation (EU, Euratom) 2014/2509,
the Steering Committee may approve the participation of contracting authorities from
EFTA States and Union candidate countries without the need for a bilateral or
multilateral treaty provided for such possibility.
10. The Steering Committee may make accession to the agreement conditional on
participating entities accepting one or more ancillary support services, as set out in
Article 37.
11. The Steering Committee shall adopt its rules of procedure, following a proposal from
the Commission.
Article 39
Applicable public procurement framework
1. A participating entity shall be deemed to have fulfilled its obligations under
applicable Union public procurement law where it acquires supplies or services by
means of contracts awarded by the Commission under this Chapter, including
through framework contracts concluded by or dynamic purchasing systems operated
by the Commission acting as a central purchasing body, or any ancillary support
services referred to in Article 37.
2. The procedural provisions applicable to Union institutions shall apply to the
procedures for the award of specific contracts under framework contracts or dynamic
purchasing systems.
3. A contracting authority that has acquired data centre services, cloud computing
services, software and AI systems from the Commission as a central purchasing body
shall ensure, in its agreements with the contracting authorities it serves, compliance
with any contractual requirements by which it is itself bound.
4. The Commission may decide to launch a procurement procedure open to
participating entities without a prior specific request from them.
5. By way of derogation from Article 168 of Regulation (EU, Euratom) 2024/2509,
participating entities may request from the Commission, throughout the period of
validity of a dynamic purchasing system, the possibility to participate in the system.
Such request shall be approved by the Commission provided that the cumulative
requests do not exceed 50% of the initial estimated quantities of the envisaged
purchases. The participation shall be approved within 10 working days of receipt of
the request and shall allow the participating entities to be included in any future
invitation to tender.
EN 69 EN
6. The possibility referred to in paragraph 5 shall be available only to participating
entities that accede to the agreement referred to in Article 38 after the dynamic
purchasing system has been launched.
Article 40
Fees for procurement activities
1. The costs arising from the procurement activities carried out pursuant to this Chapter
shall be jointly financed by the participating entities through fees levied by the
Commission.
2. The costs incurred in establishing the common procurement activities referred to in
Article 37, including the development of the common procurement platform, may be
initially borne by the general budget of the Union. In such case, they shall be
reimbursed by the participating entities over a period not exceeding three years from
the date on which they were borne by the Union. The Commission may adopt
implementing acts laying down the practical and operational arrangements for
reimbursement by the participating entities. Those implementing acts shall be
adopted in accordance with the examination procedure referred to in Article 46(2).
3. Revenues generated by the fees shall constitute internal assigned revenues within the
meaning of Article 21(3), point (a), of Regulation (EU, Euratom) 2024/2509. Those
revenues shall be assigned to cover the costs of the procurement activities carried out
pursuant to Article 37. Any revenue remaining after covering those costs shall be
entered into the general budget of the Union.
4. The fees shall be set in advance, shall be proportionate to the estimated costs of the
activities for which fees are chargeable as determined in a cost-effective way,
reflecting practices of comparable procurement frameworks, and shall be sufficient
to cover those costs.
5. 5. The Commission shall adopt implementing acts laying down detailed rules for
determining the fees, specifying the following:
(a) the estimated costs attributable to the procurement activities for which fees are
chargeable;
(b) the individual amounts of the chargeable fees;
(c) the manner and conditions under which the fees are to be paid;
(d) the conditions under which the fees are to be paid.
Those implementing acts shall be adopted in accordance with the examination
procedure referred to in Article 46(2).
Chapter V
Open source
Article 41
Promoting open source solutions andopen source first
The Union and Member States shall take the necessary measures to encourage Union entities
and public sector bodies to use and facilitate the reuse of open standards and components
released under an open source licence when building their cloud and AI ecosystem or stack,
EN 70 EN
taking into account functionalities, including security, total cost, and other relevant, duly
justified objective criteria.
Article 42
Share and reuse of software
When making software to which they hold intellectual property rights available for reuse
under an open source licence, a Union entity or public sector body shall do so using a
catalogue or repository that is connected to, and made accessible through, the EU OSS
Catalogue referred to in Article 43.
Article 43
EU Open Source Solutions Catalogue
1. The Commission shall provide and maintain an EU Open Source Solutions
Catalogue (‘EU OSS Catalogue’) as a centralised catalogue to access software made
available for reuse by Union entities and public sector bodies.
2. The EU OSS Catalogue shall be hosted on the Interoperable Europe portal referred to
in Article 8 of Regulation (EU) 2024/903 and shall be accessible electronically free
of charge.
3. The Commission shall, on the basis of objective and relevant criteria, decide on the
request of any Union entity or public sector body owning or maintaining a catalogue
or repository to have that catalogue or repository connected to and made accessible
through the EU OSS Catalogue.
Article 44
Network of Open Source Programme Offices
1. The Commission shall establish a network of Open Source Programme Offices
(‘OSPO Network’) to facilitate cooperation on the implementation of the obligations
under this Chapter.
2. Open Source Programme Offices established by public sector bodies at local,
regional or national level in a Member State, and those established by Union entities,
may request from the Commission to join the OSPO Network.
3. The OSPO Network shall have the following tasks:
(a) facilitating the exchange of information, experience and best practices between
Member States and the Commission, in particular by discussing common
technical, legal and organisational challenges, including those related to
licensing, security, maintenance and procurement of open-source software;
(b) promoting the sharing and reuse of open-source software by public sector
bodies;
(c) contributing, on a voluntary and non-binding basis, to the development of
guidance, templates or recommendations on the sharing and reuse of open-
source software;
(d) collaborating on and exchanging open-source projects of common interest to
Union entities and public sector bodies.
4. The Commission shall support and coordinate the OSPO Network.
EN 71 EN
5. The Commission shall convene and chair a meeting of the members of the OSPO
Network at least twice a year. The meetings of the OSPO Network may be organised
online.
EN 72 EN
TITLE V
FINAL PROVISIONS
Article 44
Exercise of the delegation
1. The power to adopt delegated acts is conferred on the Commission subject to the
conditions laid down in this Article.
2. The power to adopt delegated acts referred to in Article 6(4), Article 16(2), Article
20(9), Article 21(1), and Article 31(3) shall be conferred on the Commission for an
indeterminate period of time from [date of entry into force].
3. The delegation of power referred to in Article 6(4), Article 16(2), Article 20(9),
Article 21(1), and Article 31(3) may be revoked at any time by the European
Parliament or by the Council. A decision to revoke shall put an end to the delegation
of the power specified in that decision. It shall take effect the day following the
publication of the decision in the Official Journal of the European Union or at a later
date specified therein. It shall not affect the validity of any delegated acts already in
force.
4. Before adopting a delegated act, the Commission shall consult experts designated by
each Member State in accordance with the principles laid down in the
Interinstitutional Agreement of 13 April 2016 on Better Law-Making.
5. As soon as it adopts a delegated act, the Commission shall notify it simultaneously to
the European Parliament and to the Council.
6. Delegated act adopted pursuant to Article 6(4), Article 16(2), Article 20(9), Article
21(1), and Article 31(3) shall enter into force only if no objection has been expressed
either by the European Parliament or by the Council within a period of two months
of notification of that act to the European Parliament and to the Council or if, before
the expiry of that period, the European Parliament and the Council have both
informed the Commission that they will not object. That period shall be extended by
three months at the initiative of the European Parliament or of the Council.
Article 46
Committee procedure
1. The Commission shall be assisted by a committee. That committee shall be a
committee within the meaning of Regulation (EU) No 182/2011.
2. Where reference is made to this paragraph, Article 5 of Regulation (EU) No
182/2011 shall apply.
Article 47
Review
1. By [date of entry into force plus 4 years], and every 5 years thereafter, the
Commission shall evaluate this Regulation, and report to the European Parliament,
the Council and the European Economic and Social Committee.
2. Where appropriate, the report referred to in paragraph 1 shall be accompanied by a
proposal for amendment of this Regulation.
EN 73 EN
3. In carrying out the evaluation referred to in paragraph 1, to Commission shall take
into account the positions and findings of the European Parliament, of the Council,
and of other relevant bodies or sources, and shall pay specific attention to small and
medium-sized enterprises and the position of new competitors.
Article 48
Entry into force and application
This Regulation shall enter into force on the twentieth day following that of its publication in
the Official Journal of the European Union.
It shall apply from [same day and month as date of entry into force plus 1 year].
This Regulation shall be binding in its entirety and directly applicable in all Member States.
Done at Brussels,
For the European Parliament For the Council
The President The President
EN 1 EN
LEGISLATIVE FINANCIAL AND DIGITAL STATEMENT
1. FRAMEWORK OF THE PROPOSAL/INITIATIVE ................................................. 3
1.1. Title of the proposal/initiative ...................................................................................... 3
1.2. Policy area(s) concerned .............................................................................................. 3
1.3. Objective(s) .................................................................................................................. 3
1.3.1. General objective(s) ..................................................................................................... 3
1.3.2. Specific objective(s) ..................................................................................................... 3
1.3.3. Expected result(s) and impact ...................................................................................... 3
1.3.4. Indicators of performance ............................................................................................ 3
1.4. The proposal/initiative relates to: ................................................................................. 4
1.5. Grounds for the proposal/initiative .............................................................................. 4
1.5.1. Requirement(s) to be met in the short or long term including a detailed timeline for
roll-out of the implementation of the initiative ............................................................ 4
1.5.2. Added value of EU involvement (it may result from different factors, e.g.
coordination gains, legal certainty, greater effectiveness or complementarities). For
the purposes of this section 'added value of EU involvement' is the value resulting
from EU action, that is additional to the value that would have been otherwise
created by Member States alone. ................................................................................. 4
1.5.3. Lessons learned from similar experiences in the past .................................................. 4
1.5.4. Compatibility with the multiannual financial framework and possible synergies with
other appropriate instruments ....................................................................................... 5
1.5.5. Assessment of the different available financing options, including scope for
redeployment ................................................................................................................ 5
1.6. Duration of the proposal/initiative and of its financial impact .................................... 6
1.7. Method(s) of budget implementation planned ............................................................. 6
2. MANAGEMENT MEASURES................................................................................... 8
2.1. Monitoring and reporting rules .................................................................................... 8
2.2. Management and control system(s) ............................................................................. 8
2.2.1. Justification of the budget implementation method(s), the funding implementation
mechanism(s), the payment modalities and the control strategy proposed .................. 8
2.2.2. Information concerning the risks identified and the internal control system(s) set up
to mitigate them............................................................................................................ 8
2.2.3. Estimation and justification of the cost-effectiveness of the controls (ratio between
the control costs and the value of the related funds managed), and assessment of the
expected levels of risk of error (at payment & at closure) ........................................... 8
2.3. Measures to prevent fraud and irregularities ................................................................ 9
3. ESTIMATED FINANCIAL IMPACT OF THE PROPOSAL/INITIATIVE ............ 10
3.1. Heading(s) of the multiannual financial framework and expenditure budget line(s)
affected ....................................................................................................................... 10
EN 2 EN
3.2. Estimated financial impact of the proposal on appropriations ................................... 12
3.2.1. Summary of estimated impact on operational appropriations.................................... 12
3.2.1.1. Appropriations from voted budget ............................................................................. 12
3.2.1.2. Appropriations from external assigned revenues ....................................................... 17
3.2.2. Estimated output funded from operational appropriations......................................... 22
3.2.3. Summary of estimated impact on administrative appropriations ............................... 24
3.2.3.1. Appropriations from voted budget .............................................................................. 24
3.2.3.2. Appropriations from external assigned revenues ....................................................... 24
3.2.3.3. Total appropriations ................................................................................................... 24
3.2.4. Estimated requirements of human resources.............................................................. 25
3.2.4.1. Financed from voted budget....................................................................................... 25
3.2.4.2. Financed from external assigned revenues ................................................................ 26
3.2.4.3. Total requirements of human resources ..................................................................... 26
3.2.5. Overview of estimated impact on digital technology-related investments ................ 28
3.2.6. Compatibility with the current multiannual financial framework.............................. 28
3.2.7. Third-party contributions ........................................................................................... 28
3.3. Estimated impact on revenue ..................................................................................... 29
4. DIGITAL DIMENSIONS .......................................................................................... 29
4.1. Requirements of digital relevance .............................................................................. 30
4.2. Data ............................................................................................................................ 30
4.3. Digital solutions ......................................................................................................... 31
4.4. Interoperability assessment ........................................................................................ 31
4.5. Measures to support digital implementation .............................................................. 32
EN 3 EN
1. FRAMEWORK OF THE PROPOSAL/INITIATIVE
1.1. Title of the proposal/initiative
Proposal for a Regulation of the European Parliament and Council establishing a
framework of measures for strengthening Europe’s cloud and AI ecosystem (Cloud
and AI Development Act).
Short title: The Cloud and AI Development Act (CADA)
(Text with EEA relevance)
1.2. Policy area(s) concerned
The proposal concerns policy areas linked to the development of cloud computing
services, with a focus on ensuring a smooth and timely transition toward high-value,
resilient and future oriented digital ecosystems. It aims to set the regulatory
conditions for a Single Market in cloud computing services, incentivising investment
in cloud infrastructure, AI research and development and supporting emerging
requirements arising from innovative technologies. By modernising existing
frameworks, the initiative aims to facilitate efficient investment, promote sustainable
competition, and reduce technological dependencies, thereby supporting the EU’s
long-term competitiveness and resilience in the global cloud and AI landscape. The
proposal also aims to remove persisting barriers to the cross-border provision of
cloud computing services and improve regulatory coherence and predictability across
Member States.
1.3. Objective(s)
1.3.1. General objective(s)
The general objective of this initiative is to ensure the functioning of the internal
market for cloud computing services and to secure the conditions necessary for the
Union’s competitiveness and strategic autonomy.
1.3.2. Specific objective(s)
Specific objective No 1: Increase computing capacity deployed in the EU through
innovative and sustainable technologies. By 2030, the EU should at least triple its
current data centre capacity, prioritising energy-efficient technologies in new
installations. As demand continues growing, this should be considered an
intermediate objective so that by 2035, the computing capacity in the EU should
meet its needs.
Specific objective No 2: Ensure attractive conditions for the deployment of
sustainable and innovative computing capacity. While the first objective is aimed at
the deployment of capacity, this one targets the conditions for such investment and
deployment. By 2030, operators should be able to obtain all permits to build and run
a data centre in less than 18 months throughout the EU, including access to land,
permits for energy access, and connectivity, which are a major attention point for
investors.
Specific objective No 3: Decrease the overall reliance on non-European cloud
computing services. By 2035, this intervention should increase the market share of
European cloud computing service providers in the European market. Strengthening
the Union’s strategic autonomy requires reducing dependencies and ensuring that
EN 4 EN
European users have credible European alternatives to non-European incumbents. A
stronger European supply base improves the Union’s capacity to act autonomously
and enhances long-term resilience, competitiveness, and security of supply.
Specific objective No 4:Contribute to the protection of public order by enhancing
the resilience of supply of cloud computing services, in particular in the public
sector. By 2035, highly critical use cases in the public sector should be operated
using sovereign cloud and AI computing services to ensure data confidentiality,
operational autonomy and prevent harms that could undermine public order. Highly
critical use cases are those of particular systemic importance that underpin essential
functions or involve the processing of sensitive data. Ensuring that, for these, data is
protected, and service continuity is guaranteed, is a key element of attaining strategic
autonomy. That is why these use cases are a priority for the move towards services
whose provision is outside of the reach of third-country policies that could result in
data access or interruptions to service continuity, i.e. sovereign services.
1.3.3. Expected result(s) and impact
Specify the effects which the proposal/initiative should have on the beneficiaries/groups targeted.
The proposed Cloud and AI Development Act is expected to have a relevant impact
on several beneficiaries, including the public sector, the European cloud and AI
ecosystem and citizens. The initiative aims to strengthen the cloud and AI ecosystem
at Union level, pioneering energy and resource efficiency for data centres, develop
European open cloud and AI stacks, and promote the uptake of cloud computing
services. This is expected to lead to the development of advanced AI technologies
and frontier AI, the acceleration of industrial AI models and systems in strategic
sectors, and the support of cross sectoral initiatives addressing major technological
and industrial challenges. Furthermore, the proposal seeks to accelerate data centre
deployment across the Union, tripling the EU’s data centre capacity within the next
five to seven years and ensuring the deployment of resource-efficient data centres.
The Act also aims to promote autonomy, reduce dependencies on critical
technologies and increase the adoption of cloud computing services across the public
sector, enabling critical sectors to use sovereign cloud computing services.
Additionally, the proposal will lead to the establishment of a European public sector
cloud federation, common procurement activities and the promotion of open-source
solutions. Overall, the proposal seeks to empower these beneficiairies, fostering a
competitive, innovative and autonomous European cloud and AI ecosystem that
drives economic prosperity, social wellbeing and strategic autonomy. The proposal is
envisioned to have positive effects on the global competitivess of the European AI
and cloud sector, and the attractiveness of the EU for third countries’ businesses and
researchers.
1.3.4. Indicators of performance
Specify the indicators for monitoring progress and achievements.
Objective 1: Increase computing capacity in the EU through innovative and
sustainable technologies.
(a) Installed computing capacity (MW IT load) by MS
(b) Aggregate general purpose and AI-optimised compute, measured also in
FLOPs
(c) EU share of global installed computing capacity
EN 5 EN
(d) Utilisation rate of EU computing capacity; measures on PUE, WUE, location-
based emissions and related environmental impact of data centres
(e) Deployment of innovative and energy-efficient technologies (pilots launched
and uptake of new solutions)
(f) Share of clean energy in data centres and waste-heat reuse
(g) Total annual public and private investment in EU-based DCs
(h) Share of new data centre capacity deployed outside existing hubs and in
underserved regions
Objective 2: Ensure attractive conditions for the deployment of sustainable and
innovative computing capacity.
(a) Average permitting time for new data centre projects
(b) Total administrative burden for operators
(c) Share of projects delayed/cancelled due to regulatory or infrastructure barriers
(d) Number of MS with simplified permitting frameworks
(e) Cost competitiveness index
Objective 3: Decrease the reliance on non-European cloud and AI computing
services.
(a) Share of total EU cloud computing services revenue captured by European
service providers
(b) Number of public sector authorities served by sovereign providers per MS
(c) Share of installed EU DC capacity owned by European providers
(d) Share of idle capacity across MS
Objective 4: Contribute to the protection of public order by enhancing the resilience
of supply of cloud computing services, in particular in the public sector.
(a) Number of cloud services audited under levels 2,3,4
(b) Compliance rate by contracting authorities (%) with the sovereignty scheme
(c) Annual value of EU public procurement of sovereign cloud computing services
(d) Number of public sector solutions released as open source in the repository,
and their downloads by third parties.
1.4. The proposal/initiative relates to:
a new action
a new action following a pilot project / preparatory action39
the extension of an existing action
a merger or redirection of one or more actions towards another/a new action
1.5. Grounds for the proposal/initiative
1.5.1. Requirement(s) to be met in the short or long term including a detailed timeline for
roll-out of the implementation of the initiative
The Cloud and AI Development Act will be expected to enter into force within 20
days from the publication in the Official Journal. The entry into application should
be within one year of publication, with notable exceptions for rules that require
additional transition period.
39 As referred to in Article 58(2), point (a) or (b) of the Financial Regulation.
EN 6 EN
1.5.2. Added value of EU involvement (it may result from different factors, e.g.
coordination gains, legal certainty, greater effectiveness or complementarities). For
the purposes of this section 'added value of EU involvement' is the value resulting
from EU action, that is additional to the value that would have been otherwise
created by Member States alone.
The development of computing capacity in the EU currently takes place along
national lines. Each Member State operates under a distinct framework, with
different processes and requirements for data centre deployment, reflecting local
conditions and needs. However, national policies for data centre acceleration risk
further fragmentation and race-to-the-bottom with respect to sustainability.
Moreover, an increasing number of low-latency applications require close computing
capacity. More generally, the EU faces a shortage of computing capacity, a problem
that risks negatively affecting its competitiveness and requires EU-level action to
maintain a regulatory and investment environment that is easy to navigate for data
centre operators and investors, including across borders.
Closing this capacity gap and allowing European businesses and public
administrations to leverage compute capacity while ensuring sustainability requires
action at EU level. The dependence on cloud computing services supplied by non-
European providers has the same root causes across the EU and affects businesses
and public administrations in all Member States. European service providers face
difficulties to scale up across the EU, for example due to different national
trustworthiness standards, particularly in public procurement. Divergent national
procurement practices complexify the market for European providers and the
underlying situation of imperfect information is a market failure requiring an EU-
level response. Calls for EU-action to address these challenges were also made in the
public consultation.
EU action is expected to have a clear added value in addressing the problem of
limited and geographically concentrated availability of computing capacity. By
providing a common approach to data centre deployment, it would enable the
coherent planning and deployment of computing capacity in a geographically
balanced way, while avoiding a race to the bottom and reducing regulatory
complexity for investors and data centre operators. The EU is uniquely positioned to
ensure that investment and acceleration policies reflect collective priorities and avoid
fragmentation. EU-level action would ensure that all businesses and public
administrations can access sufficient compute capacity to meet their needs and is a
prerequisite for Europe to become an AI continent.
In addressing the dependence on cloud computing services supplied by non-
European providers, EU action is expected to deliver benefits that exceed what
Member States could achieve individually, especially in addressing the underlying
market failures of imperfect information. This would improve the functioning of the
internal market and enable cloud computing service providers to grow beyond their
national markets.
1.5.3. Lessons learned from similar experiences in the past
The proposal is informed by the practical experience in the implementation of
existing regulations in this field. Past experience has shown that legal safeguards are
needed but not sufficient to change dependence on non-EU providers. GDPR and
EDPS enforcement pushed public bodies and providers towards stronger contractual
EN 7 EN
controls and tighter rules on international transfers. However, these measures
produced compliance solutions rather than concrete changes. Similarly, the need for
data localisation proved to only partially solve some of the sovereignty requirements
needed for specific use cases, especially in the public sector. The EUCS cloud
certification preparatory work also demostrated that cybersecurity must be
distinguished from sovereignty requirements. The regulation also builds on and
complements the Cybersecurity Act 2.0, as well as provisions of the Data Act, which
focuses on switching costs, interoperability, portability and unfair contractual terms.
Finally, the proposal was informed by an extensive stakeholder consultation strategy.
1.5.4. Compatibility with the multiannual financial framework and possible synergies with
other appropriate instruments
The proposal is compatible with the multiannual financial framework as it would
primarily provide a new legal framework while relying on existing or planned EU
instruments for financing. Its main funding synergies should be with FP10 and the
European Competitiveness Fund (ECF), especially under the Digital Leadership
window. FP10 should support the upstream research and innovation dimension under
Pillar I, while ECF should serve as the main deployment instrument, thus
contributing to translate FP10 research outputs into operational capabilities. Other
instruments could further reinforce these efforts. IPCEIs would continue to support
large-scale, cross-border projects where cloud, edge, chips, cybersecurity or AI
infrastructure require coordination among Member States and private invesment.
EDICs could provide useful governance vehicles for groups of Member States
wishing to jointly operate common digital infrastructure. Cohesion policy
instruments, e.g. ERDF and the Cohesion Fund contribute to and may support
regional competitiveness and address territorial disparities by co-financing digital
infrastructure in less-developed regions. Synergies are also expected with InvestEU
by improving the investment environment, supporting bankable projects and
complementing financial instruments to mobilise private and public investment.
Where relevant, the proposal would also build on reforms and investments set out in
national RRPs, including National Reform Programmes and country-specific
recommendations. Finally, other ECF policy windows could support sector-specific
applications.
1.5.5. Assessment of the different available financing options, including scope for
redeployment
The assessment of available financing options has considered both redeployment
within existing Commission resources and the need for additional financing. The
implementation of the initiative is estimated to require 25 FTEs in total. Of these, 8
FTEs from DG DIGIT and 7 FTEs from DG CNECT are considered to fall within the
scope of redeployment, reflecting tasks that can be covered through the
reprioritisaton of existing activities and use of existing policy expertise. This means
that 15 of the 25 estimated FTEs could be covered through redeployment. These
redeployed posts would mainly support activities close to existing mandates, such as
policy coordination, legal analysis, stakeholder engagement, internal market
monitoring, programme management, administrative support and cooperation with
Member States. For the remaining tasks, alternative financing options have been
examined. These include targeted reinforcement of relevant budget lines and use of
existing programme envelopes. However, given that several tasks may entail
significant operational costs, fees are also introduced as an important financing
EN 8 EN
mechanism and to limit the impact on the EU budget. In particular, such fees had
been evisaged to cover activities related to common procurement activities and
administration of the EuroCloud Federation for sharing idle capacity among
interested Member States.
The remaining 10 FTEs would require additional financing as they related to new or
substantially expanded reponsibilities that go beyond current workload assumptions.
Overall, the proposed approach combines redeployment, fee-based financing and
limited additional resources. Redeployment would ensure budgetary discipline and
use of existing Commission expertise. Fees are expected to reduce pressure on the
EU budget and support budget neutrality for most resource-intensive tsasks.
Additional appropriations are thus reserved for new functions that cannot be financed
through fees or internal reallocation.
EN 9 EN
1.6. Duration of the proposal/initiative and of its financial impact
limited duration
– in effect from [DD/MM]YYYY to [DD/MM]YYYY
– financial impact from YYYY to YYYY for commitment appropriations and
from YYYY to YYYY for payment appropriations.
unlimited duration
– Implementation with a start-up period from 2028 to 2030,
– followed by full-scale operation.
1.7. Method(s) of budget implementation planned
Direct management by the Commission
– by its departments, including by its staff in the Union delegations;
– by the executive agencies
Shared management with the Member States
Indirect management by entrusting budget implementation tasks to:
– third countries or the bodies they have designated
– international organisations and their agencies (to be specified)
– the European Investment Bank and the European Investment Fund
– bodies referred to in Articles 70 and 71 of the Financial Regulation
– public law bodies
– bodies governed by private law with a public service mission to the extent that they
are provided with adequate financial guarantees
– bodies governed by the private law of a Member State that are entrusted with the
implementation of a public-private partnership and that are provided with adequate financial
guarantees
– bodies or persons entrusted with the implementation of specific actions in the
common foreign and security policy pursuant to Title V of the Treaty on European Union, and
identified in the relevant basic act
– bodies established in a Member State, governed by the private law of a Member
State or Union law and eligible to be entrusted, in accordance with sector-specific rules, with
the implementation of Union funds or budgetary guarantees, to the extent that such bodies are
controlled by public law bodies or by bodies governed by private law with a public service
mission, and are provided with adequate financial guarantees in the form of joint and several
liability by the controlling bodies or equivalent financial guarantees and which may be, for
each action, limited to the maximum amount of the Union support.
EN 10 EN
2. MANAGEMENT MEASURES
2.1. Monitoring and reporting rules
The Regulation will be reviewed and evaluated five years from its entry into force.
The Commission will report on the findings of the evaluation to the European
Parliament and the Council. To support the consistent implementation and
monitorting of this Regulation, Member States should also ensure that the relevant
information concerning their activities is available to the Commission in a timely
manner.
2.2. Management and control system(s)
2.2.1. Justification of the budget implementation method(s), the funding implementation
mechanism(s), the payment modalities and the control strategy proposed
The Regulation establishes a new policy framework for harmonised rules governing
the procurement of cloud computing services in the internal market and the
deployment of data centres across the EU, while supporting the Union’s policy
objectives of consumer trust, industrial competitiveness, security and resilience and
sustainability. The Cloud and AI Development Act aims to simplify and improve
coordination of the regulatory framework for the development of data centres and
procurement of cloud computing services. These objectives require reinforced EU-
level coordination and operational capacity, which in turn require targeted and
proportionate budgetary resources. The proposal introduces proportionate changes,
establishing a governance system with new EU-level tasks with a Single Market
dimension, while ensuring that decision-making remains at the most efficient level.
The chosen governance model aims to build on existing structures and mandates,
thereby limiting the need for new entities and allowing the budget to be implemented
through established administrative and financial arrangements, ensuring cost-
effectiveness and predictability of expenditures.
In order to carry out these new tasks, additional human resources are required. The
implementation and enforcement of the Regulation is estimated to require 6
additional FTEs for the DG CNECT and 4 FTEs for DG DIGIT. The proposed
staffing levels are proportionate to the volume and complexity of the new
responsibilities and reflect the most cost-efficient option, avoiding duplication at
national level.
Payments will follow standard EU budgetary procedures, including commitments
and payments made annually, in accordance with the Financial Regulation and
within the ceilings of the applicable Multiannual Financial Framework. This will be
supported by the annual contributions from the fees covering the costs from
additional tasks and long term costs. Expenditure will be subject to the
Commission’s internal control framework, including ex ante checks, ex post audits,
performance monitoring and reporting. This will be performed with the aim to ensure
sound financial management, legality and regularity fo expenditure and effective use
of Union funds, while ensuring the timely implementation of the Regulation.
2.2.2. Information concerning the risks identified and the internal control system(s) set up
to mitigate them
The activities proposed in the Act involve different Union entities and some
execution risks that require monitoring, oversight, coordination and guidance effort
EN 11 EN
for their mitigation. The commission will dedicate staff to, among other activities,
elaborate guidance documents, delegated acts, dependency assessments, comitology
secretariat, and oversight of the implementation of the initiative, including
monitoring its progress against established KPIs and milestones. This would allow to
promptly identify possible issues and risks in the execution of activities.
In particular for the common procurement activities, a Steering Committee shall be
established, composed of the Commission and representatives of Member States, and
will be responsible for strategic oversight of the procurement activities, including the
strategic orientations of the agenda of public procurement activities, and the strategic
orientation of each procurement procedure, ensuring compliance with this
Regulation, and transparent and non-discriminatory conditions for accession of
contracting authorities. In the case of the European public sector cloud federation, it
is important to ensure broad participation from Union entities, while ensuring the
highest level of security in the provision of services, in accordance with Union law,
to ensure feasibility, effectiveness and continuity. Therefore, it should be open for
the participation of voluntary Union institutions, bodies and agencies, as public
sector bodies from Member States willing to interconnect their cloud computing
infrastructures and deploy interoperable cloud computing services across the Union.
The Commission would establish a platform to facilitate the sharing of data centre
and cloud computing capacity accessible to its members. Due to the likely criticality
and sensitivity of the data and applications hosted in that shared capacity, the
platform would include services and mechanisms for secure access and incident
management, including shared identity management, mutual authentication tools and
incident reporting tools, and capabilities supporting the operation of the provided
services, including monitoring of service provision, resource allocation, service
activation and performance.
2.2.3. Estimation and justification of the cost-effectiveness of the controls (ratio between
the control costs and the value of the related funds managed), and assessment of the
expected levels of risk of error (at payment & at closure)
The cost of controls for this initiative have been estimated at Commission level. The
source of this information is the Commission’s internal management and control
system. The costs were estimated based on the staff and resources dedicated to the
activities foreseen as part of this initiative. The expected total costs for such controls
can be relatively high due to the complexity of the activities proposed and the need
for dedicated resources to mitigate execution risks. The control intensity will be
adapted to the nature of the expenditure, the type of beneficiaries or contractors
foreseen, the amount of financial resources concerned, and the level of risk. Specific
actions such as common procurement, EuroCloud Federation, funding management
and repository of sovereign services will require specific resources and controls,
including strategic oversight, monitoring and incident management. The cost of such
controls has been estimated to ensure that it remains proportionate to the value of the
funds managed, the complexity of each activity and the identified risks.
In terms of expected error rate, the aim is to maintain this below the 2% threshold.
This will be achieved through standard ex ante and ex post controls mentioned
above, including e.g. verification of eligibility and legality of expenditure,
procurement and contract management checks, monitoring of deliverables, risk-
based controls, audits, and recovery procedures where needed. Any deviation from
EN 12 EN
this would require a coordinated approach and would be discussed on a case-by-case
basis.
2.3. Measures to prevent fraud and irregularities
The existing fraud prevention measures applicable to the Commission will cover the
additional appropriations necessary for this Regulation.
EN 13 EN
3. ESTIMATED FINANCIAL IMPACT OF THE PROPOSAL/INITIATIVE
3.1. Heading(s) of the multiannual financial framework and expenditure budget line(s)
affected
• Existing budget lines
In order of multiannual financial framework headings and budget lines.
Heading of
multiannual
financial
framework
Budget line Type of
expenditure Contribution
Number
Diff./Non-
diff.40
from
EFTA
countries 41
from
candidate
countries
and
potential
candidates 42
From
other
third
countries
other assigned
revenue
MFF headings and budget lines to be
determined43
Diff./Non
-diff. YES NO NO NO
Fee revenue (COM will collect the fee) Diff./Non
-diff. YES NO NO YES
• New budget lines requested
In order of multiannual financial framework headings and budget lines.
Heading of
multiannual
financial
framework
Budget line Type of
expenditure Contribution
Number
Diff./Non-
diff.
from
EFTA
countries
from
candidate
countries
and
potential
candidates
from
other
third
countries
other assigned
revenue
[XX.YY.YY.YY]
Diff./Non
-diff. YES/NO YES/NO YES/NO YES/NO
[XX.YY.YY.YY]
Diff./Non
-diff. YES/NO YES/NO YES/NO YES/NO
[XX.YY.YY.YY]
Diff./Non
-diff. YES/NO YES/NO YES/NO YES/NO
40 Diff. = Differentiated appropriations / Non-diff. = Non-differentiated appropriations. 41 EFTA: European Free Trade Association. 42 Candidate countries and, where applicable, potential candidates from the Western Balkans. 43 Budget lines for the new MFF are not yet known
EN 14 EN
3.2. Estimated financial impact of the proposal on appropriations
3.2.1. Summary of estimated impact on operational appropriations
– The proposal/initiative does not require the use of operational appropriations
– The proposal/initiative requires the use of operational appropriations, as explained below
The amounts indicated below are provisional and remain subject to the final outcome of the 2028-2034 MFF negotiations. This initiative
will be financed by redeployment within the operational programmes of the next MFF, and partially by administrative expenditure for
staff. At this stage, it is not possible to indicate the contribution from each MFF heading and programme, while it is expected that a
significant contribution will come from programmes under heading 2 of the 2028-2034 MFF (e.g. the European Competitiveness Fund,
especially under the Digital Leadership window). These estimates have been prepared based on the information currently available and
are intended to provide accurate budget estimates.
3.2.1.1. Appropriations from voted budget
EUR million (to three decimal places)
Heading of multiannual financial framework Number
DG CNECT
Year Year Year Year Year Year Year TOTAL
MFF
2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
Operational appropriations
Budget line
Commitments (1a) 2.386 0.727 2.118 0.427 0.427 0.427 0.427 6.941 6.941
Payments (2a) 1.193 1.557 1.422 1.272 0.427 0.427 0.427 6.727 0.214 6.941
TOTAL Commitments =1a 2.386 0.727 2.118 0.427 0.427 0.427 0.427 6.941 0.000 6.941
EN 15 EN
DG CNECT
Year Year Year Year Year Year Year TOTAL
MFF
2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
appropriations
for DG
CNECT
Payments =2a 1.193 1.557 1.422 1.272 0.427 0.427 0.427 6.727 0.214 6.941
DG DIGIT
Year Year Year Year Year Year Year TOTAL
MFF
2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
Operational appropriations
Budget line
Commitments (1a) 2.382 2.382 2.382
Payments (2a) 1.191 1.191 2.382 2.382
TOTAL
appropriations
for DG DIGIT
Commitments =1a 2.382 0.000 0.000 0.000 0.000 0.000 0.000 2.382 0.000 2.382
Payments =2a 1.191 1.191 0.000 0.000 0.000 0.000 0.000 2.382 0.000 2.382
From 2029, the VOBU request is the net amount for which the foreseen cloud federation and joint procurement fees have been deducted.
Total (DIGIT + CNECT) Year Year Year Year Year Year Year TOTAL
MFF
2028-2034
POST
2034
GRAND
TOTAL 2028 2029 2030 2031 2032 2033 2034
EN 16 EN
TOTAL
operational
appropriations
Commitments (4) 4.768 0.727 2.118 0.427 0.427 0.427 0.427 9.323 0.000 9.323
Payments (5) 2.384 2.748 1.422 1.272 0.427 0.427 0.427 9.109 0.214 9.323
TOTAL appropriations of an
administrative nature financed
from the envelope for specific
programmes
(6) 0 0 0 0 0 0 0 0 0 0
TOTAL
appropriations
under
HEADING 2
Commitments =4+6 4.768 0.727 2.118 0.427 0.427 0.427 0.427 9.323 0.000 9.323
of the
multiannual
financial
framework
Payments =5+6 2.384 2.748 1.422 1.272 0.427 0.427 0.427 9.109 0.214 9.323
Heading of multiannual financial framework 4 ‘Administrative expenditure’
EUR million (to three decimal places)
DG CNECT Year Year Year Year Year Year Year TOTAL MFF
2028-2034 2028 2029 2030 2031 2032 2033 2034
Human resources 1.810 1.810 1.810 1.810 1.810 1.810 1.810 12.670
Other administrative expenditure 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
TOTAL
DG
CNECT
Appropriations 1.810 1.810 1.810 1.810 1.810 1.810 1.810 12.670
DG: DIGIT Year Year Year Year Year Year Year TOTAL MFF
2028-2034 2028 2029 2030 2031 2032 2033 2034
Human resources 1.794 1.794 1.794 1.794 1.794 1.794 1.794 12.558
Other administrative expenditure 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
EN 17 EN
DG CNECT Year Year Year Year Year Year Year TOTAL MFF
2028-2034 2028 2029 2030 2031 2032 2033 2034
TOTAL
DG
DIGIT
Appropriations 1.794 1.794 1.794 1.794 1.794 1.794 1.794 12.558
TOTAL appropriations
under HEADING 4 of
the multiannual
financial framework
(Total
commitments
= Total
payments)
3.604 3.604 3.604 3.604 3.604 3.604 3.604 25.228
EUR million (to three decimal places)
Year Year Year Year Year Year Year TOTAL
MFF
2028-
2034
POST
2034
GRAND
TOTAL 2028 2029 2030 2031 2032 2033 2034
TOTAL appropriations
under HEADINGS 1 to
4
of the multiannual
financial framework
Commitments 8.372 4.331 5.722 4.031 4.031 4.031 4.031 34.551 0.000 34.551
Payments 5.988 6.352 5.026 4.876 4.031 4.031 4.031 34.337 0.214 34.551
3.2.1.2. Appropriations covered from fees (EuroCloud and Joint Cloud Procurement)
EUR million (to three decimal places)
Heading of multiannual financial framework Number
DG CNECT
Year Year Year Year Year Year Year TOTAL
MFF
2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
Operational appropriations financed by fees
Budget line Commitments (1a) 0.000 2.940 4.410 5.880 2.570 2.570 2.570 20.939 20.939
EN 18 EN
Payments (2a) 0.000 2.940 4.410 5.880 2.570 2.570 2.570 20.939 20.939
Appropriations of an administrative nature financed by fees
Budget line (3)
TOTAL
appropriations Commitments =1a+3 0.000 2.940 4.410 5.880 2.570 2.570 2.570 20.939 0.000 20.939
for DG
CNECT Payments =2a+3 0.000 2.940 4.410 5.880 2.570 2.570 2.570 20.939 20.939
DG DIGIT Year Year Year Year Year Year Year TOTAL
MFF 2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
Operational appropriations financed by fees
Budget line Commitments (1a) 0.000 3.027 4.317 6.037 6.342 6.663 7.001 33.387 33.387
Payments (2a) 0.000 3.027 4.317 6.037 6.342 6.663 7.001 33.387 33.387
Appropriations of an administrative nature financed from fees
Budget line (3) 0
0
TOTAL
appropriations Commitments =1a+3 0.000 3.027 4.317 6.037 6.342 6.663 7.001 33.387 0.000 33.387
for DG DIGIT Payments =2a+3 0.000 3.027 4.317 6.037 6.342 6.663 7.001 33.387 33.387
Year Year Year Year Year Year Year TOTAL MFF
2028-2034
POST
2034
GRAND
TOTAL 2028 2029 2030 2031 2032 2033 2034
TOTAL operational
appropriations
Commitments (4) 0.000 5.967 8.727 11.917 8.912 9.233 9.570 54.326 0.000 54.326
Payments (5) 0.000 5.967 8.727 11.917 8.912 9.233 9.570 54.326 54.326
EN 19 EN
DG DIGIT Year Year Year Year Year Year Year TOTAL
MFF 2028-
2034
POST
2034
GRAND
TOTAL
2028 2029 2030 2031 2032 2033 2034
TOTAL appropriations of an
administrative nature financed from
the envelope for specific programmes
(6)
TOTAL
appropriations under
HEADING 2
Commitments =4+6 0.000 5.967 8.727 11.917 8.912 9.233 9.570 54.326 0.000 54.326
of the multiannual
financial framework Payments =5+6 0.000 5.967 8.727 11.917 8.912 9.233 9.570 54.326 0.000 54.326
3.2.2. Estimated output funded from operational appropriations (not to be completed for decentralised agencies)
Commitment appropriations in EUR million (to three decimal places)
Indicate
objectives and
outputs
Year 2024
Year 2025
Year 2026
Year 2027
Enter as many years as necessary to show the
duration of the impact (see Section1.6) TOTAL
OUTPUTS
Type44
Avera
ge
cost
N o
Cost N o
Cost N o
Cost N
o
Cost N o
Cost N o
Cost N o
Cost Total
No
Total
cost
SPECIFIC OBJECTIVE No 145…
- Output
- Output
- Output
44 Outputs are products and services to be supplied (e.g. number of student exchanges financed, number of km of roads built, etc.). 45 As described in Section 1.3.2. ‘Specific objective(s)’
EN 20 EN
Subtotal for specific objective No 1
SPECIFIC OBJECTIVE No 2 ...
- Output
Subtotal for specific objective No 2
TOTALS
EN 21 EN
3.2.3. Summary of estimated impact on administrative appropriations
– The proposal/initiative does not require the use of appropriations of an
administrative nature
– The proposal/initiative requires the use of appropriations of an administrative
nature, as explained below
3.2.3.1. Appropriations from voted budget
VOTED
APPROPRIATIONS
Year Year Year Year Year Year Year TOTAL
2028 - 2034 2028 2029 2030 2031 2032 2033 2034
HEADING 4
Human resources 3.604 3.604 3.604 3.604 3.604 3.604 3.604 25.22Ff8
Other administrative
expenditure 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000
Subtotal HEADING 4 3.604 3.604 3.604 3.604 3.604 3.604 3.604 25.228
Outside HEADING 4
Human resources 0.000 0.000 0.000 0.000 0.0000.0000.000 0.000
Other expenditure of an
administrative nature 0.000 0.000 0.000 0.000 0.0000.0000.000 0.000
Subtotal outside
HEADING 4
TOTAL 3.604 3.604 3.604 3.604 3.604 3.604 3.604 25.228
The estimated impact on expenditure and staffing for 2028 and beyond is added for
illustrative purposes only and does not pre-judge the next Multiannual Financial Framework.
The source of financing and scope of Union financial commitment in the post-2027 period
remain subject to the outcome of interinstitutional negotiations on the MFF 2028-2034 and
thereafter shall be determined through the annual budgetary procedure. All appropriations and
staffing allocations as of 2028 are indicative.
3.2.3.2. Appropriations from external assigned revenues
Not applicable.
3.2.3.3. Total appropriations
See table above.
3.2.4. Estimated requirements of human resources
– The proposal/initiative does not require the use of human resources
– The proposal/initiative requires the use of human resources, as explained
below
3.2.4.1. Financed from voted budget
Estimate to be expressed in full-time equivalent units (FTEs)
VOTED APPROPRIATIONS Year Year Year Year Year Year Year
2028 2029 2030 2031 2032 2033 2034
Establishment plan posts (officials and temporary staff)
20 01 02 01 (Headquarters and Commission’s Representation Offices)
11 11 11 11 11 11 11
20 01 02 03 (EU Delegations) 0 0 0 0 0 0 0
01 01 01 01 (Indirect research) 0 0 0 0 0 0 0
EN 22 EN
01 01 01 11 (Direct research) 0 0 0 0 0 0 0
Other budget lines (specify) 0 0 0 0 0 0 0
• External staff (inFTEs)
20 02 01 (AC, END from the ‘global envelope’) 14 14 14 14 14 14 14
20 02 03 (AC, AL, END and JPD in the EU Delegations) 0 0 0 0 0 0 0
Admin. Support
line
[XX.01.YY.YY]
- at Headquarters 0 0 0 0 0 0 0
- in EU Delegations 0 0 0 0 0 0 0
01 01 01 02 (AC, END - Indirect research) 0 0 0 0 0 0 0
01 01 01 12 (AC, END - Direct research) 0 0 0 0 0 0 0
Other budget lines (specify) - Heading 7 0 0 0 0 0 0 0
Other budget lines (specify) - Outside Heading 7 0 0 0 0 0 0 0
TOTAL 25 25 25 25 25 25 25
The estimated impact on expenditure and staffing for 2028 and beyond is added for
illustrative purposes only and does not pre-judge the next Multiannual Financial Framework.
The source of financing and scope of Union financial commitment in the post-2027 period
remain subject to the outcome of interinstitutional negotiations on the MFF 2028-2034 and
thereafter shall be determined through the annual budgetary procedure. All appropriations and
staffing allocations as of 2028 are indicative.
3.2.4.2. Financed from external assigned revenues
Not applicable.
3.2.4.3. Total requirements of human resources
The staff required to implement the proposal (in FTEs):
To be covered by
current staff
available in the
Commission
services
Exceptional additional staff*
To be financed
under Heading 7
or Research
To be financed
from BA line
To be financed
from fees
Establishment
plan posts 5 6
N/A
External staff
(CA, SNEs, INT) 10 4
* The estimated impact on expenditure and staffing for 2028 and beyond is indicative and does not
pre-judge the next Multiannual Financial Framework. The source of financing and scope of Union
financial commitment in the post-2027 period remain subject to the outcome of interinstitutional
negotiations on the MFF 2028-2034 and annual budgetary procedure and the steering mechanism.
Overall, 15 FTEs requested for the initiative are already in place and will be redeployed with the
following repatriation:
– DG CNECT: 2 FTEs in establishment plan posts and 5 FTEs in external staff
EN 23 EN
– DG DIGIT: 3 FTEs in establishment plan posts and 5 FTEs in external staff
In addition to these existing resources, the initiative requires 10 FTEs of exceptional additional staff.
These consist of 6 officials and 4 contract agents and have the following split between DG CNECT
and DG DIGIT:
– DG CNECT: 3 FTEs in establishment plan posts and 3 FTEs in external staff
– DG DIGIT: 3 FTEs in establishment plan posts and 1 FTE in external staff
These are requested on top of the current staffing levels to ensure full and effective implementation of
the initiative. The new tasks introduced by the proposal cannot be absorbed by the respective DGs’
existing human resources. The additional human resources required for this proposal cannot be
covered by redeployments within the DG/service or from redeployments from the limited Commission
redeployment pool.
Description of new tasks to be carried out by DG CNECT and DG DIGIT:
Officials and temporary staff Official and temporary staff will be tasked to:
1. Determine and manage the Work Programmes under Pillar 1 of the initiative,
including the development of detailed project plans, budgets, resource allocation,
KPIs and monitoring mechanisms.
2. Produce guidance documents, delegated acts, dependency assessments,
comitology secretariat, and oversee the implementation of the initiative, including:
a. Development of comprehensive guidelines, recommendations, templates
and tools to support Member States in their implementation efforts on
Pillar 2 and Pillar 3 of the initiative
b. Development of implementing acts, where needed, e.g. on the
requirements applicable to the members of the EuroCloud Federation and
the procedure to assess their application
c. Drafting market monitoring reports to identify cloud computing
services where the Union has a high level of dependence on a single or
limited number of third country legal entities for such services
d. Providing secretariat services including the development of documents
and reports, coordination with stakeholders
e. Monitoring progress of the initiative against the established KPIs and
milestones, identifying possible issues and risks and providing technical
assistence and support to Member States as needed
f. Establishment of a knowledge management system to capture and share
best practices, lessons leadrned and expertise across the initiative
3. Draft tender specifications for an external provider to build, develop and maintain
the repository of sovereign services, and evaluate and determine subsequent
governance procedures. After setup, regularly audit/assess a sample of certified
services, which are part of the repository, including the development of audit
protocols and assessment methodologies and/or quality control procedures to
ensure the reliability of the information provided.
4. Procure the EuroCloud Federation platform and support the setup of the
federation. Manage the governance mechanisms of the federation, e.g.
certification of requirements for membership in the federation, evaluation of
applications from Member States and other public sector bodies to join the
federation, development and enforcement of common rules and policies for
participation in the federation.
5. Setting up systems not currently in place, as part of the joint procurement
framework, including managing the complexity of AI procurement, relationships
with Member States and contracting authorities and ensuring that the framework
is effective and efficient. This will include tasks such as, setting up and managing
the procurement processes for cloud, software, and AI systems, coordinating with
participating organisations and service providers, managing the contracts and
agreements with service providers, ensuring compliance with EU regulations and
policies.
EN 24 EN
6. Develop and maintain relationships with stakeholders, including Member States,
industry representatives, and other relevant parties, to ensure that the initiative is
well-coordinated, effective, and responsive to the needs of its stakeholders,
including the organisation of regular meetings, workshops, and conferences to
facilitate communication and collaboration. This would also include close
cooperation with competent authorities on investigative and enforcement
measures to ensure compliance with Title IV of the Act.
External staff External staff will be tasked to:
1. Regularly manage the projects awarded under Pillar 1 work programmes,
including the monitoring of their progress, coordination with beneficiaries, and
verification of project deliverables and milestones to ensure they are delivered on
time, within budget and with the required quality standards. The initiative will be
implemented in the next MFF and will build on the results of existing projects
already approved under Horizon Europe and the Digital Europe Programme, e.g.
Simpl and the Smart Networks and Services Joint Undertaking, including in
coordination with other Units.
2. Establish and operate support mechanisms to provide guidance to Member States
for the deployment of data centre acceleration zones, particularly during the first
two years of the initiative. This will involve the provision of ad hoc support to
Member States as they develop and implement their national strategies and plans.
After the initial period, this will transition to a more ad hoc coordination role,
focusing on providing targeted support to Member States as needed, and ensuring
that the initiative remains aligned with the needs and priorities of its stakeholders.
3. Set up and manage a study each year to monitor the computing capacity across
Europe, including setting out the scope and methodology of the study, verifying
the data collected, reviewing deliverables.
4. Manage the joint procurement activities, in particular in relation to strategic
sourcing and market shaping, i.e. vendor management office and drafting of
specifications and technical requirements for cloud, AI and software services.
5. Manage the repository of sovereign services on a yearly basis, including by
overseeing the contractor activities related to the repository’s operations,
including technical project management, i.e. ensuring that the contractor delivers
the repository according to the agreed technical specifications, timelines, and
budget, while also monitoring the contractor's processes to ensure that the
repository meets the required standards.
6. Manage the external contractor covering the administrative, technical and
operational activities of the EuroCloud Federation platform each year, as well as
the technical aspects of the EuroCloud Federation platform, including oversight of
technical architecture and infrastructure.
3.2.5. Overview of estimated impact on digital technology-related investments
Compulsory: the best estimate of the digital technology-related investments entailed
by the proposal/initiative should be included in the table below.
Exceptionally, when required for the implementation of the proposal/initiative, the
appropriations under Heading 7 should be presented in the designated line.
The appropriations under Headings 1-6 should be reflected as “Policy IT expenditure
on operational programmes”. This expenditure refers to the operational budget to be
used to re-use/ buy/ develop IT platforms/ tools directly linked to the implementation
of the initiative and their associated investments (e.g. licences, studies, data storage
etc). The information provided in this table should be consistent with details
presented under Section 4 “Digital dimensions”.
EN 25 EN
TOTAL Digital
and IT
appropriations
Year Year Year Year Year Year Year TOTAL
MFF
2028 -
2034 2028 2029 2030 2031 2032 2033 2034
HEADING 4
IT expenditure (corporate)
0 0 0 0 0 0 0 0
Subtotal
HEADING 4 0 0 0 0 0 0 0 0
Outside HEADING 4
Policy IT expenditure on operational programmes
0 0 0 0 0 0 0 0
Subtotal outside
HEADING 4 0 0 0 0 0 0 0 0
TOTAL 0 0 0 0 0 0 0 0
3.2.6. Compatibility with the current multiannual financial framework
The proposal/initiative:
– can be fully financed through redeployment within the relevant heading of the
multiannual financial framework (MFF)
Without prejudice to the negotiations on the next MFF, the appropriations allocated
to the Cloud and AI Development Act related activities from 2028 onwards will be
covered via redeployments from ECF under the 2028-2034 MFF and partially from
fees.
– requires use of the unallocated margin under the relevant heading of the MFF
and/or use of the special instruments as defined in the MFF Regulation
– requires a revision of the MFF
3.2.7. Third-party contributions
The proposal/initiative:
– does not provide for co-financing by third parties
– provides for the co-financing by third parties estimated below:
Appropriations in EUR million (to three decimal places)
Year 2028
Year 2029
Year 2030
Year 2031
Year 2032
Year 2033
Year 2034
Total
Specify the co-
financing body
TOTAL
appropriations co-
financed
EN 26 EN
3.3. Estimated impact on revenue
– The proposal/initiative has no financial impact on revenue.
– The proposal/initiative has the following financial impact:
– on own resources
– on other revenue
– please indicate, if the revenue is assigned to expenditure lines
EUR million (to three decimal places)
Budget
revenue line:
Appropriation
s available for
the current
financial year
Impact of the proposal/initiative 46
Year
2028
Year
2029
Year
2030
Year
2031
Year
2032
Year
2033
Year
2034
Joint Cloud
Procurement
service fee
(DG DIGIT)
0.000 3.027 4.317 6.037 6.342 6.663 7.001
Cloud
Federation fee
(DG CNECT)
0.000 2.940 4.410 5.880 2.570 2.570 2.570
For assigned revenue, specify the budget expenditure line(s) affected.
Joint Cloud procurement service fee:
The assigned revenue will be used to cover the budget expenditure linked to the
brokerage system put in place for the Joint Procurement scheme. The revenue will be
allocated to support the operational costs of the system, including the costs of
external staff, tools, and systems.
The breakdown of the assigned revenue will be on average as follows: around 50%
will be allocated to cover operational and administrative costs, e.g. studies and
research to inform procurement decisions, community management tools and
systems to support communication and collaboration among participating
organisations, technical enablers and platform integration; around 30% will be
allocated to cover the costs of developing and maintaining the necessary tools and
systems, such as procurement platforms and marketplaces, contract life-cycle
management systems, portfolio and demand management tools; around 20% will be
allocated to cover the costs of additional specialised external staff, e.g. for software
procurement, data protection activities, Back-office, FinOps and ContractOps to
manage contractual aspects, portfolio and demand management.
This allocation aims to ensure that the assigned revenue is used to cover only the
operational costs of the system and the budget expenditure linked to the brokerage
system. These costs will be allocated based on effective demand needs, ensuring that
the resources are used efficiently to support the Joint Procurement scheme.
Cloud Federation Fee:
46 As regards traditional own resources (customs duties, sugar levies), the amounts indicated must be net
amounts, i.e. gross amounts after deduction of 20% for collection costs.
EN 27 EN
The assigned revenue will be used to cover the budget expenditure linked to the
services and platform put in place for the EuroCloud Federation. The revenue will be
allocated to support the operational costs of the system, including the costs of
external staff, tools, and systems.
The expenditure required to finance the federation consists of operational and
development-related costs. A significant share relates to the development and
maintenance of the platform. Development costs are expected to be highest during
the first years, reflecting the initial deployment phase. From year 4 onwards, costs
would cover ongoing operations, technical support and periodic upgrades. The
maintenance contract would include quality assurance and performance assessment
carried out by an external contractor, estimated at around 10% of the total
procurement value, based on similar procurement contracts. Operational costs
include the cost of external technical experts (10 profiles) on top of the internal FTEs
which would be responsible for coordination and project management (see above
under the description of new tasks to be carried out by DG CNECT). Their
involvement would ramp up progressively, with approximately 40% of capacity in
year 1, 80% in year 2, and reaching full capacity from year 3 onwards, in line with
the increasing operational needs of the federation. Further expenditure includes IT
infrastructure costs, such as hosting services and security-related tooling, which are
necessary to ensure the reliability, scalability and protection of the platform.
This allocation aims to ensure that the assigned revenue is used to cover the
operational costs of the system and the budget expenditure linked to the services
provided. These costs will be allocated based on effective demand needs, ensuring
that the resources are used efficiently to support the EuroCloud Federation. Initial
establishment costs may be borne by the general budget of the Union and reimbursed
by the participating authorities over a defined period through the fees.
Other remarks (e.g. method/formula used for calculating the impact on revenue or
any other information).
Joint Cloud procurement service fee:
The fee is meant to cover the operational costs for the procurement activities
(interinstitutional procurement, joint procurement and central purchasing activities)
and the possible ancillary procurement activities.
The payers of the fee would be participating contracting authorities. The fee would
increase the price for contracting authorities. However, this will be set in a way to
cover only the necessary costs for administering the procurement activities, as the
objective of this common procurement would be to be able to negotiate better
conditions, lower prices for all contracting authorities and facilitating the adoption at
scale of technologies that evolve constantly.
The fees are calculated pro-rata per transaction or via a flat fee per subsequent
awarded contract, as specified in the procurement document. The fees will be
computed based on a yearly forecast of the beneficiaries’ consumption and will be
perceived yearly before the execution of the contracts. The fees charged to the
participating contracting authorities shall not exceed the verifiable costs incurred by
the Commission. It shall be recomputed each year, proportionally to the volume of
work foreseen and costs incurred. The starting data of collection of the fee would be
2029, to allow for an initial setup process. The fee will be collected through existing
EN 28 EN
online services for invoicing and collecting payments from participating contracting
authorities.
Projected annual revenue depends on both the number of participating authorities
(beneficiaries) and their average expected purchases. The fee revenue will need to
reach approximately EUR 6 million per year to cover the forecast average annual
costs. Based on estimates on participation, it should be reachable with an average
annual fee rate of 2% applied to the spending per authority. If total spending exceeds
this level, e.g. because more authorities participate, average spending per authority is
higher, or both, the fee rate would decrease proportionately. Revenue from the fee
will be used to finance the entire operational costs of the procurement mechanism,
explained above.
The Commission shall report to the European Parliament and the Council on the
overall amount of the costs incurred for the procurement activities and the total
amount of the fees charged.
Cloud Federation fee:
The fee is meant to cover the operational costs for establishing and managing the
EuroCloud federation, including assessing membership applications and facilitating
the sharing of data centre services and cloud computing services through the
development of the EuroCloud Federation platform.
The payers of the fee would be members of the federation. The fee would slightly
increase the overall costs for participants. However, it shall be calibrated strictly to
cover the essential administrative and operational costs of running the federation.
This is expected to create efficiencies and cost savings. By enabling members to
exchange capacity and make use of otherwise idle resources, participants can access
services at more favourable rates than those available on the open market. This
shared model can reduce the need for external procurement and improve overall
resource utilisation, especially in the short to medium term. While the fee introduces
a marginal cost, it is balanced by the economic benefits derived from collaboration,
improved capacity usage and more competitive pricing within the federation.
The Commission shall adopt implementing acts laying down detailed rules relating to
determining the fees to be levied by the Commission, specifying the estimated costs
attributable to the activities for which fees are chargeable, the individual fee amounts
chargeable, as well as the ways and conditions under which the fees should be paid.
The membership fee would be structured as a cost-recovery mechanism,
proportionate to the number of participating entities in the federation. The goal is to
cover costs while progressively reducing the financial burden on each member as
participation increases.
The uptake is assumed to grow over time, with approximately 20% of possible
entities joining in the first year, 40% in the second year, 60% in the third year,
around 80% in year 4 before achieving full participation in the fifth year. The first
year would be dedicated to setting up the federation so no full membership fee would
apply. Based on the estimations done, from year 2 onward, the membership fee could
be set at around EUR 75 000 per member, reflecting the relatively low number of
participants. As more entities join and the cost base is shared across a larger group,
the fee could gradually decrease and reach around EUR 30 000 per member once full
capacity is achieved. The scheme is expected to reach a break-even point around year
4. At that stage, the fee structure would stabilise at a level sufficient to sustain
EN 29 EN
operations without generating considerable surplus, in line with the cost-recovery
principle.
The starting data of collection of the fee would be 2029, to allow for an initial setup
process. The fee will be collected through online services for invoicing and
collecting payments from participating authorities.
The estimated annual revenue would depend on the number of participating members
and the applicable membership fee. Based on an average membership fee of around
EUR 75 000 per member, annual revenues would reach around EUR 4.4 million per
year to cover the forecasted average annual operating costs of the federation,
including the initial fixed costs. As participation increases, the costs of administering
the federation would be distributed across a larger number of members.
Consequently, the fee per member would decrease proportionately, with annual
revenues stabilising at a lower level, e.g. of around EUR 2.6 million per year.
The Commission shall report to the European Parliament and the Council on the
overall amount of the costs incurred for the federation activities and the total amount
of the fees charged. The period for reporting is still to be determined.
The use of assigned revenue for the joint procurement and EuroCloud federation
offers a budgetary policy opportunity to ensure a dedicated and stable funding stream
for these initiatives. By earmarking the revenue from the fee for specific purposes,
the Commission can guarantee that the funds are used efficiently and effectively to
support the development of a European cloud ecosystem. This approach is justified
by the need to provide a clear and predictable financial framework for both initiatives
under CADA, which will allow to plan and implement them in a strategic and long-
term manner.
In terms of compliance with the budgetary principles of universality and unity, the
use of assigned revenue is justified by the fact that these initiatives are designed to
support specific and clearly defined policy objectives, i.e. the development of a
European cloud ecosystem for sharing idle capacity among Member States and the
joint procurement of cloud services to achieve efficiencies and economies of scale.
The revenues from the fees will be used to finance specific items of expenditure that
are directly related to the achievement of these objectives, as specified above in the
respective sections covering both activities. By using assigned revenue, the
Commission can ensure that the funds are used in a targeted and efficient manner,
without affecting the overall balance of the Union's budget. Furthermore, the use of
assigned revenue is consistent with the principle of unity, as it will enable the
Commission to implement these initiatives in a coherent and coordinated manner,
avoiding duplication of efforts and ensuring that the funds are used to support a
common European objective. The Commission will also ensure that the use of
assigned revenue is transparent and accountable, with regular reporting and
evaluation of the initiative's progress and impact.
In addition, the Commission will take into account the need to avoid any potential
distortions or inequalities in the treatment of different cloud service providers, and
will ensure that the use of assigned revenue is fair, proportionate, and non-
discriminatory. The Commission will also consult with relevant stakeholders to
ensure that the initiative is designed and implemented in a way that is responsive to
their needs and concerns.
EN 30 EN
It is important to note that only the operational costs of the two initatives, as outlined
above, will draw on the assigned revenue generated by the fees. Administrative costs,
on the other hand, will not be covered by the assigned revenue.
In the event that the fees do not materialise in the amount and timing assumed, the
Commission has taken a cautious approach to mitigate the risks. In the case of joint
procurement, if the fees do not materialise, it would mean that no system would be
put in place, and the initiative would not incur any significant costs. On the other
hand, the EuroCloud Federation builds on pre-existing studies and current evidence
of Member State interest in this initiative. Moreover, the major costs associated with
the initaitive are already covered under the current Multiannual Financial Framework
under the budget already assigned to the Commission, which altogether reduces this
financial risk. Overall, this approach will allow the Commission to adjust its plans,
without incurring significant financial risks and be more flexible to adapt its resource
allocation as needed, in order to ensure the success of these initiatives.
EN 31 EN
4. DIGITAL DIMENSIONS
4.1. Requirements of digital relevance
High-level description of the requirements of digital relevance and related categories (data, process digitalisation & automation, digital solutions,
and/or digital public services).
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
TITLE II
RESEARCH,
DEVELOPMENT
AND
DEPLOYMENT
ACTIVITIES FOR
THE CLOUD AND
AI ECOSYSTEM
Article 7 National
cloud and AI
strategies
Member States shall notify the Commission of their
national strategies.
European
Commission
National competent
authorities
Notification Data
TITLE III DATA
CENTRE
CAPACITIES
Article 14
Designation of data
centre strategic
projects
Applications for designation of data centre strategic
project shall provide all the necessary and relevant
information to demonstrate that the project fulfils the
relevant criteria.
Project operators
European
Commission
Application
insights monitoring Data
EN 32 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
TITLE III DATA
CENTRE
CAPACITIES
Article 15 Monitoring
the Capacity Gap
The Commission shall monitor thecompute capacity
available in the Union, including edge computing
capacity; the volume of demand for data centre capacity;
the size of the capacity gap.
European
Commission
National competent
authorities
Data collection Data
TITLE IV AUTONOMY
Article 17 Recognition
of cloud computing
service providers
A cloud computing service provider that aims to be
recognised as offering a Union assurance level, shall
submit an application for recognition to the national
competent authority of establishment.
Cloud computing
service providers
National competent
authorities
Oversight
mechanism Data
TITLE IV AUTONOMY
Article 19 Conformity
self-assessment
The cloud computing service provider may issue a
conformity self-assessment to demonstrate compliance
with Union assurance level 1. These are to be made
publicly available.
Cloud computing
service providers
National competent
authorities
European
Commission
Oversight
mechanismData
TITLE IV AUTONOMY
Article 20
Independent audit
Establishes the possibility of independent audits for
Union assurance levels 2, 3, and 4. Provisions related to
data aspects in the context of such audits.
Cloud computing
service providers
Auditing
organisations
European
Oversight
mechanism Data
EN 33 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
Commission
TITLE IV AUTONOMY
Article 21 Content
and quality of audit
evidence
Audits to be based on Annexes II (Criteria for Union
Assurance Levels) and III (Audit evidence for the audit
procedure). To be sufficiently complete and reliable.
Cloud computing
service providers
Auditing
organisations
European
Commission
Data quality Data
TITLE IV AUTONOMY
Article 22 Central
repository of cloud
computing services
The Commission shall establish and maintain a
dedicated repository of cloud computing services that
have received recognition against a Union assurance
level.
The verifying national competent authority of
establishment shall register the audited services in the
central repository. Any revocation of a recognition
shall be published, and remain published, in the
central register for 5 years.
The central repository shall be made publicly available
by the Commission and regularly updated by the
Commission and the national competent authorities of
establishment on a dedicated and easily accessible
website.
European
Commission
National competent
authorities
Repository
Digital solution
Data
Digital public
service
EN 34 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
TITLE IV AUTONOMY
Article 23
Transparency
obligations
On becoming aware of information or material change in
circumstances concerning the cloud computing services
that may affect the audit report / opinion, the service
provider shall without undue delay notify the auditing
organisation and the national competent authority of
establishment.
If the results of the audit report and opinion need to be
amended or cancelled, the auditing organisation shall
without undue delay notify the national competent
authority and the Commission.
If the results of the recognition need to be amended or
cancelled, the national competent authority of
establishment shall without undue delay notify the
national competent authorities of the other Member
States and the Commission.
Cloud computing
service providers
Auditing
organisations
National competent
authorities
European
Commission
Notification Data
TITLE IV AUTONOMY
Article 24
Penalties and
compensation
Member States shall notify, as soon as possible, notify to
the Commission the rules on penalties and compensation
and any subsequent amendment affecting them.
National competent
authorities
European
Commission
Notification Data
EN 35 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
TITLE IV AUTONOMY
Article 25 National
competent authorities
Each Member States shall designate one or more
national competent authorities. Members States shall
notify the list of authorities to the Commission. The
Commission shall maintain a public register of the
authorities.
European
Commission
National competent
authorities
Designation of
competent
authorities
Data
TITLE IV AUTONOMY
Article 27 Mutual
assistance
Data flows in the context of cooperation between
national competent authorities.
National competent
authorities
European
Commission
Data exchange Data
TITLE IV AUTONOMY
Article 89 Cross-
border cooperation
Data flows in the context of cross-border cooperation.
National competent
authorities
European
Commission
Data exchange Data
TITLE IV AUTONOMY
Article 29 Risk
Assessments
Member States must perform a risk assessment or
assessments, taking utmost account of the guidance
issued by the Commission. Member States shall
communicate the risk assessment results to the
Commission.
For the purpose of the guidance, the Commission is
entitled to request information from cloud service
providers.
Member States
European
Commission
Cloud service
providers
Risk assessment Data
TITLE IV Member States shall communicate annually to the
Commission certain information pertaining to the Member States Reporting Data
EN 36 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
AUTONOMY
Article 33 Monitoring
of procurement of
innovation in cloud
and AI
monitoring of procurement of innovative cloud
computing services and AI systems.
European
Commission
TITLE IV AUTONOMY
Article 34
Establishment of the
European public
sector cloud
federation
The Commission shall establish a platform for the
EuroCloud Federation.
European
Commission
Union entities
Public sector
bodies
Requirements for a
digital solution
Digital public
service
Digital solution
TITLE IV
AUTONOMY
Article 37
Procurement
activities of the
Commission
The Commission may establish and manage a common
procurement platform including services which may be
used for facilitating the performance of the procurement
activities under this Chapter.
European
Commission
Requirements for a
digital solution Digital solution
TITLE IV AUTONOMY
Article 42 Share and
reuse of software
When making available for reuse under an open-source
licence software on which they hold intellectual property
rights, a Union entity or public sector body shall make
the software available for reuse on a catalogue or
repository that is connected to the EU OSS Catalogue.
Union entities
Public sector
bodies
European
Commission
Catalogue Digital solution
EN 37 EN
Reference to the
requirement Requirement description
Actors affected or
concerned by the
requirement
High-level
processes Categories
TITLE IV AUTONOMY
Article 43 EU Open
Source Solutions
Catalogue
The Commission shall provide and maintain an EU
Open Source Solutions Catalogue as a centralised
catalogue to access software made available for reuse
by Union entities and public sector bodies
The EU OSS Catalogue shall be hosted on the
Interoperable Europe portal.
The Commission shall decide on the request of any
Union entity or public sector body owning or
maintaining a catalogue or repository to connect to
and be made accessible through the EU OSS
Catalogue.
European
Commission Catalogue Digital solution
4.2. Data
High level description of the data in scope
Type of data Reference to the requirement(s) Standard and/or specification (if
applicable)
Member State National Strategies
TITLE II RESEARCH, DEVELOPMENT AND
DEPLOYMENT ACTIVITIES FOR THE CLOUD AND
AI ECOSYSTEM
Article 7 National cloud and AI strategies
Minimum content requirements for
the national strategies.
Data related to applications for recognition
as a strategic project
TITLE III DATA CENTRE CAPACITIES
Article 14 Designation of data centre strategic projects
Provide all the necessary and
relevant information to demonstrate
that the project fulfils the relevant
EN 38 EN
criteria.
Statistics and data required for the
monitoring of the capacity gap
TITLE III DATA CENTRE CAPACITIES
Article 15 Monitoring the Capacity Gap N/A
Data in the context of the submissions of
applications by service providers that aim to
be recognised as offering a Union assurance
level
TITLE IV AUTONOMY
Article 17 Recognition of cloud computing service
providers
N/A
EU statement of conformity TITLE IV AUTONOMY
Article 19 Conformity self-assessment N/A
Audit report submitted for recognition and
related exchanges
TITLE IV AUTONOMY
Article 20 Independent audit
Article 21 Content and quality of audit evidence
Data quality: sufficiently complete
and reliable; minimum content for
audits.
Data on services that receive a recognition TITLE IV AUTONOMY
Article 22 Central repository of cloud computing services
The central repository shall contain
information of cloud computing
services that have been recognised
under a specific Union assurance
levels
Notification of becoming aware of
information or material change in
circumstances concerning the cloud
computing services that may affect the audit
report / opinion / recognition and related
data flows
TITLE IV AUTONOMY
Article 23 Transparency obligations N/A
EN 39 EN
Notification of rules and measures on
penalties
TITLE IV AUTONOMY
Article 24 Penalties and compensation N/A
Data on the designated national competent
authorities
TITLE IV AUTONOMY
Article 25 National competent authorities N/A
Data used in the context of cooperation
between national competent authorities
TITLE IV AUTONOMY
Article 27 Mutual assistance N/A
Data used in the context of cross-border
cooperation
TITLE IV AUTONOMY
Article 28 Cross-border cooperation
N/A
Risk assessments performed by Member
States
TITLE IV AUTONOMY
Article 29 Risk Assessments
N/A
Information provided to the Commission by
providers for guidance
TITLE IV AUTONOMY
Article 29 Risk Assessments
N/A
Data pertaining to the monitoring of
procurement of innovative cloud services
and AI systems
TITLE IV AUTONOMY
Article 33 Innovation procurement
N/A
Alignment with the European Data Strategy
Explanation of how the requirement(s) are aligned with the European Data Strategy
EN 40 EN
The proposal is consistent with the rules on switching between data processing services
introduced by the Data Act. By enabling switching and removing key sources of vendor
lock-in, the Data Act seeks to ensure that cloud service providers in the EU compete on
quality, innovation, and price. It seeks to enable cloud users to freely choose the provider
that best meets their needs and combine offers of different providers in a multi-cloud
approach. However, the Data Act does not contain elements to shape up a more competitive
offer of EU cloud services or encourage the entry into the market of a more diverse set of
cloud service providers. The Data Act opens the path towards a possible reduction of
dependencies on non-EU providers but does not build the road towards a more sovereign
and trusted EU cloud computing sector. The cloud switching and interoperability
provisions, however, make it possible for users to embrace European cloud computing
services more strongly. The Data Act is thus an enabler for the proposal.
Alignment with the once-only principle
Explanation of how the once-only principle has been considered and how the possibility to reuse existing data has been explored
The once-only principle has been duly considered and will systematically be enforced
wherever relevant. This is in particular the case for:
(a) The statistics and data required for the monitoring of the capacity gap
(b) The data on services that receive a recognition
(c) The information provided to the Commission for market analysis
Explanation of how newly created data is findable, accessible, interoperable and reusable, and meets high-quality standards
Digital solutions will be provided to ensure the findability, accessibility, interoperability,
and reusability of newly created data. Minimum content requirements will promote high-
EN 41 EN
quality standards for the data.
Data flows
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
Member State
National Strategies TITLE II RESEARCH,
DEVELOPMENT AND
DEPLOYMENT
ACTIVITIES FOR THE
CLOUD AND AI
ECOSYSTEM
Article 7 National cloud and
AI strategies
Member States European
Commission
Within three months of
the adoption of a
national strategy
Per adoption/
revision
Data related to
applications for
recognition as a
strategic project
TITLE III DATA CENTRE
CAPACITIES
Article 14 Designation of
data centre strategic projects
Applicant for
designation of a
strategic project
European
Commission
Application during
open calls for
expression of interest
//
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies
TITLE IV AUTONOMY
Article 17 Recognition of
cloud computing service
providers
Requesting cloud
computing service
provider
National
competent
authority of their
establishment
Request being made to
provide services to
Union entities and
public sector bodies
//
EN 42 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies:
Outcome of request
(reject, ask for
additional evidence,
recognise the
provider’s service)
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
National competent
authority
Requesting cloud
computing service
provider
Conclusion reached on
initial recognition
//
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies:
Distribution of
conclusion for
review
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
National competent
authority
All national
competent
authorities
Cloud computing
service provider
notified of a recognition
//
Data in the context
of the submissions of
requests by cloud
computing services
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
All national
competent
authorities
National
competent
authority
National competent
authority has objections
to a conclusion made by
a national competent
//
EN 43 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
wishing to provide
services to Union
entities and public
sector bodies:
Objections
providers authority of
establishment
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies: Cases
referred to the
Commission
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
National competent
authorities
European
Commission
National competent
authorities cannot reach
an agreement
//
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies:
Commission
decision on a
referred case
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
European
Commission
National
competent
authorities
Commission reaches a
decision on a case
where national
competent authorities
were in disagreement
//
Data in the context
of the submissions of TITLE IV
National competent
authorities
European
Commission
National competent
authorities decide to use
//
EN 44 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies:
Information
request
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
their right to an
information request
Data in the context
of the submissions of
requests by cloud
computing services
wishing to provide
services to Union
entities and public
sector bodies: Reply
to an information
request
TITLE IV
AUTONOMY
Article 17 Recognition of
cloud computing service
providers
European
Commission
National
competent
authorities
Information request
from national
competent authorities
received
//
EU statement of
conformity TITLE IV
AUTONOMY
Article 19 Conformity self-
assessment
Cloud computing
service providers
Public Self-assessment used //
Audit report
submitted for
validation and
related exchanges:
information
TITLE IV
AUTONOMY
Article 20 Independent audit
Article 21 Content and
Cloud computing
service providers
Auditing
organisation
Audit being performed //
EN 45 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
required for the
audit
quality of audit evidence
Audit report
submitted for
validation and
related exchanges:
audit reports
TITLE IV
AUTONOMY
Article 20 Independent audit
Article 21 Content and
quality of audit evidence
Cloud computing
service providers
Auditing
organisation and
national
competent
authorities
Audit finished //
Data on services that
receive a
recognition: Central
repository of cloud
computing services
TITLE IV
AUTONOMY
Article 22 Central repository
of cloud computing services
Verifying national
competent
authority
Central repository
of cloud
computing
services
(European
Commission)
Recognition granted //
Data on services that
receive a
recognition:
Publicly available
website provided
by the European
Commission
TITLE IV
AUTONOMY
Article 22 Central repository
of cloud computing services
European
Commission
National competent
authority of
establishment
Public Recognition granted //
Notification of
becoming aware of
information or
material change in
circumstances
concerning the cloud
computing services
TITLE IV
AUTONOMY
Article 23 Transparency
obligations
Cloud computing
service providers
Auditing
organisation
National
competent
authority
European
Commission
Material conditions
change on the side of
the provider
//
EN 46 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
that may affect the
audit report / opinion
and related data
flows: Initial
notification
Notification of
becoming aware of
information or
material change in
circumstances
concerning the cloud
computing services
that may affect the
audit report / opinion
and related data
flows: Notification
of the audit
outcome changing
TITLE IV AUTONOMY
Article 23 Transparency
obligations
Auditing
organisation
National
competent
authorities (of the
establishment and
of other Member
States)
European
Commission
Outcome of audit
changing
//
Notification of rules
and measures on
penalties
TITLE IV
AUTONOMY
Article 24 Penalties and
compensation
Member States European
Commission
Member States
set/amend the
framework for
imposing penalties
//
Data on the
designated national
competent
authorities:
Notification by
TITLE IV
AUTONOMY
Article 25 National competent
authorities
Member States European
Commission
National competent
authorities designated
(within a year of date of
entry into force)
//
EN 47 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
Member States
Data on the
designated national
competent
authorities: Data in
the public register
of authorities
TITLE IV
AUTONOMY
Article 25 National competent
authorities
European
Commission
Public National competent
authorities designated
and Commission
notified
//
Data used in the
context of
cooperation between
national competent
authorities:
Information
Exchange
TITLE IV
AUTONOMY
Article 27 Mutual assistance
National competent
authorities
National
competent
authorities
Cooperation (exchange
of information) between
national competent
authorities needed
//
Data used in the
context of
cooperation between
national competent
authorities:
Requests for
assistance (and
replies to such
requests)
TITLE IV
AUTONOMY
Article 27 Mutual assistance
National competent
authorities
National
competent
authorities
Assistance deemed
necessary
//
Data used in the
context of cross-
border cooperation:
Request for
assessment –
TITLE IV
AUTONOMY
Article 28 Cross-border
cooperation
National competent
authorities of a
service’s
destination
National
competent
authorities of a
service’s origin
Suspicions of non-
compliance giving rise
to a request for
assistance
//
EN 48 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
suspicion of non-
compliance
Data used in the
context of cross-
border cooperation:
Commission
informed of
suspicions of non-
compliance
TITLE IV
AUTONOMY
Article 28 Cross-border
cooperation
National competent
authorities of a
service’s
destination
European
Commission
A competent authority
decide to ask the
competent authority of
establishment for
information to assess
suspicions of non-
compliance
//
Data used in the
context of cross-
border cooperation:
Request for
assessment –
suspicion of non-
compliance
(Commission)
TITLE IV
AUTONOMY
Article 28 Cross-border
cooperation
National competent
authorities of a
service’s origin
European
Commission
National
competent
authorities
Suspicion of
noncompliance by other
competent authorities or
the Commission
//
Data used in the
context of cross-
border cooperation:
Outcome of a
request for
assessment –
suspicion of non-
compliance
TITLE IV
AUTONOMY
Article 28 Cross-border
cooperation
Competent
authorities of origin
Requestors
European
Commission
Within two months of a
request being made
//
Risk assessments
performed by
Member States
TITLE IV
AUTONOMY
Member States European
Commission
Risk assessment
occurring within one
year of entry into force
Bi-annually
EN 49 EN
Type of data Reference to the
requirement(s)
Actors who
provide the data
Actors who
receive the data
Trigger for the data
exchange
Frequency (if
applicable)
Article 29 Risk Assessments
Information
provided to the
Commission for
guidance
TITLE IV
AUTONOMY
Article 29 Risk Assessments
Cloud service
providers
European
Commission
// //
Data pertaining to
the monitoring of the
procurement of
innovative cloud
services and AI
systems
TITLE IV
AUTONOMY
Article 33 Monitoring of
procurement of innovation in
cloud and AI
Member States European
Commission
// Annually
4.3. Digital solutions
Repository of recognised cloud computing services
Digital and/or sectorial policy (when these are
applicable)
Explanation on how it aligns
AI Act Not applicable
EU Cybersecurity framework The repository will follow cybersecurity best practices of the Commission
eIDAS The repository will re-use, in so far as relevant, the eIDAS framework
Single Digital Gateway and IMI Not applicable
EN 50 EN
Others (e.g., Interoperable Europe Act) The need for another interoperability assessment under the Interoperable Europe
Act (Regulation EU 2024/903) will be evaluated once the operational details for
the repository of recognised cloud computing services become available.
EuroCloud platform
Digital and/or sectorial policy (when these are
applicable)
Explanation on how it aligns
AI Act Not applicable
EU Cybersecurity framework The platform should include mechanisms for secure access and incident
management, such as shared identity management, mutual authentication tools,
and incident reporting tools. The platform will follow cybersecurity best
practices of the Commission
eIDAS This will be specified by the Commission at a later stage.
Single Digital Gateway and IMI Not applicable
Others (e.g., Interoperable Europe Act) The need for another interoperability assessment under the Interoperable Europe
Act (Regulation EU 2024/903) will be evaluated once the operational details for
the EuroCloud platform become available.
Common procurement platform
Digital and/or sectorial policy (when these are
applicable)
Explanation on how it aligns
EN 51 EN
AI Act Not applicable
EU Cybersecurity framework The platform will follow cybersecurity best practices of the Commission
eIDAS This will be specified by the Commission at a later stage.
Single Digital Gateway and IMI Not applicable
Others (e.g., Interoperable Europe Act) Not applicable
Catalogue or repository of reusable software that is connected to the EU OSS Catalogue
Digital and/or sectorial policy (when these are
applicable)
Explanation on how it aligns
AI Act Not applicable
EU Cybersecurity framework The platform will follow cybersecurity best practices of the Commission
eIDAS The repository will re-use, in so far as relevant, the eIDAS framework
Single Digital Gateway and IMI Not applicable
Others (e.g., Interoperable Europe Act) Not applicable
EU Open Source Solutions Catalogue
Digital and/or sectorial policy (when these are
applicable)
Explanation on how it aligns
EN 52 EN
AI Act Not applicable
EU Cybersecurity framework The platform will follow cybersecurity best practices of the Commission
eIDAS The catalogue will re-use, in so far as relevant, the eIDAS framework
Single Digital Gateway and IMI Not applicable
Others (e.g., Interoperable Europe Act) Article 4 of the Interoperable Europe Act (Regulation EU 2024/903) mandates
the share and reuse of interoperability solutions between Union entities and
public sector bodies. The EU Open Source Solutions Catalogue delivers upon
this requirement.
4.4.Interoperability assessment
Digital public
service or
category of digital
public services
Description References(s) to the
requirement(s)
Interoperable
Europe
Solutions(s)
Other
interoperability
solution(s)
Union repository
of recognised
sovereign services
A dedicated Union repository of cloud
computing services that have received
recognition. To be established and
maintained by the Commission. National
competent authorities shall register the
relevant services in this repository.
TITLE IV AUTONOMY
Article 22 Central repository
of cloud computing services
// NA
EuroCloud
Federation
Article 34 establishes the European public
sector cloud federation (EuroCloud
Federation). The EuroCloud Federation
should bring together national cloud
TITLE IV AUTONOMY
Article 34 Establishment of
NA NA
EN 53 EN
initiatives providing highly trusted and
secure public sector cloud capabilities and
facilitate the sharing of such capabilities
between Union entities and public sector
bodies. This should be done via a platform
accessible to all Federation Members – the
EuroCloud Platform.
the European public sector
cloud federation
Impact of the requirement(s) as per digital public service on cross-border interoperability
Union repository of recognised sovereign services
Assessment Measure(s) Potential remaining barriers (if
applicable)
Alignment with existing digital and
sectorial policies
Please list the applicable digital and
sectorial policies identified
The implementation of the Union repository of
recognised sovereign services will take utmost account
of existing policies and the building blocks stemming
from them.
Organisational measures for a smooth
cross-border digital public services
delivery
Please list the governance measures
foreseen
The European Commission is tasked with establishing
and maintaining the repository.
The verifying national authority is responsible for
uploading the relevant data into the repository.
The organisational measures will need to be
detailed by the Commission, at a later stage.
Measures taken to ensure a shared
understanding of the data
Please list such measures
Semantic measures will be specified by the
Commission at a later stage.
EN 54 EN
Use of commonly agreed open
technical specifications and standards
Please list such measures
Technical measures will be specified by the
Commission at a later stage.
EuroCloud Federation
Assessment Measure(s) Potential remaining barriers (if
applicable)
Alignment with existing digital and
sectorial policies
Please list the applicable digital
and sectorial policies identified
The implementation of EuroCloud federation will take utmost
account of existing policies and the building blocks stemming
from them. In particular: NIS2 for cybersecurity, Simpl, Data
Spaces.
NA
Organisational measures for a
smooth cross-border digital public
services delivery
Please list the governance
measures foreseen
The detailed governance of the Eurocloud platform will be dealt
with through secondary legislation.
NA
Measures taken to ensure a shared
understanding of the data
Please list such measures
Semantic measures will be specified by the Commission at a later
stage.
NA
Use of commonly agreed open
technical specifications and
standards
Technical measures will be specified by the Commission at a
later stage.
NA
EN 55 EN
Please list such measures
4.5. Measures to support digital implementation
Description of the measure References(s) to the
requirement(s)
Commission
role Actors to be involved Expected timeline
The Commission is empowered to
adopt implementing acts
specifying (i) the technical,
operational and organisational
measures and (ii) the procedure to
participate in the EuroCloud
Federation as referred to in Article
40(2).
Article 40, Article 41 Drafting one or
more
implementing
acts
European Commission
Participating Member States
TBD
EN EN
EUROPEAN COMMISSION
Brussels, 3.6.2026
COM(2026) 502 final
ANNEXES 1 to 3
ANNEXES
to the
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strenghtening Europe's cloud and AI
ecosystem (Cloud and AI Development Act)
{SEC(2026) 502 final} - {SWD(2026) 502 final} - {SWD(2026) 503 final}
EN 1 EN
ANNEX I
GRAND CHALLENGES
1. Grand Challenge 1: Environmental sustainability, performance and security of the
Union’s data centres
Testing and deploying technologies for data centres across the Union to surpass state-of-the-
art energy-efficiency and resource efficiency.
This includes achieving lower Power Usage Effectiveness (PUE) and enabling significantly
higher server utilisation rates. Examples include:
(1) Lowering average Power Usage Effectiveness: improving the environmental
sustainability and performance of the Union's cloud and edge data centres to an
average Power Usage Effectiveness (PUE) of 1.15 across the Union. The main focal
areas include enabling the development of:
(a) advanced data centre energy efficiency technologies such as cooling, waste
heat recovery;
(b) quantum computing technologies for cloud and compute infrastructure
operations;
(c) grid integration and advanced energy management systems;
(d) pilot lines for the validation of next-generation energy-efficient technologies at
operational scale.
(2) Raising average server utilisation rates of data centres: raising average server
utilisation rates across the Union’s data centres towards 50%, by integrating for
example, AI-powered technologies for dynamic server utilisation management,
runtime workload management and scheduling or for balancing utilisation, energy
cost, thermal constraints, and latency requirements.
(3) Enhancing the security and resilience of data centres: enhancing the security and
resilience of data centres’ value chain and supply by integrating semiconductor
technologies and quantum technologies designed and manufactured in the Union, and
by improving their resistance to physical and cybersecurity threats, including
targeted attacks.
2. Grand Challenge 2: Cloud stacks
Building end-to-end hardware and software cloud stacks, including AI tools, infrastructure,
services and management layers to bridge the Union's critical capacity gaps.
This includes building AI servers powered by semiconductors and quantum technologies
designed and manufactured in the Union for distributed and decentralised cloud and edge
computing for AI.
Pilot programmes could help demonstrate the capabilities of the European open cloud stacks
in strategically important sectors.
3. Grand Challenge 3: Frontier AI
Developing the next generation of multimodal frontier AI models and systems and pioneering
novel capabilities.
The focus will be on the architectural design and development of next-generation multimodal
models and systems that push the boundaries of current algorithmic capabilities for achieving
superior performance in advanced reasoning, cross-modal understanding and agentic
EN 2 EN
capabilities; investigating novel approaches to model efficiency, cognitive modelling, and
alternative computational structures, etc.
The potential applications could include foundational science such as scientific discovery and
complex data interpretation, and the development of world models for improved
reasoning, automated management simulation and planning.
4. Grand Challenge 4: Physical AI
Developing advanced physical AI models and systems that operate autonomously and safely
for delivering robust, manipulation and navigation in unstructured environments.
The focus will be on co-designing software and its underlying hardware architectures and on
combining frontier AI techniques with world models supporting physical reasoning for
delivering robust manipulation, navigation, and interaction capabilities with minimal human
supervision.
The potential applications could include autonomous robots, industrial systems and
drones operating in dynamic real-world environments.
5. Grand Challenge 5: Industrial AI
Accelerate the development and deployment of European industrial AI across the Union’s
strategic sectors.
The focus will be on developing European industrial AI models and systems capable of
serving high-value industrial applications. Such models and systems should be adaptable to
sector-specific use cases and enable secure deployment.
The initiatives launched under this grand challenge should rely on specialised computing
resources and testing facilities necessary to validate AI systems in real-world environments
before supporting their large-scale deployment and uptake, including at regional and local
level.
In the automotive sector, those initiatives may facilitate the development and deployment of
innovative software platforms and AI models for automated driving, while in manufacturing,
they may enable the creation of specialised models that optimise production processes. Other
strategic sectors that could benefit from industrial AI may include healthcare, energy, agri-
food and defence.
6. Grand Challenge 6: Cooperative European Industrial Models
Developing cooperative European industrial AI models and systems for strategic sectors by
enabling collaboration at European industrial scale without exposing commercially sensitive
data between participants.
The focus will be on advanced confidentiality-preserving technologies. Those mechanisms
include federated and distributed training approaches where algorithms are brought to the data
rather than data being transferred centrally; secure execution environments, encryption-based
processing, anonymisation and pseudonymisation techniques, access compartmentalisation,
and protections against the extraction of commercially sensitive information from trained
models.
Strategic sectors that could benefit from cooperative European industrial AI models and
systems may include aerospace, pharmaceutics, cybersecurity, mobility, autonomous vehicles
and drones, energy and defence.
EN 3 EN
7. Grand Challenge 7: AI Agents Platform
Developing a European AI agent orchestration framework, providing the essential
middleware for the resilient and secure deployment of autonomous agents at scale.
The focus will be on (i) exploring innovative technological paradigms that enable multiple AI
agents to collaborate effectively, surpassing the capabilities of standalone systems while
maintaining rigorous security standards; and (ii) on the creation of resilient, cloud-based open
platforms dedicated to the large-scale management of AI agents.
The potential applications could include healthcare (such as clinical decision support and
research coordination), cybersecurity (such as threat detection and response), as well as
foundational science.
8. Grand Challenge 8: Public Sector AI
Developing AI models and systems, based on high-quality data from the public sector
targeting critical domains (such as healthcare, public administration, law and crisis
management as well as public services)
The focus will be on public service solutions that are expected to have a high positive impact
on the most critical public services and are shared across different levels of public sector
organisations.
One target will be to enable data sharing and frontier model development across national
public services to increase the impact on the overall Union’s public sector, including also in
areas handling sensitive data. Privacy-preserving frameworks, (such as federated learning and
high-fidelity synthetic data generation), that make it possible to train of models without
compromising the confidentiality of underlying datasets, and measures to accelerate the
broad uptake of those modems, including at regional and local level, will also help achieve
this target.
EN 4 EN
ANNEX II
CRITERIA FOR UNION ASSURANCE LEVELS
This Annex sets out the criteria to be met by cloud computing service providers and their
cloud computing services in order to be recognised as offering services at Union assurance
levels 1, 2, 3 and 4. For the purpose of the criteria under Union assurance levels 1, 2, 3, and 4,
‘software’ within the meaning of Regulation (EU) 2024/2847, Article 3, point (4) falls within
the scope of this Annex and Annex III to this Regulation. ‘Hardware’ within the meaning of
Regulation (EU) 2024/2847, Article 3, point (5) is outside of the scope.
1. Union assurance level 1
1.1. For Union assurance level 1, cloud computing service providers must meet the
following cumulative criteria:
(a) the cloud computing service provider is established in the Union;
(b) the infrastructure and assets of the cloud computing service provider, including those
of its subcontractors which are involved in the provision of the service, are located in
the Union unless the public sector body explicitly requires otherwise;
(c) the customer data, including metadata and telemetry data, that is processed, stored
and transferred by the cloud computing service provider, and by the subcontractors,
which are involved in the provision of the service, remain exclusively within the
Union, unless the public sector body explicitly requires otherwise and at any time,
including before, during or after the configuration or use of the service;
(d) where the cloud computing service provider outsources the technical and operational
support or assistance, including any subsequent sub-outsourcing arrangements, to
third-party service providers outside of the Union, the necessary legal, technical and
organisational measures are implemented to ensure traceability, security and
governance of those operations and those operations do not, in any way, compromise
the operational autonomy of the cloud computing service provider;
(e) the cloud computing service provider demonstrates that the service complies with the
state-of-the-art cybersecurity standards;
(f) the cloud computing service provider provides full transparency around the use of
subcontractors. The cloud computing service provider subjects subcontractors to due
diligence, contractual obligations and ongoing oversight to meet Union legal
obligations;
(g) Where the cloud computing service provider is subject to the control of a third
country or a legal entity established in a third-country, the cloud computing service
provider guarantees that there are no existing laws and practices in that third country,
demonstrated by independent sources, that require the cloud computing service
provider to report information on software vulnerabilities to authorities of that third
country prior to those vulnerabilities being known to have been exploited.
1.2. For Union assurance level 1, the subcontractors referred to in the first paragraph
must be subcontractors that are third parties that have a direct contractual
relationship with the cloud computing service provider and that contribute to the
provision and the delivery of the cloud computing service.
2. Union assurance level 2
EN 5 EN
2.1. For Union assurance level 2, cloud computing service providers must meet the
following cumulative criteria:
(a) the audited provider and the subcontractors which are involved in the provision of
the audited service are established in the Union;
(b) the infrastructure, assets, and personnel of the audited provider, including those of its
subcontractors which are involved in the provision of the service are located in the
Union;
(c) the customer data, including metadata and telemetry data, that is processed, stored
and transferred by the audited provider and the subcontractors which are involved in
the provision of the service, remain exclusively within the Union, unless the public
sector body explicitly requires otherwise and at any time, including before, during or
after the configuration or use of the service;
(d) if the public sector body determines that imposing additional personnel screening and
Union citizenship requirements are necessary, the audited provider should ensure that
presonnel meeting those requirements are available;
(e) the audited service obtains a European cybersecurity certificate of at least assurance
level ‘substantial’ under a European cybersecurity certification scheme covering
cloud computing services to be established under Regulation (EU) 2019/881,
provided that such a scheme has been established under that Regulation and is
available to cloud computing service providers. Until the establishment of such a
scheme, national cybersecurity certification schemes shall apply, where they exist.
Where no Union or national cybersecurity certification schemes exist, the audited
provider is to demonstrate that the service complies with the highest cybersecurity
standards under applicable Union law;
(f) the data generated by using the audited service are not used to train or fine-tune any
AI system operated by a third country or a legal entity established in a third-country ,
and are not transferred outside the Union in any case;
(g) if the audited provider and the subcontractors which are involved in the provision of
the audited service are subject to the control of a third country or a legal entity
established in a third-country, they demonstrate that the necessary legal, technical
and organisational measures have been implemented to ensure that the:
i. control of the third country or the legal entity established in a third-country
over the audited provider is not exercised in a manner that restrains or restricts
the provider’s ability to perform and deliver the service, imposes limitations on
the infrastructure, assets, and personnel required for the service provision, or
undermines the capabilities and standards necessary to perform the audited
service;
ii. access by a third country or by a legal entity established in a third-country to
customer data is prevented;
iii. possibility of disruption of the service continuity and/or the degradation of the
service quality by a third country or a legal entity established in a third country
is prevented;
iv. control of the third country or the legal entity established in a third-country
over the audited provider is not exercised in a manner that obliges the audited
provider to implement, enforce, give effect to, or comply with restrictive
measures such as sanction regimes, embargoes, or any equivalent legal or
EN 6 EN
administrative measures adopted by a third country, unless such measures are
legitimate under the national laws of Member States or Union law
(h) the technical and operational support or assistance related to the audited service,
including subsequent sub-outsourcing arrangements, are initiated and performed
exclusively within the Union;
(i) the audited provider demonstrates that the following software supply chain measures
are in place:
i. a complete and up-to-date software bill of materials (SBOM), as defined in
Article 3, point (39), of Regulation (EU) 2024/2847,and a list of identified
dependencies relevant to the provision of the service are documented and made
available to the auditing organisation;
ii. where software components as defined in Regulation (EU) 2024/2847 Article
3, point 6 or products are provided, owned, and licensed by a legal entity
established in a third country, controls are implemented and documented to
block any remote features that could materially tamper with or disrupt a device,
system, or software (including during updates) and to ensure that the security-
relevant components from third-country software manufacturers, as defined in
Regulation (EU) 2024/2847 Article 3, point 13, are subject to source code
audits, and have a documented migration plan in the event that the vendor fails
or a third country imposes restrictions;
iii. where the cloud computing service provider is subject to the control of a third
country or a legal entity established in a third-country, the cloud computing
service provider guarantees that there are no existing laws and practicesin that
third country, demonstrated by independent sources, that require the cloud
computing service provider to report information on software vulnerabilities to
authorities of that third country prior to those vulnerabilities being known to
have been exploited;
(j) where software released under an open-source licence is used for the provision of the
service, the audited provider demonstrates that it has implemented and documented
the appropriate controls to prevent the use of any remote features or mechanisms that
could be used to materially tamper with or disrupt a device, system, or software;
(k) to the extent that the audited provider provides its services globally and maintains a
subsidiary in a third country, the audited provider has implemented the necessary
measures to ensure and enforce the effective legal, technical and organisational
separation between the Union parent company and any such third-country subsidiary.
2.2. For Union assurance level 2, the subcontractors referred to in the first paragraph
must be subcontractors that are third parties that have a direct contractual
relationship to the cloud computing service provider and that contribute to the
provision and delivery of the cloud computing service.
3. Union assurance level 3
3.1. For Union assurance level 3, cloud computing service providers must meet the
following cumulative criteria:
(a) the audited provider and the subcontractors which are involved in the provision of
the audited service are established in the Union;
EN 7 EN
(b) the infrastructure, assets, and personnel of the audited provider, including those of
the subcontractors which are involved in the provision of the service, are located in
the Union;
(c) the customer data, including metadata and telemetry data, that is processed, stored
and transferred by the audited provider and the subcontractors which are involved in
the provision of the service, remain exclusively within the Union unless the public
sector body explicitly requires otherwise and at any time, including before, during or
after the configuration or use of the service;
(d) the personnel, including the personnel of the subcontractors which are involved in
the provision of the audited service are Union citizens and where appropriate, the
personnel must also have the necessary national security clearance issued by a
Member State when handling classified information, as defined in Article 2, point
(21), of Regulation (EU) 2021/697;
(e) the audited service obtains a European cybersecurity certificate of at least assurance
level ‘substantial’ under a European cybersecurity certification scheme covering
cloud computing services to be established under Regulation (EU) 2019/881,
provided that such a scheme has been established under that Regulation and is
available to cloud computing service providers. Until the establishment of such a
scheme, national cybersecurity certification schemes shall apply, where they exisit.
Where no Union or national cybersecurity certification schemes exist, the audited
provider is to demonstrate that the service complies with the highest cybersecurity
standards under applicable Union law;
(f) the data generated by using the audited service are not used to train or fine-tune any
AI system operated by a third country or a legal entity established in a third-country
and are not transferred outside the Union in any case;
(g) the audited provider and the subcontractors which are involved in the provision of
the audited service are not subject to the control of a third country or a legal entity
established in a third-country. By way of derogation to this criterion, a cloud
computing service provider and its subcontractors which are involved in the
provision of the audited service that are subject to the control of a third country or a
legal entity established in a third-country may be audited for Union assurance level
3 where the Commission has adopted an implementing act under Article 19. Where
the Commission has adopted an implementing act under Article 19, the audited
provider and the subcontractors which are involved in the provision of the audited
service must also demonstrate that the necessary legal, technical and organisational
measures have been implemented to ensure that the:
i. control of the third country or the legal entity established in a third-country
over the audited provider is not exercised in a manner that restrains or restricts
the provider’s ability to perform and deliver the service, imposes limitations on
the infrastructure, assets, and personnel required for the service provision, or
undermines the capabilities and standards necessary to perform the audited
service. The audited provider should allow for reasonable access to the code;
ii. access by a third country or by a legal entity established in a third-country to
customer data is prevented;
iii. possibility of disruption of the service continuity and/or the degradation of the
service quality by a third country or a legal entity established in a third country
is prevented;
EN 8 EN
iv. control of the third country or the legal entity established in a third-country
over the audited provider is not exercised in a manner that obliges the audited
provider to implement, enforce, give effect to, or comply with restrictive
measures such as sanction regimes, embargoes, or any equivalent legal or
administrative measures adopted by a third country, unless such measures are
legitimate under the national laws of Member States or Union law;
(h) the technical and operational support or assistance related to the audited service,
including subsequent sub-outsourcing arrangements, are initiated and performed
exclusively within the Union, by personnel that are Union residents, and by third
parties that are not subject to the control of a third country or a legal entity
established in a third country;
(i) the audited provider demonstrates that the following software supply chain measures
are in place:
i. a complete and up-to-date SBOM and a list of identified dependencies relevant
to the provision of the service are documented and made available to the
auditing organisation;
ii. where software components or products are provided, owned, and licensed by a
legal entity established in a third country, controls are implemented and
documented to block any remote features that could materially tamper with or
disrupt a device, system, or software (including during updates) and to ensure
that the security-relevant components from third-country manufacturers are
subject to source code audits, and have a documented migration plan in the
event that the vendor fails or a third country imposes restrictions;
iii. where the cloud computing service provider is subject to the control of a third
country or a legal entity established in a third-country, the cloud computing
service provider guarantees that there are no existing laws and practices in that
third country, demonstrated by independent sources, that require the cloud
computing service provider to report information on software vulnerabilities to
authorities of that third country prior to those vulnerabilities being known to
have been exploited;
(j) where software released under an open-source licence is used for the provision of the
service, the audited provider demonstrates that it has implemented and documented
the appropriate controls to prevent the use of any remote features or mechanisms that
could be used to materially tamper with or disrupt a device, system, or software;
(k) to the extent that the audited provider provides its services outside of the Union and
maintains a subsidiary in a third country, the audited provider demonstrates that it
has implemented the necessary measures to ensure and enforce the effective legal,
technical and organisational separation between the Union parent company and any
such third-country subsidiary.
3.2. For Union assurance level 3, the subcontractors referred to in the first paragraph
must be subcontractors that are third parties that have a direct contractual
relationship to the cloud computing service provider and that contribute to the
provision and the delivery of the cloud computing service, and that may require
access to classified or sensitive information, as defined in Article 2, point (22), of
Regulation (EU) 2021/697.
4. Union assurance level 4
EN 9 EN
4.1. For Union assurance level 4, cloud computing service providers must meet the
following cumulative criteria:
(a) the audited provider and the subcontractors which are involved in the provision of
the audited service are established in the Union;
(b) the infrastructure, assets, and personnel of the audited provider, including the
subcontractors , which are involved in the provision of the service, are located in the
Union;
(c) the customer data, including metadata and telemetry data, which, following a risk
assessment, is identified as sensitive, that is processed, stored and transferred by the
audited provider and the subcontractors which are involved in the provision of the
service, remain exclusively within the Union and at any time, including before,
during or after the configuration or use of the service;
(d) the personnel, including the personnel of the subcontractors , which are involved in
the provision of the audited service are Union citizens and, where appropriate, the
personnel must also have the necessary national security clearance issued by a
Member State when handling classified information;
(e) the audited service obtains a European cybersecurity certificate of at least assurance
level ‘high’ under a European cybersecurity certification scheme covering cloud
computing services to be established under Regulation (EU) 2019/881, provided that
such a scheme has been established under that Regulation and is available to cloud
computing service providers. Until the establishment of such a scheme, national
cybersecurity certification schemes shall apply, where they exist. Where no Union or
national cybersecurity certification schemes exist, the audited provider is to
demonstrate that the service complies with the highest cybersecurity standards under
applicable Union law;
(f) the data generated by using the audited service are not used to train or fine-tune any
AI system operated by a third country or a legal entity established in a third-country,
and are not transferred outside the Union in any case;
(g) the audited provider and the subcontractors which are involved in the provision of
the audited service are not subject to the control of a third country or a legal entity
established in a third-country;
(h) the technical and operational support or assistance related to the audited service,
including subsequent sub-outsourcing arrangements, are initiated and performed
exclusively within the Union, by personnel that are Union residents, and by third
parties that are not subject to the control of a third country or a legal entity
established in a third country;
(i) the audited provider demonstrates that the following software supply chain measures
are in place:
i. a complete and up-to-date SBOM and a list of identified dependencies relevant
to the provision of the service are documented and made available to the
auditing organisation;
ii. measures in place to retain effective control over the software components or
products by demonstrating that a third country or a legal entity established in a
third country does not hold or exercise effective control over the design,
development, maintenance, and evolution of those components or products.
Effective control includes the ability to materially influence the technical
EN 10 EN
evolution, maintenance priorities, security remediation, and long-term
continuity of the component;
(j) where software released under an open-source licence is used, the audited provider
demonstrates that it has implemented and documented the appropriate controls to
prevent the use of any remote features or mechanisms that could be used to
materially tamper with or disrupt a device, system, or software;
(k) to the extent that the audited provider provides its services outside of the Union and
maintains a subsidiary in a third country, the audited provider demonstrates that it
has implemented the necessary measures to ensure and enforce the effective legal,
technical and organisational separation between the Union parent company and any
such third-country subsidiary.
4.2. For Union assurance level 4, the subcontractors referred to in the first paragraph
must be subcontractors that are third parties that have a direct contractual
relationship to the cloud computing service provider, that contribute to the provision
and delivery of the cloud computing service, and that may require access to classified
or sensitive information in order to carry out the service provision.
EN 11 EN
ANNEX III
AUDIT EVIDENCE FOR THE AUDIT PROCEDURE
Auditing organisations should request the audit evidence listed in this Annex from the audited
provider when assessing the compliance of the audited service against the applicable audit
criteria under Annex II.
This Annex is indicative and does not limit the evidence that may be requested or considered
by the auditing organisations. Auditors may seek any additional information necessary to
ensure a comprehensive and accurate assessment of compliance to conduct audits. While the
evidence requested may be the same, the aspects that need to be analysed will differ
depending on the assurance level criteria and their strictness.
1. Audit criterion A – Union establishment
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (a) of Annex II based on the following:
(1) Any evidence demonstrating that the audited provider is incorporated under the law
of a Member State in the Union or otherwise constituted in line with company law of
a Member State in the Union.
(2) Any evidence that the registered office, central administration, and main
establishment of the audited provider is established within the Union.
(3) The auditing organisation should verify the applicable Union legal framework of the
audited provider and verify whether their establishment is genuine and stable or
whether the audited provider instead qualifies only as a non-EU provider offering
services in the Union.
(4) The auditing organisation should verify whether the provider is legally incorporated
in a Member State of the Union. Evidence of this could include, but is not limited to,
the national company extracts, tax residency documentation, business licences, VAT
registration, verification of whether the provider is registered in the Business
Registers Interconnected System (BRIS) and the VAT information Exchange System
(VIES).
(5) The auditing organisation should verify the stable and effective presence in the
Union of the audited provider. The auditing organisation should therefore verify that:
(a) EU physical offices or operational premises exist (for example, through lease
contracts, utility bills or property documents);
(b) permanent staff is located in the Union and that customer support operations
are carried out in the Union (for example, through employment contracts,
payroll records, personnel timesheets);
(c) contractual operations are handled in the Union (for example, through activity
management records, incident reporting records);
(d) banking and accounting functions are exclusively in the Union (for example,
through financial statements and statutory audit reports).
(6) The auditing organisation should verify the presence of EU Member State
establishment units or branches (for example, through lease contracts, utility bills,
property documents, employment contracts, timesheets, payroll records or purchase
orders).
EN 12 EN
2. Audit criterion B – Location of infrastructure, assets, and personnel
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (b) of Annex II based on the following evidence:
(1) Location of infrastructure
(a) A list with relevant details of the location of the infrastructure and data storage
locations used in the provision of the audited service. This list should include
the precise location (number, street, city, postal code and country) of the
infrastructure demonstrating that all elements remain within the Union for the
provision of the audited service. This includes the location for the primary,
backup, disaster recovery and log storage.
(b) Any other evidence that the IT infrastructure is located in the Union such as
lease agreements, property deeds, maintenance contracts, service contracts,
facility access logs.
(c) Network diagrams and architecture documents illustrating the exclusive use of
Union-based infrastructure for data storage and processing, including backup
and replicated data.
(2) Location of assets
(a) A list and relevant details of the assets used in the provision of the audited
service, such as an asset register.
(b) Evidence that servers, equipment, and operational assets are located in the
Union, such as records identifying the server and its location, purchase
invoices, delivery notes, licence agreements, subscription contracts, or invoices
for software purchases or subscriptions, invoices with delivery proofs for
hardware.
(c) Evidence that service delivery capabilities are based in the Union, such as
deployment records, installation records, service status reports, configuration
reports, monitoring outputs, admin logs showing usage of the service.
(3) Location of personnel
(a) A list and relevant details of the personnel involved in the provision of the
audited service.
(b) Evidence that the personnel involved in the provision of the audited service are
located in the Union, including employment contracts, payroll records,
timesheets, activity records, and organisational charts showing Union-based
staff with operational responsibilities.
(4) Considerations regarding the infrastructure, assets and personnel
(a) The auditing organisation should also assess where such infrastructure, assets,
or personnel:
i. store, transmit, access, process or otherwise handle customer data;
ii. provide, enable, or could enable administrative access to, control over,
configuration of, or visibility into customer data;
iii. if compromised, misconfigured, made unavailable or disrupted could
reasonably result in the disruption or unavailability of the audited service.
EN 13 EN
NB: ‘Infrastructure’ means physical infrastructure, including but not limited to, data centre
infrastructure or colocation infrastructure, network, cooling, and IT systems that allow for
the management of the datacentre.
‘Assets’ means hardware and software, including, but not limited to, libraries, the internal
network needed for software components to communicate, cryptographic materials that
enables the provision of the cloud computing service.
‘Personnel’, including personnel managed by subcontractors, means individuals who support
the delivery, administration, security, availability, or operation of the audited service.
3. Audit criterion C – Data localisation in the Union
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (c) of Annex II based on the following evidence:
(1) Evidence demonstrating that customer data are stored and processed exclusively in
the Union and that third-parties or subcontractors that do not meet the conditions
under this Annex are in no circumstances technically or operationally able to access,
obtain, make unavailable, destroy or more generally process customer data without
prior authorisation. Examples include access logs, support access policies, privileged
access records, backup retention policy, data flows diagram demonstrating where the
customer data are stored, processed, replicated and backed up. When processing
personal data, contracts with the subcontractors that demonstrate compliance with
Regulation (EU) 2016/679.
(2) Evidence of logs and monitoring records demonstrating that all data are stored and
processed exclusively within the Union. Examples include master service
agreements, data processing agreements, data residency contractual agreements or
any EU data boundary.
(3) Evidence (such as contractual agreements, logs, and procedures offered to public
sector bodies ) demonstrating that the audited provider and the subcontractors which
are involved in the provision of the audited service have put in place the necessary
measures to ensure that:
(a) no customer data, including encrypted data, are transferred outside of the
Union without public sector body approval;
(b) no data are transferred to any third-party other than subcontractors which are
involved in the provision of the service or recipients expressly authorised by
the public sector body.
(4) A data flow diagram showing the flows of data between the cloud computing service
provider and customer data, as well as with third-party services and subcontractors.
The diagram must clearly identify the source and destination of data and demonstrate
that the data does not leave the Union.
NB: For the purpose of this Annex, a customer means a public sector body who has entered
into a contractual or other legally binding arrangement with the cloud computing service
provider for the purpose of accessing or using the cloud computing service.
‘Customer data’ could mean any data under the control of the cloud computing service
customer, whether by legal, contractual, or other means, that are:
EN 14 EN
(a) input into the cloud computing service by or on behalf of the customer,
including authentication credentials;
(b) produced through the customer’s use of the cloud computing service.
Customer data may also include data under the audited providers control that are derived as
a result of interaction with the audited service by the cloud service customer. This includes
customer data and any data resulting from the usage of the cloud computing services (i.e.
telemetry, metadata).
‘Cloud compunting service derived data’ includes the portion of log data containing records
of who used the service, at what times, which functions, types of data involved and so on. It
can also include information about the numbers of authorised users and their identities. It can
also include any configuration or customisation data, where the cloud computing service has
such configuration and customisation functionalities.
4. Audit criterion D – Union citizenship
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (d) of Annex II based on the following evidence:
(1) The audited provider should provide the auditing organisation with proof that it has
implemented the measures to ensure that, if a public sector body were to request
Union citizenship, the personnel involved in the provision of the audited service are
Union citizens. This can be demonstrated through valid official government issued
documents (e.g. valid passport and national identity card);
(2) The audited provider should provide organisational charts and job descriptions
confirming that it can ensure, where requested, that only personnel with Union
citizenship have access to the audited service's operation, management, maintenance,
and support.
(3) The audited provider should provide documents demonstrating access control
policies and audit trails showing that only authorised personnel who are Union
citizens can access the service's systems and data.
(4) The audited provider should demonstrate that it has put in place procedures
describing how citizenship is verified before assignment and how compliance with
this audit criterion is maintained throughout employment.
NB: Personnel involved in the provision of the audited service could include personnel who
have logical or physical access to infrastructure and assets used to operate the cloud
computing service, as well as those who are responsible for customer support, and all
personnel who have management control of the cloud compunting service provider.
5. Audit criteria E – European cybersecurity certification scheme adopted under
Regulation 2019/881
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (e) of Annex II based on the following evidence:
(1) A valid European cybersecurity certificate issued by a competent conformity
assessment body demonstrating that the audited service has been assessed and found
compliant with the requirements corresponding to the ‘basic’, ‘substantial’ or ‘high’
assurance levels under a European cybersecurity certification scheme adopted under
Regulation (EU) 2019/881, provided that such shas been established ;
EN 15 EN
(2) A certification report including a description of the main components used for the
development and operation of the cloud computing service that is covered by the
audit certificate.
(3) Until the European cybersecurity certification scheme covering cloud computing
services has been established, the audited provider can demonstrate compliance
through valid cybersecurity certifications. This can include, but is not limited to, the
following:
(a) a valid certificate issued by a competent conformity assessment body (in line
with CEN/CLC/TS 18072:2025) demonstrating that the cloud computing
service has been assessed and found compliant with the requirements
corresponding to the ‘basic’ or ‘substantial’ or ‘high’ assurance levels defined
under CEN/TS 18026:2024;
(b) A valid certificate issued by the relevant national competent authority
demonstrating that the cloud computing service has been assessed and found
compliant with the requirements under the national cybersecurity scheme
currently in place in the Member State;
(c) in the absence of a national scheme, evidence demonstrating adherence to the
highest level of cybersecurity standards available on the market.
6. Audit criterion F – AI systems operated by a third country or third country
legal entity
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (f) of Annex II based on the following evidence:
(1) Contractual clauses stating that data processed or generated by using the audited
service, including customer-derived data, logs and telemetry, will not be used to train
or fine-tune any AI model or system operated by a third country or a third-country
legal entity, and are not transferred outside the Union in any case.
(2) Contractual clauses specifying that data are processed solely for the delivery of the
audited service and not for service improvements or model or system enhancements
or any other secondary purpose.
(3) Data flow diagrams documenting the end-to-end flow of data, covering data
ingestion, storage, processing and deletion. The diagrams should also show where
the AI pipelines or machine learning operations (MLOps) connect with customer
data.
(4) MLOps or deployment records demonstrating that the build, test and release
locations are in the EU.
(5) Model or system cards covering the model or system name, version, training and
validation sources, including statements that the data generated by using the audited
service does not leave the Union.
(6) Data lineage polices and related implementation documentation that shows that the
provider operates data lineage and provenance tools and that can demonstrate (per
record) what the data has been used for.
(7) A list of the subcontractors (indicating their country of establishment) that access the
data generated by using the audited service.
EN 16 EN
7. Audit criterion G – Absence of third-country control or third-country entity
control
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (g) of Annex II based on the following evidence:
(1) The auditing organisation should identify and analyse:
(a) all direct and indirect shareholders, up to the ultimate owners;
(b) the cap table documenting the company’s ownership structure;
(c) the body or bodies empowered to take strategic decisions (general assembly of
shareholders, supervisory board, board of directors, etc.);
(d) the rules for the appointment/election/removal of governing bodies and the
actual composition of the governing bodies (e.g. to identify if any shareholder
is entitled to nominate a board representative or has majority seats in the
board);
(e) the quorums and majority required for adopting strategic decisions, in order to
determine if any shareholder can take a strategic decision (either because they
have the required majority to approve such a decision or because they can
block such a decision through a veto or other specific rights even if they cannot
impose such a decision on their own, etc.);
(f) the possible influence on strategic decisions through commercial links,
financial links or other means, etc.
(2) The audited provider should request all the above information from its subcontractors
and make it available to the auditing organisation.
7.1. Assessment of ownership and control:
(1) The audited provider should provide the auditing organisation with the following
evidence related to the headquarters:
(a) the location and full address of the global headquarters and/or head office;
(b) the locations of the executive management structures.
(2) The audited provider should provide the auditing organisation with the following
evidence related to the ownership structure and specific rights:
(a) A detailed list describing any owners that:
i. hold, directly or indirectly, at least 5% of the capital or at least 5% of
the voting rights, including through any content, understanding,
relationship or intermediary. This includes voting agreements between
shareholders that would together have more than 5% of the voting
rights or 5% of the capital;
ii. have one or more of the following specific rights in relation to their
ownership: (a) right to veto a transfer of shares; (b) pre-emption rights;
(c) right to purchase additional shares or investment subject to
conditions.
(b) The auditing organisation should request the following supporting documents
to assess the elements in paragraph 2(a):
EN 17 EN
i. commercial registry extracts and shareholders’ books of the
organisation and any other relevant document that clearly indicate the
shareholders and their voting rights or percentage of interest;
ii. shareholders’ agreement, memorandum of understanding among
shareholders, statutes, articles of association or other relevant
documents regarding the decision-making procedures within the legal
entity, investment agreements between the shareholders, etc.;
iii. for any shareholders that are legal persons that hold at least 5% in the
capital or at least 5% of the voting rights:
(1) a graph describing the different ownership layers/chain of control
until the ultimate owners;
(2) the articles of association, bylaws or equivalent constitutional
documents;
(3) a register of directors, officers and signatories.
(3) The audited provider should provide the auditing organisation with the following
evidence related to the corporate governance:
(a) The audited provider should provide the auditing organisation with a
description of:
i. the decision-making bodies, their composition as well as their
nationality or place of establishment (where applicable);
ii. the rules regarding election, appointment, nomination or tenure of
members of the decision-making bodies or other management
positions;
iii. the decision-making procedures, including information on the required
majority and/or quorum needed for decisions;
iv. internal governance policies describing how ownership and control
decisions are recorded and approved;
board and management decisions reflecting the stated control structure;
board minutes and resolutions for control changes.
(b) The audited provider should provide the auditing organisation with supporting
documents setting out or describing: the decision-making bodies and the rules
on their election, appointment, nomination or tenure, decision-making
procedures, voting rights, veto rights, appointment rights, approval rights
within the legal entity (e.g. articles of association bylaws, reports on corporate
governance, etc.). The supporting documents and information should be
provided for each intermediate legal entity that directly or indirectly holds 5%
or more of the capital or voting rights, up to the ultimate owners of all the
layers involved.
(4) The audited provider should provide the following control-related evidence to the
auditing organisation:
(a) The audited provider should provide the auditing organisation with the
evidence of the commercial links conferring control. This includes, but is not
limited to, a list of individuals or legal entities with whom the audited
providers (or the owners of the audited provider, including intermediate layers
EN 18 EN
until the ultimate owners) have a commercial relationship that (a) leads to a
similar level of control on management and resources as the ownership of
shares or assets; and (b) is of very long duration (e.g. very important long-term
supply agreements or credits provided by software manufacturers/customers,
coupled with structural links).
(b) The supporting documents should include cooperation agreements with the
public sector body or software manufacturers , etc.
(5) The audited provider should provide the auditing organisation with the following
evidence related to the financial links conferring control:
(a) The audited provider should list the individuals or legal entities (including
controlling shareholders or owners) on whom the audited provider (or the
owners) are financially dependent in a way that could allow them to obtain
concessions in strategic business areas.
(b) The supporting documents should include loan documents, by-laws, documents
showing the financial link; etc.
(6) The audited provider should provide the auditing organisation with the following
evidence related to other sources of control:
(a) The audited providers should indicate to the audited organisation if there is any
other means, process or link ultimately conferring control to another third
country or a legal entity established in a third country (similar level of control
on management and resources as the ownership of shares or assets and of long
duration).
(b) Supporting documents should provide evidence of any such control or a
declaration that there is no such control (this declaration may come from the
management board of the service provider).
NB: The elements that should be taken into account when assessing control are the
ownership structures and specific rights, corporate governance, commercial links
conferring control, financial links conferring control and any other sources of
control.
7.2. Additional steps based on the conclusion of the ownership and control test
If the auditing organisation determines that the audited provider is subject to the control of a
third country or a third-country legal entity, it should request the following additional
evidence:
(c) Demonstrating that the Commission has adopted a decision pursuant to Article
19 regarding the third country for which the cloud computing service provider
is subject to the control of;
(d) All evidence demonstrating that the audited provider and any
subcontractorinvolved in the provision of the audited service has implemented
the necessary measures to enforce the effective legal, technical and
organisational separation between the cloud computing service provider and
any third country or legal entity established in a third country, ensuring that the
cloud computing service provider is unable to comply, legally, technically and
operationally, with any request to access customer data, including encrypted
data, or to disrupt service continuity or to degrade service quality.
EN 19 EN
(e) All evidence demonstrating that the public sector body will be informed of any
such request and a confirmation that the request has been refused;
(f) All evidence demonstrating the maintenance of an up-to-date record of any
request to access customer data, to disrupt service continuity or to degrade
service quality from a third country or a legal entity established in a third
country, containing at least the request and the response to the request.
8. Audit criterion H – No technical and operational support outside of the Union
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (h) of Annex II based on the following evidence:
(1) evidence that the audited provider has implemented binding contractual clauses
stating that all support, administration, maintenance, monitoring, incident response,
and operational activities must be initiated and performed exclusively in the Union.
This could include contractual clauses requiring advanced disclosure of all
subcontractors and support locations, prior written approval before engaging any new
subcontractors, and a right to reject any subcontractors located outside of the Union;
(2) evidence that the audited provider maintains an up-to-date subcontractor register;
(3) evidence that the audited provider does not subcontract or transfer such activities
outside of the Union;
(4) evidence that the audited provider has implemented the necessary legal, technical
and organisational measures to ensure that there can be no remote access for
technical and operational support from outside the Union for the audited service.
(5) evidence that the audited provider’s help desk/support services, infrastructure
administration, operations of its security operations centre (SOC) or network
operations centre (NOC), privileged access, backup handling, and disaster recovery
operations of the audited service are exclusively provided from the Union, including
the access path to operate the service;
(6) evidence that the audited provider ensures that the personnel upon the departure from
the company have no further access to the audited service and revokes all access
policies;
(7) evidence that the audited provider has implemented the necessary technical and
organisational measures to ensure that administrative access to systems used to
operate the audited service is provided through access paths located within the
Union. This can be demonstrated through the implementation of geographically
restricted network controls, Union-based administrative infrastructure, privileged
access management controls, and monitoring mechanisms;
(8) evidence that the audited provider has procedures in place that there is no effective
control of a third country or a legal entity established in a third country, including for
subsequent sub-outsourcing.
9. Audit criterion I - Ensuring the transparency of the software supply chain
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (i) of Annex II based on the following evidence:
(1) The audited provider should make available to the auditing organisation a complete
and up-to-date software bill of materials (SBOM) for all software components,
including open-source software (OSS).
EN 20 EN
(2) The audited provider should make available to the auditing organisation a list of
dependencies. This should include:
(a) all software modules, libraries or application programming interfaces (APIs)
used, as well as development tools;
(b) origin of software: where (country of origin) and by whom the software is
designed, developed and maintained, the location and jurisdiction governing
the software distribution, and updates;
(c) degree of reliance on non-EU vendors, facilities, or proprietary technologies;
for level 3, evidence that in case the software stack is provided by a third
coutnry entity, no unduly unjustified licensing restrictions are in place.
(d) degree of reliance on open-source software;
(e) visibility into the entire software manufacturer and sub-manufacturer chain,
including audit rights.
N.B. The requirements above imply that joint ventures made, e.g., of a Union entity with a
legal entity established in a third country can qualify for this level
(3) The audited provider should provide:
(a) evidence of a risk-based process for identifying and mitigating dependencies
on external software manufacturers relevant to the operation of the cloud
computing service;
(b) evidence that it has identified one or more alternative software solutions,
including open-source software. If equivalent software cannot be identified, a
solution ensuring minimal viable functionality must be identified. Tests must
be implemented and a switchover plan enabling migration to such alternative
solutions;
(c) evidence that it can migrate to an alternative solution in the event of any defect
or failure of the vendor or restrictions from a third country or a legal entity
established in a third country;
(d) provide a list of open standards that are followed as part of the audited
providers policies regarding the audited service;
(4) The audited provider should ensure transparency through remote access and source
code auditability by:
(a) making available to the auditing organisation a list of evidence to prove that
there is no use of any remote features or mechanisms that could be used to
materially tamper with or disrupt a device, system, or software. This should
include:
i. evidence related to the testing of the software component to prevent the
use of any remote features or mechanisms that could be used to
materially tamper with or disrupt a device, system, or software (test
procedure, test reports, test plan, etc.);
ii. evidence that the organisation’s change management procedures cover
any change in firmware, bios and software updates as well as
integration of a new components to prevent the use of any remote
feature or mechanism;
EN 21 EN
iii. evidence that the maintenance procedure is updated to include
preventing any remote feature or mechanism that could be used to
materially tamper with or disrupt a device, system or software.
(b) The audited provider must ensure that the third-party independent auditor is
granted the right to access and audit the source code of such software. The
audited provider must also ensure that all documentation, technical material,
information necessary to evaluate and audit the source code are made available
to the auditing organisation in a complete, accurate, and accessible format.
10. Audit criterion J – Open-source software
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (j) of Annex II based on the following evidence:
(1) The audited provider should ensure transparency through remote access and source
code auditability by:
(a) making available to the auditing organisation a list of evidence to prove that
there is no use of any remote features or mechanisms that could be used to
materially tamper with or disrupt a device, system, or software. This should
include:
i. evidence related to the testing of the software component to prevent the
use of any remote features or mechanisms that could be used to
materially tamper with or disrupt a device, system, or software (test
procedure, test reports, test plan, etc.);
ii. evidence that the organisation’s change management procedures
include any change in firmware, bios and software updates as well as
integration of new components to prevent the use of any remote feature
or mechanism;
iii. evidence that the maintenance procedure is updated to include
preventing any remote feature or mechanism that could be used to
materially tamper with or disrupt a device, system or software;
(2) The audited provider should provide:
(a) evidence of a risk-based process to identify and mitigate: (i) a weak ecosystem
and community support of the OSS; (ii) a failure to continuously monitor the
updates released; (iii) cases where the OSS is deprecated or is no longer
maintained.
(b) evidence that the audited provider has applied the up-to-date OSS without
undue delay;
(c) evidence that the audited provider has identified one or several alternative
open-source solutions. If the audited provider cannot identify an equivalent
software, it must identify a solution ensuring minimal viable functionality. The
audited provider must implement tests must and a switchover plan enabling
migration to the alternative solutions.
(3) Where the audited provider uses software released under an open-source licence, the
audited provider should implement mechanisms to detect and provide timely notice
to the public sector body if the software is acquired by or comes under control of a
third country or a legal entity or foundation established in a third country.
EN 22 EN
11. Audit criterion K – Global services and subsidiaries in third-countries
The auditing organisation should assess the audit criterion listed under Union assurance levels
2, 3, and 4 paragraph (k) of Annex II based on the following evidence:
(1) The auditing organisation should verify that the subsidiary is legally and
operationally independent from the audited provider.
(2) The audited provider must demonstrate that the subsidiary has no access to systems
processing or storing the customer data.
(3) The audited provider must demonstrate that the subsidiary has no privileged accounts
within the Union production environments, including cloud administration, Identity
and Access Management (IAM), Privileged Access Management (PAM), monitoring
or database administration privileges.
(4) The auditing organisation should verify that the personnel of the subsidiary cannot
obtain access to Union customer data.
(5) The auditing organisation should verify that the subsidiary has no authority to
instruct Union operational staff to disclose customer data or bypass security
procedures.
(6) The auditing organisation should verify that all foreign government requests received
by the subsidiary are formally redirected to the competent Union entity for legal
assessment under Union and Member State law.
EUROPEAN COMMISSION
08.05.2026
SEC(2026) 502
REGULATORY SCRUTINY BOARD OPINION
{COM(2026) 502}
{SWD(2026) 502-503}
Impact assessment / Cloud and AI Development Act
________________________________ Commission européenne/Europese Commissie, 1049 Bruxelles/Brussel, BELGIQUE/BELGIË - Tel. +32 22991111 [email protected]
EUROPEAN COMMISSION REGULATORY SCRUTINY BOARD
Brussels, RSB
Opinion
Title: Impact assessment / Cloud and AI Development Act
Overall 3rd opinion: POSITIVE
(A) Policy context
The Cloud and AI Development Act was a recommendation of the Draghi Report and in
the Competitiveness Compass. The main objective of this initiative is to address the EU’s
gap in cloud and AI infrastructure around 3 pillars: research & innovation aimed at
accelerating research and innovation in cutting-edge AI computing technologies, the
sustainable cloud and AI capacity pillar to boost EU’s data processing infrastructure, and
increased secure data processing capacities of cloud providers located in the EU for the
high-critical use cases.
(B) Key issues
The Board notes the improvements to the report, in particular with regard to policy
options and assessment of impacts.
The Board gives a positive opinion. The Board considers that the report should be
further improved with respect to the following aspects:
(1) The report should sufficiently describe the policy measure establishing a set of
non-price award criteria and the one encouraging the use, reuse, development
and sharing of open-source assets by the public sector.
(2) The internal coherence between the policy measures addressing sovereign cloud
and AI services and those addressing critical dependencies and supporting the
use of Open Source should be sufficiently explained and assessed.
(3) The costs related to migration and porting and the related uncertainties should
be sufficiently reflected in the assessment of the total costs in the main report.
2
(C) What to improve
(1) The policy measures aimed at reducing critical dependencies (PM16/PM19) should
be better described, in particular why the criterion “outside of the country of
dependencies” is used rather than “within the EU”, on the basis of which criteria the
dependency threshold is set at 50%, how the non-price award criteria are to be used
by public buyers, how they will be used and assessed. The report should also better
discuss the representativeness of the assumptions in PM15 regarding what
percentages of use cases would respectively require sovereignty levels 1, 2, 3 and 4.
(2) Policy measure PM20 encourages the use of Open Source in the public sector and
requires the contracting authorities to assess the equivalence/superiority of Open
Source over proprietary solutions in the tendering procedure. PM22 encourages EU
level joint public procurement. Building on the annex, the report should better analyse
the impact of these measures on public authorities, including the need for specialised
expertise and interplay with envisaged Open-Source Programme Offices, as well as
how a uniform approach across the EU will be ensured. The impact of these measures
on the companies offering the relevant IT services should also be discussed.
(3) Regarding external coherence: as the policy measure PM11, defining the sovereignty
levels and the associated requirements, includes criteria based on the European
Cybersecurity Certification Scheme for Cloud Services (EUCS), which has not yet
been adopted, the report should describe in more detail the interplay of the two
initiatives as well as implementation uncertainties for EUCS.
(4) In view of the impact on the total costs of the initiative, the costs related to migration
and porting, described in box 2 and in the annex, should be reflected in a summary
table of costs and benefits.
(D) Conclusion
The lead Service may proceed with the initiative.
The lead Service should take these recommendations into account when revising the
report and its executive summary in accordance with the Board’s findings before
launching the interservice consultation.
Full title Cloud and AI Development Act
Reference number PLAN/2025/815
Submitted to RSB on 30 April 2026
Date of RSB meeting “Written procedure”
3
EUROPEAN COMMISSION REGULATORY SCRUTINY BOARD
Brussels, RSB
Opinion
Title: Impact assessment / Cloud and AI Development Act
Overall 2nd opinion: NEGATIVE
(A) Policy context
The Cloud and AI Development Act was a recommendation of the Draghi Report and in
the Competitiveness Compass.
The general objective of the intervention is to ensure the functioning of the internal market
for cloud and AI computing services and to secure the conditions necessary for the
Union’s competitiveness and strategic autonomy
(B) Key issues
The Board notes the information provided and the changes in the objectives and
measures, and improvements regarding presentation of some limitations of the
report.
However, the Board maintains its negative opinion because the revised report still
contains the following significant shortcomings:
(1) The key proposed policy measures are not sufficiently specified to allow for the
assessment of those measures and whether they can address the identified
problems of EU competitiveness and strategic autonomy. Their proportionality
is not adequately demonstrated.
(2) The analysis of effectiveness and efficiency does not adequately reflect all the
costs (i.e. transition costs). The benefits appear to be over-estimated.
(3) The analysis of coherence must be reinforced as there are apparent overlaps
with existing and upcoming legislation.
4
(C) What to improve
(1) Given the market situation and the stated specific objectives, the interplay between
the criteria for authorising sovereign cloud services and existing legislation in third
countries should be clarified and the resulting impacts analysed. The report must provide
a robust estimate of the number of entities affected by mandatory requirements regarding
sovereign solutions, as well as which sovereign cloud services EU service providers
currently do not offer. The report should explain how “countries with dependencies” will
be identified as well as how and by whom the criterion of equivalence/superiority will be
established for the prioritisation of open source. Without these clarifications the impact
and proportionality of the intervention in terms of costs, including transition costs, and
benefits cannot be credibly assessed.
(2) The analysis of benefits must be based on robust and verifiable assumptions and
methodologies. Although the report acknowledges the arbitrary choice of certain
assumptions and parameters used for assessing the impact of key measures under the
preferred option, those (for example increased market shares, risk parameters allocation
between the various options, mark-up percentage for sovereignty increased costs) appear
to drive the results of the analysis.
(3) The report must provide a more comprehensive analysis of technical, functional and
operational consequences of key measures (i.e. PM 20, 21). As certain measures (in
particular PM 21, embedding PM 11, 15 and 16) aim at building an EU sovereign cloud,
the impacts of the limitations of supply from third countries must be assessed, including
the time and costs estimated for EU service-providers to offer equivalent substitutes. The
report should assess the pass-through of these costs.
(4) The report should provide a detailed analysis of coherence, in particular of the
authorisation scheme for cloud and AI computing services with the security risk
assessment in the Cybersecurity Act and of public procurement measures with the
upcoming revision of the public procurement rules. This should be accompanied by an
analysis of the potential additional legal complexity and administrative burden.
(5) Given the above, the report does not adequately inform the decision-making.
(D) Conclusion
The Board’s opinion is, as a rule, final. The lead Service can seek political guidance
on whether, and under which conditions, this initiative may proceed further.
Full title Cloud and AI Development Act
Reference number PLAN/2025/815
Submitted to RSB on 25 February 2026
Date of RSB meeting “Written procedure”
5
EUROPEAN COMMISSION REGULATORY SCRUTINY BOARD
Brussels, RSB
Opinion
Title: Impact assessment / Cloud and AI Development Act
Overall opinion: NEGATIVE
(A) Policy context
The Cloud and AI Development Act was a recommendation of the Draghi Report and in the
Competitiveness Compass. The main objective of this initiative is to address the EU’s gap
in cloud and AI infrastructure around 3 pillars: research & innovation aimed at accelerating
research and innovation in cutting-edge AI computing technologies, the sustainable cloud
and AI capacity pillar to boost EU’s data processing infrastructure, and increased secure data
processing capacities of cloud providers located in the EU for the high-critical use cases.
(B) Key issues
The Board notes the additional information provided and commitments to make
changes to the report.
However, the Board gives a negative opinion because the report contains the following
serious shortcomings that the lead Service must address:
(1) The report does not sufficiently establish the root cause(s) of the competitiveness
problems of the EU industry and the regulatory and market failures affecting the
cloud ecosystem in the EU.
(2) The aspects of EU strategic autonomy are insufficiently reflected in the report.
(3) The proposed policy measures, including on EU ‘sovereign’ solutions, are not
clearly defined and the causal links in the intervention logic are not sufficiently
substantiated.
(4) The analysis of effectiveness and efficiency does not sufficiently reflect the
adjustment costs, technical feasibility and unintended consequences linked to the
proposed policy measures.
6
(C) What to improve
1) The report should better analyse the root causes of the insufficient cloud capacities in
the EU and the declining market share of European services providers. Based on an analysis
of the main drivers of competitiveness, such as price, scope of services, innovation and
quality, including cyber security, it should analyse potential market failures. It should also
analyse potential regulatory failures based on available evidence of investment decisions. It
should analyse more thoroughly the current practices and regulatory frameworks in Member
States and how they impact the Internal Market. It should describe the risks of reliance on
non-EU providers and the potential trade-offs with aspects of strategic autonomy.
2) The report should be clear whether – and if yes how - the objectives of this initiative
relate to the EU’s strategic autonomy. What level of autonomy is to be achieved and in which
domains? The content of policy measures should be clearly defined, including the criteria for
“sovereign solutions” and “EU-made”. The report should better justify how the target of
market share of EU providers is set.
3) The report should assess the potential impacts related to the criteria for “sovereign
solutions”. It should analyse in sufficient detail the levels of sovereignty to be reached and
be transparent about technical prerequisites needed. The report should assess all the
adjustment costs and quantify them to the extent possible and be clear on who will bear these
costs. The report should provide a more comprehensive analysis of unintended
consequences, including on cyber security, in particular related to sovereignty and open-
source requirements.
4) The analysis of benefits should be based on robust assumptions, in particular for PO2B
and PO2C where stated benefits are very high. For example, all implications should be
factored in the analysis, such as investments needed to maintain cyber security, and the scale
needed to make them viable in economic terms. In terms of the environmental footprint, the
report should explain the mechanisms leading small(er) EU cloud providers to significantly
reduced Power Usage Effectiveness values and how European-funded R&D initiatives are
expected to deliver superior results when compared to the efforts of global hyperscalers.
5) The report should allow for verifiability of the economic modelling. It should quantify
and monetise only benefits which can be clearly linked to the policy measures and whose
quantification can be substantiated by evidence. Currently the benefit cost ratios presented
in the report are not plausible. After revision of the analysis of costs and benefits, the
sensitivity analysis should be reviewed in order to allow to assess uncertainty related to the
most impactful variables behind the projections.
6) The coherence, complementarity and potential synergies with other initiatives - such as
the Cybersecurity Act, Data Act or the public procurement revision - should be analysed in
more detail.
Some more technical comments have been sent directly to the author Service.
7
(D) Conclusion
The DG must revise the report in accordance with the Board’s findings and resubmit
it for a final RSB opinion.
Full title Cloud and AI Development Act
Reference number PLAN/2025/815
Submitted to RSB on 10 December 2025
Date of RSB meeting 14 January 2026
EN EN
EUROPEAN COMMISSION
Brussels, 3.6.2026
SWD(2026) 502 final
PART 1/2
COMMISSION STAFF WORKING DOCUMENT
IMPACT ASSESSMENT REPORT
Accompanying the document
Proposal for a
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strengthening Europe’s cloud and AI
ecosystem (Cloud and AI Development Act)
{COM(2026) 502 final} - {SEC(2026) 502 final} - {SWD(2026) 503 final}
i
Table of contents
1. INTRODUCTION: POLITICAL AND LEGAL CONTEXT ............................................................... 4
1.1. Political context ................................................................................................. 4
1.2. Legal context ..................................................................................................... 5
2. PROBLEM DEFINITION .................................................................................................................... 6
2.1. Problem context ................................................................................................. 6
2.2. What are the problems? ..................................................................................... 9
2.3. What are the problem drivers? ......................................................................... 15
2.4. How likely is the problem to persist? .............................................................. 25
3. WHY SHOULD THE EU ACT? ........................................................................................................ 26
3.1. Legal basis ....................................................................................................... 26
3.2. Subsidiarity: Necessity of EU action ............................................................... 27
3.3. Subsidiarity: Added value of EU action .......................................................... 27
4. OBJECTIVES: WHAT IS TO BE ACHIEVED? ............................................................................... 28
4.1. General objectives ........................................................................................... 28
4.2. Specific objectives ........................................................................................... 28
5. WHAT ARE THE AVAILABLE POLICY OPTIONS? ..................................................................... 29
5.1. What is the baseline from which options are assessed? .................................. 29
5.2. Description of the policy options ..................................................................... 31
5.3. Options discarded at an early stage ................................................................. 56
5.4. Possible combination of options ...................................................................... 56
6. WHAT ARE THE IMPACTS OF THE POLICY OPTIONS? ........................................................... 57
6.1. Economic impact ............................................................................................. 57
6.2. Social impact ................................................................................................... 73
6.3. Environmental impact ...................................................................................... 74
7. HOW DO THE OPTIONS COMPARE? ............................................................................................ 78
7.1. Effectiveness .................................................................................................... 78
7.2. Efficiency ......................................................................................................... 85
7.3. Coherence ........................................................................................................ 86
7.4. Subsidiarity and proportionality ...................................................................... 89
7.5. Sensitivity analysis .......................................................................................... 89
7.6. Comparison per criteria ................................................................................... 90
8. PREFERRED OPTION ....................................................................................................................... 93
8.1. Outcome of comparison of policy options....................................................... 93
8.2. Application of the “One In One Out” (OIOO) Approach................................ 95
9. HOW WILL ACTUAL IMPACTS BE MONITORED AND EVALUATED? .................................. 97
ANNEX 0 - ENDNOTES ........................................................................................................................... 100
ii
Glossary
Term or acronym Meaning or definition
AI Artificial Intelligence
AZ Availability Zone
CEN European Committee for Standardization
CENELEC European Committee for Electrotechnical Standardization
CfE Call for Evidence
Colocation data centre
A data centre in which one or more customers install and
manage their own network or networks, servers and storage
equipment and service
Colocation data centre operator
An organisation who manages and leases/sells space, security,
network access, power and cooling capacity from a colocation
data centre to one or more customers who install and manage
their own network or networks, servers and storage equipment
and services
CSP Cloud service provider
Data Centre (DC)
A data centre (DC) is defined as a structure or a group of
structures used to house, connect and operate computer
systems/
servers and associated equipment for data storage, processing
and/or distribution, as well as related activities.
DSO
Distribution System Operators operate and manage local and
regional distribution networks, delivering electricity to end
users from the transmission system, such as Data Centres up to
a given size. See also TSO.
EC European Commission
Enterprise data centre operator
Enterprise data centre operator is a physical or legal person who
manages the entire enterprise data centre, including the building
and the use of the information technology services delivered
EED Energy Efficiency Directive
EU European Union
FLAPD Refers to the data centre markets located in Frankfurt, London,
Amsterdam, Paris and Dublin.
iii
FTE Full Time Employee
Hyperscaler
Hyperscalers are very large cloud and AI computing service
providers. They are characterized by their ability to provide
cloud computing at very large scale. Hyperscalers include
Amazon (Amazon Web Services), Microsoft (Azure), and
Google (Google Cloud Platform).
IA Impact Assessment
ICT Information and Communication Technology
IPCEI Important Project of Common European Interest
MS Member State(s)
OSS
Open Source Software is software that is released under a
license that allows users to view, modify, and distribute the
source code.
PUE
Power Usage Effectiveness, or PUE, is a metric used to
measure the energy efficiency of data centres. It is defined as
the ratio of total facility energy to the energy used by IT
equipment. A PUE of 1.2 implies that for each unit of energy
spent in powering IT equipment, 0.2 units are spent for non-IT
equipment such as cooling. A lower PUE indicates better
energy efficiency, as it means less energy is being used for non-
IT purposes.
Simpl Smart middleware platform for data spaces
SME(s) Small- and Medium-sized Enterprise(s)
TFEU Treaty on the Functioning of the European Union
TS Technical Specification
TSO
Transmission System Operators, or TSOs, manage the high-
voltage transmission networks that transport electricity over
long distances, ensuring stability and reliability across regions.
Above a given size, Data Centres connect to the grid directly
through TSOs. See also DSO.
4
1. INTRODUCTION: POLITICAL AND LEGAL CONTEXT
Today, cloud computing and AI are more than enablers of innovation. They have become the
driving forces reshaping industry, public service, and daily life in the EU. The cloud provides the
backbone necessary for all to access and deploy digital solutions efficiently, while AI unlocks
unprecedented opportunities for automation, data-driven decision-making, and personalised
services. A strategic approach to their uptake is thus essential for a competitive, secure, sovereign
and future-ready EU. Annex 8 explains the technical terms used in this assessment such as the
notion of cloud and AI computing services, which, in the context of this assessment, means
offering computing resources for the running (inference) of AI systems.
1.1. Political context
The Draghi report on the future of European competitivenessi recognised the importance of
sufficient access to cloud services and of increasing computational capacity in the EU, particularly
for the EU’s ability to develop and adopt AIii. The Competitiveness Compass calls for the EU to
provide sufficient cloud and data infrastructure required for AI leadership, as enablers of
competitivenessiii. The Economic Security risk assessment on AI technologies identified
dependencies on a limited number of foreign cloud providers and frontier AI providers as
significant vulnerabilities in the European AI ecosystemiv. The Cloud and AI Development Act
(CADA) is a headline digital policy initiative listed in the Mission letter to EVP Virkkunen,
alongside a single EU-wide cloud policy for public administrations and public procurementv. The
Competitiveness Compass calls for action in Europe to provide sufficient cloud and data
infrastructure required for AI leadership, as enablers of competitivenessvi. Furthermore, the AI
Continent Action Planvii announces the goal of tripling the EU’s data centre (DC) capacity within
the next five to seven years and emphasises that sovereignty and operational autonomy need a
greater reliance on highly secured EU-based cloud capacity. The Digital Decade Policy
Programme (DDPP) sets the target of 75% of EU businesses adopting cloud services and of 10
000 edge nodes being rolled out by 2030viii.
The European Parliament’s own-initiative report (tabled for plenary since June 2025) on
European technological sovereignty and digital infrastructure voices concerns about the EU's
excessive dependence on non-European actors in critical areas like cloud infrastructure. It calls on
the Commission to introduce the CADA to strengthen European data infrastructure, promote
European cloud service providers (CSP), build a European single market for cloud and AI, and
propose a definition for sovereign cloud and its scope of applicationix. In its conclusions of
December 2025 on European Competitiveness in the Digital Decade, the Council calls for CADA
to include common criteria for sovereign cloud services, allowing for addressing market
transparency and risks associated with dependencies, including extraterritorial effects of
legislation adopted by third countries for highly critical use casesx.
Several non-EU countries have adopted policy initiatives to develop AI computing capacity,
including DCs. In the US, building on a long tradition of supporting their cloud sector with
policiesxi and large public contractsxii, the July 2025 AI Action Planxiii massively boosts US DC
capacity and cloud services. An Executive Order on accelerating federal permitting of DC
infrastructurexiv establishes nation-wide fast-track procedures, lowers environmental protections,
and makes federal land available for DC build-out. The Executive Order on promoting the export
of the American AI Technology Stack establishes federal support for full-stack AI export
packages bundling AI-optimised computer hardware, DC storage, cloud services, and networking,
which will be exclusively sold in US-providers-only packagesxv. China launched the
infrastructure project “Eastern Data, Western Computing” in 2022, coordinating DC construction
by concentrating facilities in the West of the country. This led to a surge in government
procurement of DCsxvi. In July 2025, the UK proposed AI Growth Zones to better equip the UK
5
for running training and inference and support UK companies in developing sovereign, sustainable
and secure compute technologies and servicesxvii. The UAE pursues investments in renewable
energy, power transmission, and hyperscale-ready infrastructure to expand its DC capacityxviii.
1.2. Legal context
The EU lacks a framework to foster the strategic investment in computing capacity beyond AI
Factories and Gigafactories1. The DDPPxix only sets a deployment target for edge nodes, but not
for DCs. The legislative framework for DCs focuses on enhancing their sustainability through
transparency measures without explicitly incentivising deployment. The Energy Efficiency
Directive (EED)xx establishes an annual sustainability reporting and lays the basis for a rating
scheme. The Taxonomy for Sustainable Finance enhances transparency on DC sustainability for
financial market participantsxxi. DCs must comply with minimum performance and information
rules of the Ecodesign Regulationxxii. While DC projects are not subject to mandatory
environmental impact assessments based on EU rules, such assessments are often required by
Member States2. The upcoming Industrial Accelerator Act will create clusters for accelerating
industrial activity for the manufacturing sector, from which DCs will not benefit. Other EU
initiatives target key input factors for DC deployment: the upcoming Digital Networks Actxxiii will
improve connectivity; the Grids Packagexxiv will ensure that electricity grids can serve growing
demands, with focus on improving permitting, planning but also providing tools to accelerate grid
connections procedures via a dedicated guidance on grid connections; the Savings and
Investments Unionxxv will improve access to capital in the EU. While these frameworks can
benefit DC deployment, they are not tailored to the sector’s specific needs. The upcoming
Regulation on accelerating and streamlining environmental assessments establishes a toolbox with
provisions for faster environmental assessments, applicable to strategic sectors when sectoral
legislations refer to it, something that CADA will leverage for DC projects.
Similarly, the EU lacks a framework for incentivising the uptake of European cloud and AI
computing services and for ensuring security of supply. The DDPP sets high-level adoption
targets3, and the Apply AI strategyxxvi supports AI adoption in strategic sectors, but without
concrete measures geared at European services. Other existing frameworks address market
practices: the Data Actxxvii regulates cloud switching and interoperability. The Digital Markets
Actxxviii considers cloud services as core platform services where providers can be designated as
gatekeepers and subject to specific obligations4. The Digital Networks Act will address the
scenario where a CSP operates a connectivity network. The AI Actxxix sets out risk-based rules for
providers and deployers of AI systems and general-purpose AI models, fostering trustworthy AI
but without addressing computing. Other frameworks address cybersecurity: the Cybersecurity
Act (CSA)xxx, currently under review, enables the adoption of an EU-wide cybersecurity
certification scheme for cloud services5 and addresses supply chains by tackling high-risk vendors
but without addressing public procurement. The Digital Operational Resilience Actxxxi targeting
financial entities, and the NIS2 Directivexxxii defining sectors of high criticality, require entities
like CSPs to implement risk management and other security measures. The use of cloud and AI
computing services in the public sector is horizontally governed by the Public Procurement
framework, currently under revision, which enshrines transparency, equal treatment, open
1 These initiatives focus on High Performance Computing (HPC) and do not address the need for more decentralised computing capacity. 2 According to the Environmental Impact Assessment (EIA) Directive, changes to which may come from the Environmental Omnibus: Directive -
2014/52 - EN - EIA - EUR-Lex. 3 75% business adoption of cloud, AI, or big data by 2030. 4 So far, no provider has been designated as a gatekeeper for the provision of cloud services, but on 18 November 2025, the Commission opened
three market investigations on cloud computing services under the DMA. 5 This requires an Implementing Regulation. ENISA has been working on developing the European cybersecurity certification scheme for cloud services (EUCS) since 2019, which has not been adopted yet. Two technical specifications by CEN-CENELEC have resulted from this work on the
security requirements and the accreditation of the conformity assessment bodies and the conformity assessment methodology.
6
competition and sound procedures as well as respect for EU’s international commitmentsxxxiii.
However, these initiatives fall short of addressing sovereignty considerations in this sector.
2. PROBLEM DEFINITION
The analysis conducted identified two key problems that warrant policy attention. This chapter
presents these problems in detail, exploring their underlying root causes structured around four
main problem drivers. It further assesses the associated risks and potential consequences should
these issues remain unaddressed. To better define and characterise the identified problems, it is
useful to first examine the competitive dynamics in the cloud computing market and the
functioning of cloud service demand and supply, as a critical contextual element.
2.1. Problem context
Cloud computing emerged in the early 2000s in the United States, with Amazon Web Services
(AWS) being established in 2002 and commercial cloud offers from AWS, Google and others
becoming popular from 2008 onwards. The early 2010s set the stage for the massive adoption of
cloud servicesxxxiv, which accelerated further in subsequent years. The global cloud and data
infrastructure market grew by around 35% per year since 2016xxxv, with expected growth rates
above 20% for subsequent years. In the EU, this high-growth environment led the share of
enterprises buying cloud services to increase from 18% in 2014 to 53% in 2025, with adoption
almost tripling in a decade; for large firms, the figure exceeds 80%xxxvi. This rapid growth in
market demand can be linked to three major trends:
• The digitalisation of the economy and a structural shift from on-premises IT models:
businesses across sectors have migrated from traditional on-premises infrastructure to cloud
solutions, considered to offer lower upfront costs, more flexibility and a richer ecosystem of
services in a single interface. Part of this cloudification has also been supply-driven, as
providers pushed from installed licences to cloud subscription models6. Major CSPs,
massively supported by system integrators, expanded their service portfolios, pricing models
and migration programmes in ways that created and deepened the demand for cloud services.
This initial trend from on-premises to infrastructure-as-a-service (IaaS) has seen a second
wave towards more advanced platforms deployed as platform-as-a-service (PaaS) and
software-as-a-service (SaaS). The growing weight of PaaS and the shift towards AI (mostly
deployed as SaaS) has been a key market dynamic in the last years and remains so to date.
• The emergence of new cloud native digital services: demand has increased with the rise of
cloud-native platforms such as social media, video and music streaming, and a broad range of
SaaS applications such as Customer Relations Management tools.
• Artificial Intelligence and data-driven business models: the diffusion of AI, advanced analytics
and data-intensive applications have increased demand for compute power, often requiring
processing capacity close to end-users to meet low-latency requirements, thus reinforcing the
shift towards cloud and edge solutions.
From 2017 to 2021, the European cloud market expanded rapidly, notably during Covid-19, which
increased the use of digital services and boosted demand for cloud computing across the EU.
However, most of the incremental demand was captured by US hyperscalers, whose share rose
from around half to two-thirds of the market, while European providers’ collective share nearly
halvedxxxvii. During this period, European CSPs were predominantly national or regional players
serving domestic markets. They reacted to foreign competition by either specialising in use cases
6 For example, Office 365 monthly active users grew from around 60 m in 2015 to 200 m in 2019, while traditional Office product revenues fell by around 21%. See: Office 365 Number of Users Reaches 345 Million Paid Seats and FY23 Q4 - Productivity and Business Processes Performance -
Investor Relations - Microsoft
7
with stronger data sovereignty and privacy demandsxxxviii or by developing partnerships with the
hyperscalers7, rather than matching the broad footprint of US providers. This was the case too of
European Telco providers who ventured into cloud services, but mostly partnering with US CSPs,
becoming de facto resellers. The 2021 European industrial technology roadmap for the next
generation cloud-edge offering, prepared by key European players, argued that the European
digital market was “fragmented into local realms, individually lacking the critical mass for
players to scale and compete” with the US and Chinaxxxix. Despite strong revenue growth, the
market share of European providers declined. By contrast, large US providers have been able to
benefit given their early, large-scale deployment of cloud offerings, which put them in a position
to capture this market growth. This was made possible by three elements.
Firstly, US CSPs were able to build on an early growth driven by US government adoption. In
2013, the Central Intelligence Agency awarded a USD 600 m contract to AWSxl, followed by the
award of a fifteen-year multi-billion-dollar contract for the development of the “Commercial
Cloud Enterprise” to AWS, Google, IBM, Microsoft and Oracle. The Department of Defence
awarded additional sizeable contracts to these providers for developing secure cloud services or
enabling the migration to the cloud of agencies like DARPA8. These early and large-scale public
contracts ensured fixed revenues and allowed these providers to grow their portfolio, often with
particularly secure services resulting from the stringent requirements of the US administration.
These providers carried their first-mover advantage to Europe demand for cloud services was just
emergingxli, driving early European cloud users to turn to US providers9. As well, in the shift from
on-premises to cloud solutions, hyperscalers extensively used partner networks. By offering
dedicated training and skill certifications, they engaged in partnerships with consulting firms and
resellers, through which they could rapidly expand in local marketsxlii.
Secondly, the absence of a thriving tech industry in the EU prompted US CSPs to leverage their
domestic advantage by partnering with large global technology companies when expanding into
the European market. Indeed, figures show that cloud adoption in the EU is driven by companies
working in the ICT sectorxliii, which are often not European and tend to buy the same cloud
services as they buy domestically, i.e. US CSPs. Some examples: Netflix, which serves more than
50% of the European video-on-demand market, relies exclusively on AWS for its core cloud
infrastructurexliv; the Amazon (retail) Marketplace serves as an anchor customer for AWS, its own
cloud services offering.
With respect to market dynamics, competition in cloud services, like other capital-intensive
network industries, is characterised by distinct supply and demand side elements.
On the supply side, cloud markets are defined by considerable sunk costs associated with data
centre deployment, with long investment times and substantial economies of scale and scope.
These characteristics tend to benefit large providers which can fund expensive compute capacity
and distribute costs across a wide customer base and a diverse service portfolio. When looking at
infrastructure, high fixed costs constitute a major barrier to entry. The cloud and AI infrastructure
market is capital-intensive: building and equipping data centres requires large upfront investment,
and providers are able to secure financing if they expect sufficient demand and market share gains.
Differences in deployment procedures among Member States create transaction costs within the
single market, affecting the profitability of new projects. These costs are more easily absorbed by
7 For example, German provider plusserver working with AWS, Azure and Google Cloud in 2019 by offering hybrid solutions interconnected with
hyperscalers. See: https://www.juniper.net/content/dam/www/assets/case-studies/us/en/plusserver.pdf 8 An excerpt of awarded contracts from different US Departments and agencies to US hyperscale cloud providers: C2E - $600 m (CIA, 2013- present), Wild & Stormy – USD 10 bn (NSA, 2021 – 2023), JWCC – USD 9 bn (DoD, 2022 – 2028). See also : DARPA plans shift from AWS and
on-prem to fully cloud by 2022; General Dynamics again wins DOD's cloud email & collaboration contract; Big Tech and the US Digital-Military-
Industrial Complex - Intereconomics. 9 In the Netherlands, for example, the strong reliance of the public sector on hyperscalers is laid down by the Netherlands Court of Audit, which
analyses selected critical cloud contracts awarded by the Dutch central government.
8
larger providers but remain prohibitive for smaller firms. Access to funding is also slower and
more complex for small enterprises. By contrast, AWS, Microsoft and Google were able to
collectively invest around EUR 12 bn in European infrastructure in 2020 alone, marking a 20%
increase compared to the previous yearxlv. This investment capacity was underpinned by vertically
integrated businesses and diversified revenue streams. The French Autorité de la Concurrence, in
its 2023 opinion on cloud services, emphasised that hyperscalers benefit from “conglomerate”
structures, i.e. their presence in multiple digital markets allows them to develop credit systems and
discounts using their market power to accelerate cloud growth. Similarly, a more recent work by
the OECD on competition in the cloud market notes that hyperscalers are best equipped to
mitigate the risks of aggressive cloud expansion by “portfolio diversification or cross-subsidising
losses”10. On the innovation side, the asymmetry with European players is also evident. The EU
Industrial R&D Investment Scoreboard showed that by 2018-2020 Amazon, Alphabet and
Microsoft were among the top global corporate R&D investors, each spending in the order of
billions of dollars per year. In the same period, Europe’s largest R&D spenders were concentrated
in the automotive and pharmaceutical sector, while European cloud providers like OVHcloud were
small companies with revenues in the hundreds of millions and modest R&D budgetsxlvi. This
disparity in innovation spending, combined with hyperscalers large capital expenditures, projected
to reach USD 335 bn in 2025xlvii, has given them an important advantage over smaller providers to
offer broader service portfolios and integrated ecosystems.
On the demand side, customer choices are shaped by two key factors: the value placed on
flexibility and the breadth of services available through single platforms operating seamlessly. On
these two key factors, hyperscalers have been better placed than European providers from the
start. Although European cloud offerings encompass a diverse range of services, customers need
to collaborate with multiple providers to match the quality and breadth of services offered by
leading global cloud providers. End-users desire simplicity and have grown accustomed to one-
stop shops delivering everything from Infrastructure-as-a-Service (IaaS) to Software-as-a-Service
(SaaS) on a global scale, a level of integrated service that hyperscalers readily supply. In contrast,
European providers typically have more limited catalogues, often focused on specific
infrastructure or industry niches, making it challenging to secure substantial, multi-country
enterprise deals. In several European industries, value chains are organised around networks of
smaller providers that combine their specialised products across the single market, supported by
common standards and technical specifications. This was achieved in the telecom sector after the
introduction of the GSM standard. However, a similar market structure has not materialised in the
cloud sector. As mentioned above, instead of pooling resources and federating, European
providers decided to (i) build partnerships with US hyperscalers, (ii) focus on a specific region, or
(iii) specialise in a single layer of the cloud stack, e.g. IaaS or Saas offerings, rather than offering a
comprehensive portfolio. By contrast, US hyperscalers deliver an integrated, end-to-end service,
operating as “IT supermarkets”xlviii. The French competition authority clearly noted that
hyperscalers’ large ecosystem of offers implies that, for several workloads, competition takes
place “for the market” rather than “on the market” as customers tend to choose a single supplier
able to cover their entire needs.
The repercussions of these competition dynamics result in a rigidity when it comes to switching
providers and considering alternative offers. Leading providers impose complex pricing structures,
egress fees and restrictive licensing terms. When switching providers, customers face high costs
stemming from egress charges, the use of proprietary data formats or APIs and long-term
contractsxlix, i.e. different forms of vendor lock-inl. Even if a better or cheaper provider exists,
customers may therefore not switch, weakening competitive pressures and resulting in alternative
10 OECD, “Competition in the provision of cloud computing services”
9
providers struggling to attract customers due to factors outside of their control11. The Dutch
competition authority’s cloud market study concludes that users encounter technical hurdles to
portability and incur significant financial costs for data transfer, which collectively make
switching providers challenging and effectively lock users into the chosen cloud provider for
lengthy periods, contributing to market consolidationli. Moreover, once a provider becomes the
‘default’ or ‘safe choice’, the customer starts creating the tooling and culture around this provider,
resulting in inertia as an effective blocker of new incumbents. In this context, EU providers
struggle to attract customers away from well-established US providers. The challenges described
above, along with difficulties in partnering with system integrators, consultancies and
intermediaries to promote European solutions, have collectively contributed to the gradual erosion
of market share for EU CSPs.
Last but not least, there are signals of new opportunities for European providers to gain market
shares. AI adoption, the third demand driver identified above, is bringing change to the market
landscape, and competitive positions could change. While US providers are already well
established mostly thanks to their integrated offers, European providers might still find
opportunities to capture growth, especially in relation to more specialised, sector-specific offers,
where their proximity to customers and ability to provide safeguards in terms of data localisation
and integrity are important. The Apply AI strategy adopted by the Commission is a signal in that
direction. AI has the potential to alter industry business models, with cloud and AI services
extending beyond basic IT commodity functions to become integral parts of a company’s
competitive edge. Furthermore, the issue of sovereignty, exacerbated by geopolitical
considerations, has gained prominence and could be an opportunity for European cloud and AI
providers. According to Gartnerlii the sovereign cloud IaaS market is forecast to grow at a yearly
rate of 38% in the next five years, with developments in terms of moving away from global cloud
providers and new business migrating to a sovereign cloud environment. The positioning of
European cloud and AI providers as sovereign is, however, today hindered by a lack of clarity and
lack of harmonisation in the market in terms of how sovereign services are defined.
2.2. What are the problems?
2.2.1. Problem 1 - Limited and geographically concentrated availability of
computing capacity in the EU
AI is driving an unprecedented demand for computing capacity, not only for the High-
Performance Computing (HPC) capacity required to train models, but also for the capacity to
enable inference, fine-tuning and service integrationliii. The European AI market is projected to
exceed EUR 300 bn by 2030liv, growing at more than 26% between 2025 and 203012. Beyond AI,
adoption of cloud computing and other digital services continues to accelerate, adding further
pressure on the available computing capacity13. In 2025, uptake of new DC capacity in Europe14 is
expected to reach a new high of 854 MW15, exceeding new supply for the third consecutive yearlv.
Vacancy rates in major EU DC hubs have declined to historical lows, and the share of co-location
11 While these contextual factors are referred to in the problem drivers described below, they are not themselves considered as drivers for this
analysis as they are addressed through a dedicated intervention under the Data Act (see section 7.3 and annex 7) and through competition cases. See
for example the European Commission’s investigation into Microsoft Teams: Statement of Objections to Microsoft. 12 Enterprise adoption of AI technologies remains limited at 13.5% according to the latest data from the Digital Decade Policy Programme, partly
due to infrastructure constraints and cost barriers. The Digital Decade 2025 report highlights the need for the data centre industry to expand and
adapt to support the rapid growth of AI technologies. 13 In 2025, per Eurostat, EU business cloud uptake stood at 52.7% - far from the 2030 target of 75%, but that also includes the adoption of cloud, AI
or data analytics. As more European businesses adopt cloud and AI computing services, demand for DCs is thus expected to rise further. 14 When looking at the DC market, Europe typically includes the UK. 15 DC capacity is typically expressed in megawatts (MW) or gigawatts (GW) because power availability plays a key role for both the operation of
the servers and the cooling systems.
10
capacity that is pre-leased16 before becoming operational continues to rise. These are clear signs of
high demandlvi.
Across the EU27, the expansion of data centre capacity is not keeping pace with this rapidly
growing demand. Despite increased investmentlvii and installed capacity expected to reach 12.4
GW in 202517, available supply remains insufficient, resulting in an estimated gap of almost 3 GW
relative to current demand.
Moreover, the existing capacity is concentrated in a limited number of established hubs,
mostly in Northern and Western Member States: in 2025, Germany (Frankfurt), France (Paris), the
Netherlands (Amsterdam) and Ireland (Dublin) account for 65% of the EU27 DC market18 (Figure
119). These locations have historically provided more favourable factors for DC deployment, e.g.
strategic geographic location, proximity to end users, and connectivity to other world regions.
Ireland, for example, is geographically positioned as a gateway between Europe and the US, with
extensive undersea fibre-optic cable networks. This makes it an ideal location for low-latency
transatlantic data transfers for US CSPs serving European markets from overseas. Ireland’s low
corporate tax rate and supportive government policies have attracted significant foreign direct
investment, particularly from US tech companies. These companies, including the cloud
hyperscalers, have established major operations in Ireland, contributing to the rapid expansion of
nearby data centre infrastructure20.
Figure 1. Data centre capacity across EU 27 MS in 2025
Source: Technopolis et al. (2025)lviii
Expressed per 100 000 inhabitants, overall capacity amounts to 2.75 MW per 100 000 people and
is concentrated in a few Member States, with Ireland, the Netherlands, Denmark, Sweden, Finland
and Luxembourg standing out (
Figure 2).
16 Pre-leased capacity in DC means that customers commit to renting space and power before the facility becomes operational. Market information
points to a growing share of data centre spaces already reserved before delivery, reflecting very high demand and tight supply. 17 Technopolis Group, Wavestone, Timelex, STL Partners, OpenForum Europe and KAPA Research (2025), "Study: Cloud and AI". The methodology is based on all known commercial data centre sites listed in the Data Center Map, additional sites identified through the survey and
any publicly known hyperscaler sites. The figures do not include private enterprise sites, under which most dedicated HPC facilities fall. 18 Technopolis Group et al. (2025), "Study: Cloud and AI". 19 Core EU DC hubs include Frankfurt, Amsterdam, Paris and Dublin. This concentration is explained by operators leveraging metro areas and
exploiting the best locations in terms of connectivity (Dublin, Paris, Frankfurt, Amsterdam), proximity to economic hubs (Paris, Frankfurt,
Amsterdam) or low corporate tax (Dublin). 20 In a second phase, hyperscalers made substantial investments in locations with strong sectoral demand, such as Frankfurt for banking services
and Belgium for the pharmaceutical industry.
11
Figure 2. Data centre capacity per 100 000 people across EU 27 MS in 2025
Source: Technopolis et al. (2025)lix
In addition, the EU lags behind other regions in both scale and ownership of digital infrastructure
(see Figure 3). In 2022, Europe had approximately 1 MW of installed data centre capacity per 100
000 people, while the US had 12 MWlx. The situation persists today: despite similar GDPs, in
2025 the EU accounts for only 20% of global installed data centre capacity compared to the US
(42%)lxi.
Figure 3. Comparison of GDP and data centre capacity for leading markets in 2025
Source: Technopolis et al. (2025)lxii
The existing capacity gap and strong geographical concentration point to structural inefficiencies
in the allocation of resources in the market for data centre capacity, with effects already visible
today:
The limited supply has led to rising prices for existing capacity, negatively affecting businesses
which rely on such capacity. Since 2022, average asking prices in the European colocation
markets have increased by 51% for 100 kW leases lxiii. New co-location capacity is often leased to
large service providers: Amazon Web Services (AWS), Microsoft, Google21. By 2028, they are
expected to drive 65% of the demand for DCs in Europe, an increase of 12% during the same
timeframelxiv. This reinforces the competition dynamics discussed above: when most of the new
capacity is effectively reserved by a handful of players, they can shape where and how new
infrastructure is built, secure access to suitable sites and further capitalise on economies of scale.
Already today, vacancy rates in Europe’s top DC markets are at a record low of 7.4%lxv. Due to a
lack of available space, co-location providers are expected to raise prices in 2025 by 10% in
leading DC marketslxvi. The negative effects on businesses’ ability to access capacity are
21 Co-location data centre operators often lease their facilities to hyperscalers to obtain a quicker return on investment.
12
particularly pronounced in regions with a high concentration of DCs, where businesses are faced
with higher prices.
As noted above, geographic concentration puts a strain on affected regions: in Ireland, DCs
accounted for 22% of electricity demand in 2025, up 5% from 2015. The resulting grid stress has
led local TSOs to enact a de facto moratorium for new DC applications in Dublin due to fears of
overloading the grid and compromising energy securitylxvii. Municipalities in Noord-Holland have
also enacted several moratoria on building new DCs, the latest one in 2023 in Amsterdamlxviii. In
the case of Dublin, the moratorium has reportedly stalled EUR 8-10 bn in planned DC
investments. The effects of grid stress also impact other user groups22. In Luxembourg, the plans
for a large DC by Google have been subject to a review of its energy grid planning that includes a
new connection to Germanylxix because local electricity generation was not sufficient. This
situation, with the saturation of existing infrastructure, has further slowed deployment, raised
energy system pressures and diverted investment, rather than leading to an efficient redistribution
of capacity across the Union.
While cloud and AI computing services can be technically delivered cross-border, regions with a
low DC presence are also disadvantaged by this geographic imbalance, as exemplified by the
higher prices in regions with low DC presencelxx. Moreover, the lack of nearby computing
capacity drives up latency, limiting the availability and quality of low-latency services, thus
placing local end-users at a competitive disadvantage compared with regions that have better
access to DC capacity23 (see also section 2.3). It also points to under exploited investment
opportunities if comprehensive business cases can be built around these cases. Over time, this can
slow digital transformation in the affected Member States and weaken overall competitiveness
within the Digital Single Market. Stakeholders have also warned that this limited availability of
computing capacity in the EU acts as a barrier for the development and uptake of cloud and AI
computing services: for instance, Mistral AI has warned that a lack of DC capacity could become
a roadblock for developing and applying AI models in Europelxxi. Current market dynamics risk
reinforcing existing concentrations, increasing regional disparities and limited access to
computing resources for business and public authorities outside main hubs.
2.2.2. Problem 2 - Dependence on cloud and AI computing services supplied by non-
European24 providers
The European cloud services market is growing significantly. It was worth around EUR 70 bn in
2022 and is estimated to reach over EUR 200 bn by 2028lxxii. In 2024, the European Infrastructure
as a Service (IaaS) and Platform as a Service (PaaS) market was dominated by three US
companies, the so-called hyperscalers (Table 1). Worldwide, AWS held a market share of 32% in
Q2 2024, Microsoft 23% and Google Cloud 12%. No other provider held more than 4%lxxiii.
In Europe, those three providers account for around 70% of the market, while the largest European
providers (SAP and Deutsche Telekom) each serve 2%lxxiv. The European cloud market has
grown, but the market share of European providers has decreased from 29% in 2017 to 15% in
2022 and has since remained stablelxxv. A recent study for the European Parliament reaches the
same conclusion, noting that US firms dominate all major software layers, including cloud, and
that European providers account only for a small share of the infrastructure marketlxxvi. Despite
22 In Ireland, for example, interest representatives point to increasing delays for electricity connection of housing projects: Govt warned of rising
household bills as data centres strain grid. 23 Taking the example of Microsoft’s Azure regions, round-trip latency (the time it takes for a data pack to travel from one point in the network to another and back again) from Poland (Central Europe) to Frankfurt (Western Europe) is ca. 10–15ms, and latency from Poland to Amsterdam or
London is ca. 15–20ms. By contrast, latency within Western Europe (e.g. Frankfurt to Amsterdam) is typically <5ms. A fintech business in Warsaw
thus faces significantly higher latency than a competitor in Western Europe. 24 Whereas section 2.1.1. deals with the presence of computing capacity in the EU, including capacity provided by non-EU companies; this section
deals with services provided by companies headquartered in the EU.
13
European providers playing an important role in some Software as a Service market segments,
non-EU providers dominate in important fields, such as office automation and productivity
software25.
Table 1. Market leaders for cloud services (IaaS, PaaS and hosted private cloud revenues), Q2 2024
World US China26 Rest of APAC Europe Rest of World
#1 Amazon Amazon Alibaba Amazon Amazon Amazon
#2 Microsoft Microsoft Tencent Microsoft Microsoft Microsoft
#3 Google Google China Telecom Google Google Google
#4 Alibaba Oracle Huawei NTT Oracle Salesforce
#5 Oracle Salesforce China Unicom Alibaba Salesforce Oracle
#6 Salesforce IBM China Mobile Fujitsu IBM IBM
Source: Synergy Research Group
Hyperscalers thrive in the European cloud market thanks to their global scale, significant financial
resources, and overall ecosystem gravity, composed of integrated digital ecosystemslxxvii,
partnership programmes and marketplaces, among others.
On the supply side, this problem impacts European providers of cloud and AI computing services
in terms of foregone commercial opportunities. Astères approximates their magnitude by
estimating that EU companies’ annual purchases of cloud software add EUR 264 bn to the US
economylxxviii. On the demand side, this problem impacts private and public sector users of cloud
and AI computing services. Some reports suggest that, by relying so strongly on a small number of
providers, users pay more for their cloud and AI computing services than by relying on
alternatives27. In its most basic form, this dependence can lead to significant economic costs.
Without alternative, users are defenceless in light of price increases. This is illustrated by
Broadcom’s acquisition of VMWare (a provider of leading virtualisation technology, for which
little European alternatives exist), which resulted in unilateral licensing changes and price
increases of 800% to 1500% according to users28. More generally, there is evidence suggesting
that the dependence on non-European providers may cause European users to over-pay, with some
reports suggesting that European providers offer digital resources at lower prices29 and with lower
egress charges.
European AI computing service providers lag their global competitorslxxix, also due to their
competitive disadvantage in accessing computing service providerslxxx.
Dependence also introduces significant tail risks: where users rely on a small number of providers,
cloud outages can have far-reaching consequences, including business interruptions, and
substantial financial or data losseslxxxi. Strong reliance on a small set of providers also means that a
single failure can simultaneously affect several critical services, as seen in the recent Crowdstrike
incident or AWS outageslxxxii. It also limits the EU’s operational autonomy and system resilience
as these providers may be exposed to third-country policies restricting service access, for example
in the context of sanctions or economic coercion. This risk recently materialised in the suspension
of service provision to the Chief prosecutor of the International Criminal Court, on whom the US
had previously imposed sanctionslxxxiii. This case illustrates the challenges which exposure to
third-country policies can cause to operational autonomy and system resilience in the EU. Another
illustrative example is the planned takeover of the Dutch cloud provider Solvinity by the US IT
25 For example, in 2023, SAP held 49.6% of the specialised market for Travel and Expense Management Software. 26 In China, foreign invested companies are not allowed to provide so-called Internet Data Centre Services according to the Promulgating the Classification Catalogue of Telecommunications Services and must rely on local Chinese partners in the form of a technology cooperation. q For example, Leitmotiv Digital finds that European providers offer digital resources at prices that are five to ten times lower than the current
incumbents: Leitmotiv - Toward our Digital Future. 28 https://www.theregister.com/2025/05/22/euro_cloud_body_ecco_says_broadcom_licensing_unfair/ 29 https://leitmotiv.digital/publications/breaking-the-cloud-monopoly
14
company Kyndryllxxxiv. Solvinity offers cloud services to the IT company of the Dutch
administration, including the back-end of digital wallet, used by 16.5 million Dutch citizens to
verify their identities in order to get access to the tax administration and other departments. A non-
European policy affecting the provision of Kyndryl’s services in the EU would thus cause
disruption for 90% of Dutch citizens. Threats to operational autonomy and of data access are
particularly concerning in highly critical use cases relying on cloud and AI computing
serviceslxxxv. For example, their use in healthcare, defence and certain public sector services often
involves the processing of highly sensitive data. Service interruption in these sectors can have
significant adverse effects on the EU economy and society. Some public sector actors are
undertaking steps to limit exposure to such risks. For example, the Dutch parliament called on the
government to reduce its reliance on US cloud serviceslxxxvi, and France mandates sensitive data to
be exclusively stored and processed using SecNumCloud certified services30 provided either
solely by EU providers or Joint Ventures between EU undertakings and US CSPs (see section
2.3.4). This issue was also recently illustrated by the postponement of Finland’s electoral
management system cloudification31, previously awarded to AWS, and France’s switch from US
solutions to open source equivalents to serve public sector video conferencing needs32. The
European Central Bank requires banks to use hybrid architectures and a multi-cloud
approachlxxxvii. The US requires Federal Agencies to use only cloud services authorised under the
Federal Risk and Authorization Management Program (FedRAMP)33. While dependence is widely
considered as a critical risk to service continuity and operational autonomy, and demand for
sovereign cloud solutions is growing34, a coherent framework to address this challenge is still
lacking at EU level.
Dependence on US providers also exposes EU user data to extraterritorial laws and potential US
government access resulting from the US Clarifying Lawful Overseas Use of Data (CLOUD)
Act35 in the context of criminal proceedings, or Section 702 of the US Foreign Intelligence
Surveillance Act36. Access can be granted unilaterally, without the involvement of EU judicial or
public authorities. In a sworn testimony before the French Senate, a representative of Microsoft
affirmed that the company could not guarantee that French data would not be transmitted to US
authorities, even in the absence of explicit authorisationlxxxviii. Any data managed by a US-
headquartered provider or its subsidiaries is potentially exposed, including sensitive business
information, intellectual property, and personal data of EU citizens. Such exposure, especially of
sensitive business or public data, can raise significant concerns when considering strategic entities,
public authorities, and high-profile individuals. More generally, a lack of trust in cloud and AI
computing services is slowing down European users’ adoption of other digital serviceslxxxix. This
lack of trust is also caused by technical aspects such as not having control over the supply chain,
30 For instance, the platforms for electronic invoices shall be ISO/IEC 27001 certified and stored on one of the providers qualified by
SecNumCloud: Facturation électronique et plateformes partenaires. 31 See Finland's Ministry of Justice delays plans for cloud migration - DCD 32 Souveraineté numérique : l’État généralise « Visio », sa solution de visioconférence sécurisée et souveraine à destination des agents publics –
Presse – Ministère des Finances 33 FedRAMP is an authorisation process for cloud services used by US federal agencies. Obtaining such authorisation is mandatory for all CSPs wishing to work with US federal agencies. It provides a standardised approach for the security assessment of cloud services in three impact levels
(low, medium, high). In terms of European providers, three SAP products are authorised at level ‘moderate’ and the Accenture Insights Platform
(originating in US, headquarters in Ireland) is authorised at level ‘high’. 34 84% of European cloud users are already using or planning to use sovereign cloud solutions. See: How Digital Sovereignty Is Influencing Cloud
Solution Choice 35 The CLOUD Act facilitates the access of US law enforcement authorities to electronic data held by service providers that fall within US jurisdiction in the context of criminal proceedings. It affects providers of electronic communication services and remote computing services that are
subject to the authority of US courts, either because they are established in the US or because they have a sufficient presence to be subject to US
personal jurisdiction. When presented with a valid legal process, including a warrant for the content of communications meeting the high standards of probable cause, these entities can be obliged to disclose the contents of electronic communication and any related record or other information
pertaining to their customers, regardless of whether the data are located within or outside of the US. 36 FISA Section 702 permits the US government to conduct targeted surveillance of foreign persons located outside the US to acquire foreign intelligence information. Under Section 702, the US Attorney General and Director of National Intelligence may issue directives compelling US
electronic communication service providers to provide such information, including via bulk data collection.
15
not being able to carry out audits and penetration testing, or the uncertainty of where the data is
locatedxc.
2.3. What are the problem drivers?
The two identified problems are closely interlinked, as they relate to consecutive stages of the
same value chain, with a clear supply-demand relationship: cloud and AI computing services
require adequate underlying infrastructure, while data centre investments rely on expected demand
for such services. Problem 1 concerns the physical infrastructure layer, involving data centre
operators, utilities, local authorities and investors. Problem 2 relates to the downstream market for
cloud and AI computing services, characterised by distinct market structures, competitive
dynamics and regulatory frameworks. Accordingly, the analysis of the root causes behind these
two problems is presented separately, while acknowledging their interdependence. Four
underlying drivers are identified: one specific to Problem 1, two specific to Problem 2 and one
common driver that affects both problems.
Problem 1 relates to the limited and geographically concentrated distribution of compute capacity
in the EU. It stems from two underlying factors: one related to market dynamics and the other to
regulatory fragmentation. On the first one, first-mover advantage of non-EU providers reinforced
by path dependency and network effects has led to market entry barriers that limit the scale and
scope of EU cloud and AI computing service providers (PD1), thus hindering the deployment of
capacity in the EU territory. As for the second factor, regulatory fragmentation and bottlenecks of
a different nature have further slowed down the expansion of data centres (PD2), reducing
international investors’ interest, and thus contributing to the limited computing capacity, while
also reinforcing geographical imbalances.
Problem 2 concerns the dependence on cloud and AI computing services supplied by non-
European providers. Like problem 1, it is driven by the lack of scale and scope of EU cloud
service providers (PD1), which makes reliance on non-European providers inevitable. The limited
public sector uptake of cloud and AI computing services supplied by European providers (PD3),
even in segments where local presence is essential, further hampers the creation of a large, unified
demand for such services. This weakens EU providers’ ability to invest, which results in foregone
economic opportunities and deepens dependence. Finally, the absence of clarity around the
concept of sovereign cloud and AI computing services (PD4) keeps European providers from
commercially leveraging sovereignty as a distinguishing factor for themselves, further
contributing to the problem.
Figure 4. Problem tree
16
2.3.1. Problem Driver 1 (PD1): Lack of scale and scope of European cloud and AI
computing service providers
An important driver behind the limited and geographically concentrated availability of EU-native
computing capacity and services is rooted in the comparatively small scale of European providers
relative to global hyperscalers, as a result of the rapid development of the data centre and cloud
sector in the EU. Notably between the years 2017 and 2021, this led to a significant market
concentration around only a few non-European market incumbents well-placed to respond to the
growing demand at scale which was out of reach for their European competitors.
As discussed under section 2.1, historically, the earlier cloud computing services were developed
in the United States. Their early scale advantage reinforced their position as the world’s leading
data centre hubxci, which US providers leveraged for global service provision. Hyperscalers
invested aggressively in infrastructure, and building networks of data centres connected via
submarine cables that enabled data flows into European marketsxcii. This first-mover advantage
reinforced their leadership in the continent, setting a high entry and growth barrier for European
providers.
Hyperscalers leveraged their scale to standardise services, invest in innovation and accumulate the
financial resources to deploy compute capacity globally: today the AWS cloud spans 117
availability zones (AZs) within 37 geographical regionsxciii. By comparison, OVHcloud has 37
AZsxciv. Due to their comparatively higher scale and financing advantagesxcv, the ongoing
expansion of data centre capacity in the EU is driven by non-European CSPsxcvi. As demand for
local capacity grows, hyperscalers are able to secure land and grid access for their own data
centres and (pre)lease large amounts of co-location data centre space, thus becoming anchor
tenants for larger sites, while European CSPs struggle to access the same resources for developing
digital infrastructurexcvii and are relegated to residual capacity under less attractive conditions. This
is visible in the geography of Europe’s data centres: the FLAPD hubs (Frankfurt, London,
Amsterdam, Paris and Dublin) comprise over 60% of operational capacity and around half of the
new and planned capacity is mainly driven by hyperscalers pre-leasing and committed
projectsxcviii. Scale and scope advantages of large providers reinforcethe convenience of building
in existing hubs.
Faced with these market dynamics, most European providers remain smaller, specialised and
nationally fragmented, lacking the scale, scope or cross-border integration to deliver comparable
cloud services. Comparatively lower revenues reduce their ability to invest in infrastructure
expansion or innovationxcix. Google, for example, spent USD 9.6 bn on DCs across the globe in
Q3 2025 onlyc. The capital expenditure of OVHCloud for the entire financial year 2025 amounts
to EUR 361.4 m with EUR 51 m devoted to infrastructure and networksci. Taken together, these
factors (a) reinforce the structural advantage of large non-EU providers in securing key resources
at scale, (b) further constrain the ability of European providers to expand, and (c) contribute to the
current shortage and spatial concentration of computing capacity in a few established hubs,
leaving significant areas of the Union under-served in terms of local compute.
The lack of scale and scope of European cloud and AI computing service providers is also closely
linked to the second problem, i.e. the dependence on cloud and AI computing services supplied by
non-European providers.
As mentioned under section 2.1, hyperscalers leverage bundles of services to operate as “walled
gardens” meeting “all” customer needs, often within their own closed marketplacescii. The
convenience of accessing all cloud-related services from a single provider is a key driver for
customer choiceciii. AWS and Microsoft Azure each have 200+ services across their portfolios,
and Google Cloud delivers 100+ specialised offerings available at their proprietary marketplaces.
The absence of a comprehensive vendor-neutral marketplace creates an imperfect information
environment about available services and their characteristics and negatively affect the
17
comparability of services across providers and general awareness of users about the services
offered by smaller, including European, providers (discovery problem). End-users desire
simplicity and have grown accustomed to one-stop shops delivering everything from
Infrastructure-as-a-Service (IaaS) to Software-as-a-Service (SaaS) on a global scale, a level of
integrated service that hyperscalers readily supply. Customers may therefore make suboptimal
choices because they cannot easily compare offers. Imperfect information constitutes a failure in
the market for cloud and AI computing services and reduces competitive pressure on incumbents.
Hyperscalers also offer dedicated professional training37 and certification programmes38, including
through partnerships with key IT consultants and system integrators tailoring hyperscaler tools to
specific customer needsciv that re-sell or promote their servicescv. This has led to high transaction
costs of leaving the hyperscalers ecosystems and a form of skills lock-in39.
Lock-in practices and the absence of technical interoperability further bolster the hyperscalers’
market position by making it less likely for customers to switch to another provider40. In a market
characterised by network effects and high switching costscvi, the initial choice of cloud and AI
computing service provider limits future possibilities for diversification. This has posed high entry
barriers for ‘late’ entrants, notably European ones41. Large providers leverage their strong position
in the cloud market to attract customers with low entry prices, for example through cloud credits42,
but also to expand into the emerging market of AI computing servicescvii, through tying and
bundling practices - linking together new AI applications with existing cloud products - and
preferential access to computing powercviii. Their extensive service portfolio is also a competitive
advantage for attracting AI companies as customers since they rely on the advanced offering of
large providers to provide better usability.
European providers do not offer the same breadth of services. OVHcloud and Scaleway, for
example, offer around 80 servicescix whereas the US hyperscalers, such as Amazon, offer around
200. Annex 13 provides a comparative analysis of a sample of service of European vs American,
demonstrating that EU providers usually cover well core IaaS and PaaS services but also highlight
the actual gaps which tend to be more specific. These relate mainly to native AI/ML platforms,
serverless, and integrated analytics pipelines where European providers have not yet reached
hyperscaler maturity.
A smaller catalogue of services does not equate to lower quality, as argued later in this section.
Gap analyses show that European providers offer equivalent quality in their services to those
offered by the hyperscalers43 but with rather a more limited integration: To obtain access to a
similar breadth of services as those offered by a single hyperscaler, a customer must combine
37 During the COVID-19 pandemic, for example, AWS and Google Cloud offered free resources for training cloud skills and getting them certified:
See here and here. AWS offers exam vouchers for certifying one’s cloud skills to participants of EU-certified cloud courses for SME employees and jobseekers: Free EU-certified Cloud & Gen AI courses for SME employees and jobseekers — now open for enrolment! | Digital Skills & Jobs
Platform. 38 AWS for example reports 1.05m of individuals certified to their products. AWS Certification. 39 See Ofcom’s Cloud Services market study – October 2023. This structural dependence on non-EU proprietary solutions has also been cited by
several SMEs as part of their Call for Evidence contribution. 40 Competition authorities are increasingly scrutinising behaviours such as switching barriers of technical, contractual, and financial nature (including egress charges). Other relevant factors are a lack of standardisation and interoperability among services and the ability of large providers
to attract customers through free credits and volume discounts. As discussed in section 1.2 and Annex 7, the Data Act creates a right for customers
to switch and tackles relevant barriers. 41 The entry barriers for EU providers are, for example, described in the Autorité de la Concurrence’s opinion on competition in the cloud sector:
Avis 23-A-08 du 29 juin 2023. 42 Cloud credits are typically offered as short-term credits to attract new customers or motivate existing customers to adopt a new service or as long- term credits for selected customers, especially start-ups, to allow them to grow a cloud-native business in the environment of a given CSP. See:
Report Covers. Ofcom has found that AWS offers up to $100k, Microsoft offers up to $150k, and Google offers up to $100k for each year over two
years (so a total of $200k), as part of their credit programs for ‘start-ups’. See: Cloud services market study final report. 43 As discussed in this section, the simple fact of widespread uptake does not equate to a superior quality of services by non-European providers.
Instead, the breadth and integrated nature of services is a major factor of distinction.
18
solutions from multiple providerscx – each requiring different skills44, using different tooling, and
employing different operational procedures. In the absence of tools that easily enable such
combination of resources, such as cloud brokers45, customers must invest their own resources to
integrate services from different providers through custom developments.
A broader and a better integrated catalogue offers an advantage to US providers but this comes
with a caveat. The latest Eurostat data shows that the services EU enterprises actually rely on most
are email, office software, file storage, database hosting, and compute power46, which are core
IaaS workloads: virtual machines, managed databases, object storage, and nowadays also PaaS,
with managed Kubernetes (orchestration). This points to the fact that customers seldom use a
service catalogue to its full extent, something confirmed through interviews with European CIO
public and private organisations, as well as by reports that find that only few customers use a large
breadth of services from hyperscalers.47
Another factor for the low adoption of European offers is the limited visibility of said offers, as
exemplified by market reports which rarely mention European providerscxi. All this places a high
commercial barrier to greater uptake of European cloud and AI computing services.
The hyperscalers’ size also allows them to better overcome barriers for offering their services
across the EU48. Consequently, European providers often focus on highly specialised market
segments, such as the treatment of sensitive workloads requiring high-quality and secure
servicescxii. For example, the Polish provider CloudFerro supports space-related use cases such as
the ESA Civil Security programme and CODE-DE, a German Copernicus Data and Exploitation
Platform. Italy’s leading provider Aruba provides the country’s eID services. The general high
quality of services is highlighted, for example, by the recognition of OVHCloud and Scaleway in
Gartner’s list of best Strategic Cloud Platform Servicescxiii. Nevertheless, European CSPs are
typically geared towards their home market. For example, OVHcloud generates 48% of its
revenues in France compared to 29% in other European countries and 26% in the rest of the world.
Similarly, IONOS generated 56% of its revenue in 2024 in Germany, as compared to only 8% in
Spain, 5% in France, and 3% in Poland as the next biggest markets in the EU49. This phenomenon
is even more pronounced in the Italian market. 89.7% of Aruba customer base is in Italy, followed
by only 0.6% in Spain and 0.5% in France50. In 2021, the European Alliance on Industrial Data,
Edge and Cloud gathering all main European cloud providers pointed out their lack of ability to
exploit the Single Market and offer their services across borders efficiently as one of the main
barriers to scaling up and being able to compete with non-European providers51. This, paired with
the absence of effective interoperability mechanisms across services or solutions that allow for a
federation of cloud resources among CSPs, prevents European providers from achieving
economies of scale (see also section 2.2.2).
Taking the considerations above, and under the same market conditions, US hyperscalers have
been able to thrive in the EU while their local competitors have not. The hyperscalers are able to
exploit massive economies of scale, integrated ecosystems with default interoperability across
services and high capital expenditure, which allow them to offer cheaper and broader services.
44 SMEs have also pointed to workforce certifications being skewed towards non-EU vendors, thus reducing the possible adoption of European alternatives. 45 A cloud broker is a 3rd-party that adds value to cloud services on behalf of cloud service consumers. It delivers: 1) aggregation from multiple
cloud services (from possibly different providers) into a unified offering 2) arbitration, allowing switch among multiple providers dynamically based on e.g. cost 3) intermediation. Gartner, Alonso et al. “Federated Cloud Service Broker (FCSB): An Advanced Cloud Service Intermediator
for Public Administrations” . 46 See Eurostat 47 See the blog post introducing the Civo 2024 report ‘Has Cloud Computing lost its way to complexity and cost?’ 48 Diverging national cloud cybersecurity certification schemes require providers to get certified several times. 49 https://firmsworld.com/ionos-group-comprehensive-profile 50 List of Aruba Cloud Customers 51 European_CloudEdge_Technology_Investment_Roadmap_for_publication_pMdz85DSw6nqPppq8hE9S9RbB8_76223.pdf
19
Due to their global scale and capital, hyperscalers are able to spread compliance costs across the
different national markets over a large revenue base, amortizing them across scale so the average
compliance cost per unit of revenue (e.g. service) falls with scale. Conversely, European providers
remain sub-scale and confined to niche segments, focused mainly on their own national markets
and legal regimes. The services offered by EU providers, while equal in innovation and quality,
are not as integrated as the services offered by the hyperscalers nor they benefit from a large
ecosystem gravity, which prevents demand from pooling into contracts large enough to justify
more investments. Due to the sub-scale, compliance costs for EU providers remain higher, which
may slow them the release of advanced features that would allow them to meet the expectations of
customers operating in certain critical markets.
The outcome is a Single Market failure: in a genuinely unified market, all providers should be
able to scale across borders with relatively uniform rules, standards, and buying channels,
allowing efficient firms to grow and compete on equal footing. However, persistent regulatory and
contractual heterogeneity and fragmentation (see also next sections), as well as siloed demand,
mean that the EU’s market remains de facto segmented, so network effects and economies of scale
accrue disproportionately to already scaled non-EU incumbents rather than enabling EU-based
providers to reach competitive size.
2.3.2. Problem Driver 2 (PD2): Bottlenecks slowing down data centre expansion
The second driver concerns bottlenecks and regulatory fragmentation, which have contributed to
slowing the expansion of data centre deployment, reducing attractiveness of Europe for investors
and as a result contributed to limited computing capacity and persistent geographic imbalances.
Companies seeking to build data centres in the EU are faced with a fragmented policy and
regulatory environment52. For example, Germany’s Energy Efficiency Act imposes Germany-
specific obligations on data centres related to the use of renewable energy and heat recovery53. In
the Netherlands, stricter rules are applied on new facilities in the Amsterdam region as opposed to
other parts of the country54. In Poland, different permits are needed for different types of data
centres55. Bulgarian zoning regulations altogether lack an adequate category for data centres. In
the context of infrastructure buildout, the most relevant bottlenecks and frictions are found in
permitting, availability of suitable land, grid access, and access to capital.
Permitting procedures for infrastructure development – which encompass zoning and land
allocation, building permits, utilities and grid connection authorisations, and environmental
permitting – involve multiple, often inconsistent layers of national and local regulations with
multiple uncoordinated stakeholders and non-centralised processes. DCs are often not mentioned
in national or municipal planning regulations, creating uncertainty and requiring additional
rezoning processes in some jurisdictionscxiv. Permitting regimes typically ignore the strategic
dimension of DCs for the EU economy, notably their enabling role for the EU uptake of digital
solutions. Most jurisdictions require environmental review, ranging from basic assessments to full
Environmental Impact Assessments that further lengthen the permitting processcxv. In most
Member States, permitting involves repetitive requests and lengthy timelines, which can be
aggravated by community opposition and appeals56.
52 For the complete analysis on the different regulations concerning data centre deployment in 12 MS, see Annex 4 section 9. See also here pp 19- 20: 2025-Data-Center-Site-Selection-Dynamic-Brief.pdf 53 Watson Farley & Williams (2024). Data centres: An international legal and regulatory perspective—Spotlight on Germany. Available at:
https://www.wfw.com/articles/data-centres-an-international-legal-and-regulatory-perspective-spotlight-on-germany 54 Royal HaskoningDHV (2023Navigating Dutch data centre challenges and opportunities.
https://www.haskoning.com/en/newsroom/blogs/2023/navigating-dutch-data-centre-challenges-and-opportunities 55 Dudkowiak, M. (2025). Data Center Investments in Poland | Law & Development Guide 2025. Dudkowiak Kopeć & Putyra. https://www.dudkowiak.com/invest-in-poland/data-centers-investments-in-poland/ 56 See as an example, the case of Apple in Athenry in Ireland.
20
The example of Germany illustrates well how and why permitting has emerged as a bottleneck for
data centre deployment in light of the recent significant increase in demand for data centre
capacity (see section 2.1.1). Germany’s central location in Europe and its proximity to business
users have made it a major data centre hub, despite comparatively long timelines for permitting.
Early investments have resulted in strong path dependency and network effects, with new capacity
continuing to cluster around established hubs despite rising congestion costs, grid limitations, and
regulatory barriers. However, with demand for data centres growing, permitting is now a
bottleneck holding back the fast expansion of data centre capacity that would be necessary to
reduce the capacity gap. Illustrative cases in Ireland and Luxembourg show how administrative
delays caused 57data centre projects to be cancelled or delayed. In response, Member States or
regions have adopted policies to accelerate DC deployment, for example by offering transparent
planning procedures (Denmark), pre-designating areas (France), naming DC projects as of
“strategic economic interest” (Spain at regional level), enabling parallel rather than sequential
permitting (Germany), encouraging DCs to locate within existing industrial areas (Finland) or
through tax incentives for DC operators (Sweden). While these acceleration measures facilitate
short-term capacity expansion at national level, they contribute further to the regulatory
fragmentation across the Single Market.
Another factor slowing down data centre expansion is land availability, the difficulty for DC
operators to identify suitable sites. DC sites must display specific characteristics in terms of access
to utilities (energy, water) and connectivity, making suitable real estate scarce. Connectivityis a
particularly relevant pull factor for selecting appropriate sites for DC deployment58, thus
reinforcing concentration in existing hubscxvi. Land scarcity is aggravated by large cloud service
providers acquiring land and reserving energy capacity, without building DCs on the sitecxvii.
Energy availability. With respect to energy needs, different modelling approaches indicate that
DC energy consumption is foreseen to grow at an average annual rate of 13% between 2023 and
2030, twice the 2018-2023 growth ratecxviii. Energy prices in Europe are two to three times higher
than in the US and China, as shown in the table below, while they amount for 40-50% of DCs’
operational costs, creating a competitive disadvantage for the EUcxix.
Table 2. Industrial electricity prices (€/kWh – all data 2025, except China 2024)
US59 China60 EU-2761 Ireland Germany Italy Sweden Finland
€ 0.090€ 0.081 € 0.190 € 0.296 € 0.275 € 0.271 € 0.121 € 0.101
Access to energy grid constitutes an important bottleneck in the construction of DCs62. In parts of
Europe, grid constraints have resulted in moratoria on new DC connections, especially close to
established hubscxx. Connecting a DC to the grid can take between 3 and 10 years, depending on
the Member State: around 3-5 years in emerging markets (Italy or Spain) and 7-10 years in
established hubs (Frankfurt, Amsterdam, Paris or Dublin), with some projects experiencing delays
of up to 13 years due to grid congestion63cxxi. In the case of brownfield sites where there is a
57 Apple: see e.g., Guardian (2018), available (here) & Data Centre Dynamics (2022), available (here). Google: see e.g., Delano (2025), available
(here) & RTL (2024), available (here). 58 Access to connectivity was also mentioned as a key problem by most of the SMEs in the public consultation. 59 Average Price of Electricity to Ultimate Customers by End-Use Sector, available here. 60 China's Industrial Power Rates 2025: A Guide for Investors. 61 Eurostat’s non-household prices refer to the standard medium industrial band (annual consumption 500 to 1999 MWh) including all non- recoverable taxes and levies. These are widely used as an official benchmark for industrial power costs and for comparing Member States.
Available here. 62 This was also flagged by all the SMEs responding to the questionnaire as an important or highly important problem their organisation has encountered when expanding or building their infrastructure in the EU. SME also underlined that availability and affordability of (low carbon)
energy, and of a utility provider’s infrastructure nearby where the key factors that have driven their decision to select a DC’s location. 63 It is important to note that these delays are the result of grid congestion and limited availability. When it comes to grid connection, a 3 months deadline for receiving information on treatment of the connection request (i.e., the result of the permitting procedure) has been introduced by 2024
amendments to the Directive (EU) 2019/944.
21
previous power grid connection, this interval is highly reducedcxxii, but these sites can be
insufficient to meet the needs of large DC projects. To overcome grid interconnection delays, DC
operators are increasingly considering producing their own energy, with a dedicated microgrid. In
some cases, such energy production could be based on gas turbinescxxiii, which would significantly
increase their scope 1 greenhouse gas emissions64.
Another bottleneck holding data centre expansion at the necessary speed is limited access to
capital: over the past five years 58% of global DC investment occurred in the US with a record in
2023cxxiv, with Europe accounting for a smaller sharecxxv. AI computing infrastructure can be ten to
thirty-times more expensive than general-purpose DCscxxvi, requiring massive capital investment.
The EU’s fragmented financing landscape lacks the depth of an integrated capital market.
Compared to the US, the UKcxxvii and Chinacxxviii, European investors have treated DCs as a niche
market rather than a distinct assetcxxix. In addition, many parameters affect the bankability of DC
projects65 such as the size of the operator, permitting and access to grid. Combined with Europe’s
fragmented capital markets66 and the high risk associated with smaller EU CSPs, notably outside
the FLAPD markets, these challenges slow down capital mobilisation for DCs.
In parallel to these bottlenecks across EU markets, water availability can be a critical factor for
DC deployment for DCs that use cooling technologies which rely on water. With evaporative
cooling techniques, a 1-megawatt DC can use up to 25.5 m litres of water annuallycxxx, a potential
issue in water-stressed locationscxxxi. While this is not a systematic investment bottleneck in
Europe compared with other constraints67, its importance in site selection and environmental
assessments highlights the need to consider water risk as a relevant aspect of future infrastructure
planning.
In responding to this unprecedented demand growth, the factors highlighted above are holding the
EU back from swiftly building the computing capacity needed to respond to the increase in
demand, including for socially valuable infrastructure. The identified bottlenecks have historically
not weighed as heavily as today: the demand for data centres has surged with the growing role of
digital services and, more recently, with the advent of AI, which also require data centres located
closer to users. This results not only in a capacity gap but also in the reinforcement of geographic
imbalances, driven by regulatory fragmentation and the risk of regulatory arbitrage. This driver
can be considered a coordination failure among investors, energy system operators, and public
authorities, coupled with regulatory failures, as a result of fragmented practices and divergent
regulations. The first hinders effective communication of where new capacity could yield the
greatest net benefits, while individual regulations have created obstacles to the proper functioning
of the internal market, increasing compliance costs and preventing the cross-border scaling for
smaller operators. Even though larger companies face the same rules, they are able to internalise
such costs and regulatory heterogeneity given their global scale.
2.3.3. Problem Driver 3 (PD3): Limited public sector uptake of cloud and AI
computing services supplied by European providers and diverging
procurement practices
This driver is innately linked to PD1. While some analystscxxxii argue that the dependence on non-
European cloud and AI computing services stems from the limited scale and scope of these
64 Today’s DC scope 1 emissions are largely limited to diesel back-up generators which typically operate only a few days per year for testing. 65 In the public consultation, most SMEs flagged having difficulties in getting funding to develop capacities. Among the respondents to follow-up questions, several flagged access to finance and/or high interest rates when procuring next generation GPUs as a key limitation to expand capacity.
On accessing finance, SMEs frequently underlined difficulties in accessing capital, such as loans or equity. 66 EU capital markets remain fragmented along national lines. The absence of a pan-European financing framework with common legal structures, or cross-border Real Estate Investment Trusts regimes underpin the difficulties of raising capital quickly. 67 Alternative cooling technologies can be used (often at the expense of a lesser energy efficiency).
22
services (PD1), others argue that the scale and scope of these services can only grow based on
stable anchor customerscxxxiii.
As discussed in relation to PD1, the initial growth of today’s largest cloud and AI computing
service providers was fuelled by public contracts. Similar opportunities did not emerge for
European providers. Generally, public sector cloud uptake and corresponding spending on cloud
and AI computing services is relatively low. Looking at 2024 publicly available data from the
portal of Tenders Electronic Daily, approximately 3.13% of ICT awards were dedicated to cloud
and AI computing services. Out of all award notices published on the portal in 2024, 0.34% were
related to cloud and AI computing services. Budgetary constraints and a shortage of a digitally
skilled workforce to build their own private cloud constructs68 make public authorities dependent
on cloud and AI computing services from external providerscxxxiv. The systematic uptake of cloud
and AI computing services by the public sector is still a relatively recent phenomenon. For
example, Italy only set up its Polo Strategico Nazionale to provide public administrations with
access to cloud infrastructure in 2022cxxxv and Germany launched its government cloud in March
2025cxxxvi. While all Member States have developed national AI strategies and 21 of them have in
place dedicated cloud policies, Member States vary widely in terms of maturity in the uptake of
cloud services or in how they allocate and report public funding associated with these initiatives69.
In some Member States, public procurement practices impede the procurement of pay-per-use
cloud and AI computing services as they cater more towards fixed-price contractscxxxvii. Beyond
pricing modalities, the lack of harmonised approaches to procuring cloud and AI computing
services creates difficulties in developing and evaluating call for tenders70. Moreover, the
approach to using cloud and AI computing services laid down in cloud strategies and policies
varies significantly, even within the same public entity and does not always include risk
assessments71. Where they do purchase cloud and AI computing services, an increasing number of
public sector entities relies on services provided by non-European companies72, sometimes
concluding direct partnerships73. For example, Finland's State Treasury uses Oracle and Microsoft
Azure, Denmark's state public services use OpenStack's private cloud, Belgium's public authorities
use a hybrid G-cloud, which runs on the clouds by IBM, Microsoft and Oracle, and the Dutch Tax
Office uses Microsoft Azurecxxxviii. Similarly, the Flemish government recently partnered with
Microsoftcxxxix. Some public tenders directly refer to services from leading cloud and AI
computing service providers74. A 2026 study by FOTI, the Future of Technology Institute, shows
the pervasiveness of non-European providers also in the defence sector75. This results in foregone
economic opportunities for European CSPs and closes off one avenue for obtaining the stable
public contracts that have allowed US CSPs to scale (see PD1).
68 Public authorities’ ability to develop and operate cloud and AI solutions in-house is also often hampered by lower salaries in the public sector.
Regarding limited ability to attract talent, see for example Strengthening the attractiveness of the public service in France – OECD – 2023, or the struggle by Italian data protection authority to recruit AI experts. Regarding budgetary constraints, see for example Data foundations for
government: From AI ambition to action – Capgemini – 2025, where 66% public sector respondents report limited budgets for on-premise solutions
as a factor limiting widespread adoption of Generative AI. 69 OECD, 2025, Progress in Implementing the European Union Coordinated Plan on Artificial Intelligence (Volume 1) (EN) 70 SMEs responses to the Call for Evidence have highlighted how public tenders tend to be designed for non-EU incumbents with scarce
consideration for European SMEs. 71 For the case of the Central government of the Netherlands, this is highlighted in a recent report by the Netherlands Court of Audit: Dutch central
government in the cloud | Netherlands Court of Audit. Similarly, the 2022 EDPB Coordinated Enforcement Action on the Use of cloud-based
services by the public sector reports that only 1/3 of the services procured by the Member States involved was subject to the necessary Data Protection Impact Assessment.72 For example, the Netherlands’ Court of Audit has found that out of 1588 cloud services audited, more than half were procured from AWS,
Microsoft and Google. See also: Trotz Abhängigkeit und Datenschutzrisiken: Behörden gehen in die Microsoft-Cloud | Heise Online. 73 See for example the recent announcement of the State of Bavaria: Vertrag soll bis Jahresende stehen: Bayern will in die Microsoft-Cloud | heise
online. 74 See for example this tender for the provision of cloud services from the portfolio of the cloud provider AWS by an authorised cloud reseller: Bereitstellung von Cloud-Services für den Betrieb der Förderzentrale Deutschland (FZD) | BMWE. 75 Cloud Defence: an exposed European flank
23
While the described barriers to public sector cloud adoption affect public sector demand for all
service providers, they have particularly pronounced effects on comparatively smaller European
providers which do not have the resources to navigate diverging approaches to public
procurement, and that prevent them from the benefits of the Single Market (regulatory failure).
Moreover, budgetary and human resources constraints76 as well as a lack of awareness for
alternative solutions drive the public sector towards the purchasing of integrated service packages
from large incumbentscxl (see PD1). At the same time, public sector organisations voice strong
concerns related to the possible loss of operational autonomy and control over the data and
associated infrastructure77. Different national approaches have emerged to identify what
constitutes a sovereign cloud provider and leveraging public procurement, as evidenced for
example by the recent Franco-German sovereignty task force78. Similarly, there are multiple and
diverging national cybersecurity certification schemes, as described in section 2.2.1 of the report,
with some of them integrating a sovereignty dimension, demonstrating a clear fragmentation in the
internal market (systemic failure – fragmentation of the internal market). While potentially
effective in pursuing individual Member States’ policy objectives with respect to their enhanced
autonomy, different national efforts come at the cost of increased regulatory fragmentation. This
undermines the ability of comparatively smaller providers to easily navigate the European cloud
market and offer their services across borders, including to public procurers. As discussed in the
Draghi report, “multiple different national rules in public procurement generate high ongoing
costs for cloud providers. The net effect of this burden of regulation is that only larger companies
– which are often non-EU based – have the financial capacity and incentive to bear the costs of
complying79.”
This limited public sector uptake of cloud and AI computing services has spillover effects on the
modernisation of public services. The adoption of cloud and AI is also a driver of public sector
modernisation, as it enables administrations to move away from often fragmented, legacy IT
systems toward more flexible, scalable, and interoperable digital infrastructures. By leveraging
cloud and AI, public authorities can deploy new applications faster, improve service reliability,
and respond more effectively to changing policy needs or crises. They also facilitate data sharing
across departments and levels of government, supporting more integrated and user-centric public
services. In addition, they reduces the burden of maintaining on-premises infrastructure, allowing
resources to be redirected towards core missions. Overall, cloud and AI adoption underpins a shift
toward a more agile, efficient, and digitally capable public administration.
Finally, cloud and AI-savvy public authorities are increasingly embracing OSScxli, thanks to its
overall lower total cost of ownership but their benefits remain to be scaled. While OSS deployed
on EU-based data centres addresses some concerns over confidentiality of data and enables
custom-built solutions, it entails some difficulties regarding time, skills and lack of easily re-
usable public procurement award criteria (regulatory failure – administrative burden). This is
exacerbated by public authorities lacking a common approach to open source, notably when
agreeing to coordinated developments80, or to the sharing and maintenance of existing code81.
76 The 2023 Opinion 23-A-08 of the French Autorité de la Concurrence on competition in the cloud sector points to a decline in organisations’
overall in-house IT skills, due to higher reliance on managed services provided by private operators, which allow for savings but affect negatively such organisations strength of negotiation, choice and mastery of IT tools. 77 64% of public sector organizations surveyed as part of the 2025 Capgemini Data foundations for government: From AI ambition to action study
express concern about data sovereignty, 58% about cloud sovereignty, and 52% about AI sovereignty as a factor in deciding about future technology choices. 78 https://uk.diplomatie.gouv.fr/en/summit-european-digital-sovereignty-delivers-landmark-commitments 79 97e481fd-2dc3-412d-be4c-f152a8232961_en, p. 13. 80 Several Member States (IT, FR) require a comparative analysis of existing open source and commercial solutions and mandate that applications
developed for public administrations be released in open, public repositories. Others take a more voluntary approach (CZ). 81 When public authorities release OSS, the maintenance and further development often end up depending on just one or two main contributors, undermining resilience and innovation. See for instance the log4J case where there was only one active member contributing to a library largely
used to log actions in IT systems.
24
Additionally, many of the open source components widely used in today’s services and
applications are maintained by single individuals and small teams, creating critical points of
failure (see the log4j case, mostly maintained by one individual and which is considered the
second most commonly exploited vulnerabilitycxlii). Thus, while open source offers more
transparency by releasing code openly for more eyes to review and faster bug discovery, it
requires constant and expert monitoring to avoid that vulnerabilities are exploited by malicious
actors.
2.3.4. Problem Driver 4 (PD4): Absence of clarity around the concept of sovereign
cloud and AI computing services
In the absence of an agreed definition and criteria to evaluate sovereignty, users are left without
the necessary information to assess whether a service is sovereign or not. At the same time,
providers are left without a reliable opportunity to distinguish themselves commercially.
‘Sovereignty’ is often used in the ICT domain without a commonly accepted definition. Literature
defines it as“possessing the ability and competences to have reliable access to a technology it
deems critical for its own system, without any structural, uncontrollable dependency from third
countries”cxliii.Currently, non-European service providers, including US-based CSPs, are at the
forefront of offering sovereign-branded solutions for the EU market based on diverse
characteristics. For example, AWS European Sovereign Cloud guarantees data residency in
Europe with physical and logical separation from other regions and operation entirely run by EU
residentscxliv. Oracle’s EU Sovereign Cloud locates customer support, DC support, and DC
operations fully in the EU and ensures management by a dedicated EU entitycxlv. An alternative
approach takes the form of joint ventures, such as Bleu (partnership between Capgemini and
Orange offering Microsoft services)cxlvi or Clarence (joint venture between Proximus and
LuxConnect based on Google technology)cxlvii.
The price of sovereign cloud offers over traditional ones is subject to diverse points of view,
something not surprising given their recent arrival to the market. An analysis carried out by BCG82
estimates that listed prices are 10% to 30% higher compared to the public cloud. According to
BCG, “Google Sovereign Cloud is priced 10% to 20% over the public cloud, while Oracle EU
Sovereign Cloud charges a 15% to 30% price premium”, whereas “Microsoft Azure Government
carries a 15% to 25% price premium”. An empirical comparison of AWS pricing between the
sovereign cloud offers and the eu-central-1 region (Frankfurt) of 6 AWS cloud services in January
2026 using AWS’ provided calculator shows that the average price premium for the previously
mentioned sovereign services is of 15%83. Against these observations, European service providers,
possibly because sovereignty in the EU is easier for them to reach, declare in bilateral interviews
that their prices should not be affected by sovereignty requirements and could even be lower than
non-EU non-sovereign services. Real observed prices in actual competitive tenders, to which this
assessment had access in confidence, show price differences ranging from +12% to -10% for the
same level of sovereign service.
This can also be observed for “sovereign AI”: For example, Oracle advertises AI solutions
running on its sovereign cloud as “sovereign AI”cxlviii, and OpenAI announced agreements with
Germany as well as the UK Ministry of Justice, on the expansion of UK sovereign AI
capabilitiescxlix. The descriptions of these offers often also refer to high levels of cybersecurity.
However, a technically cybersecure service may still be exposed to non-EU laws requiring the
provider to grant data access to third-country authorities or to third-country policies intended to
82 See BCG Cloud Cover: Price Swings, Sovereignty Demands, and Wasted Resources 83 These are S3, FSX Windows, EC2, Lambda, RDS for PostgreSQL and DynamoDB. See: AWS European Sovereign Cloud (ESC) – Launch,
Pricing, and What’s Next
25
limit service supply. Moreover, the processing of data in the EU does not in itself prevent the
control over the software by a non-EU entity84.
The above-mentioned sovereignty claims are currently made without demonstrating the
safeguarding of operational autonomy and the protection of data against foreign interference or
access (see section 2.2.2). This imperfect information environment implies a failure in today’s
market for cloud and AI computing services leading to an information asymmetry as users do not
have a reliable means of verifying whether a service is truly sovereign or not85. At Member State
level, existing schemes on cloud cybersecurity certification sometimes contain a sovereignty
dimension. For example, the French SecNumCloud aims to ensure the ‘sovereignty of CSPs’
based on strict technical cybersecurity requirements and non-technical – sovereignty - criteria. The
adoption of SecNumCloud by cloud services remains rather low, and most European customers
are unaware of the benefits in the services provided by European CSPs, despite an increased use of
the notion of sovereignty to market their offerscl.However, in the absence of clarity around the
concept, this has not yet translated into meaningful commercial advancementscli.
Thus, the lack of such a clear common understanding and enforceable criteria for sovereignty
along with the solutions adopted by different Member States is resulting in further fragmentation
of the Single Market. Announcements such as AWS investment of EUR 7.8 bn in an EU
sovereign cloudclii indicate that today’s incumbents are likely to capture the nascent market for
sovereign cloud and AI computing services. In the absence of clarity on what constitutes a
sovereign service, the definition will de facto be set by today’s leading providers in a way that
further enshrines today’s dependence.
2.4. How likely is the problem to persist?
The problems can be expected to become increasingly acute. Despite continued investment in
DCs, the gap between supply and demand of computing capacity will likely grow. The current
timeline for DC deployment is likely to remain complex as rising demand puts additional strain on
the already lengthy permitting processes.
Member States’ national policies to attract DCs and accelerate their deployment will likely persist,
complexifying the European market for operators and creating geographical imbalances regarding
the deployment of DCs towards certain regions. Considering the importance of low latency, this
will lead to unequal opportunities for businesses across the EU, notably in central and southern
Europe where latency performance can fall short of AI or IoT solutions execution requirementscliii.
Ultimately, insufficient access to computing capacity will limit businesses' ability to integrate AI
into their operations, negatively affecting their competitivenesscliv.
While lowering operating costs will incentivise DC operators to adopt energy efficient
technologies, the industry’s demand for energy will continue to grow. Without strategic energy
planning and a focus on sustainable infrastructures, DC expansion will particularly challenge
existing DC hubs and regions with high strain on natural resources, at the risk of crowding out
electrification objectives in other sectors and generating increasing public oppositionclv. The
ongoing revision of the infrastructure planning under the TEN-E Regulation (2022/869) has a
potential to reduce the scale of the problem.
84 Delos Cloud (an SAP subsidiary operating based on Microsoft software for use by public administrations), for example, is deemed by the Interior
Ministry of the State of Baden-Württemberg to not be fully sovereign beyond its infrastructure. As the application layer remains Microsoft software, the Ministry cautions that the software and data processed remains subject to the requirements of the US CLOUD Act, giving rise to data
access without the customer’s awareness. 85 This has given rise to the term ‘sovereignty washing’, see for example: Sovereignty Washing - When 'Sovereign Cloud' Isn't Really Sovereign – VSHN AG. Similarly, SMEs respondents to the Public Consultation have stressed that foreign-controlled firms market services as “European”
despite extra-territorial dependencies.
26
Considering these factors, the expansion of DCs in the EU will continue, but at a slower pace than
needed to meet growing demands. This will continue to raise prices, a trend already noticeable for
colocation and access to GPU capacity in Europeclvi. This shortage of suitable computing capacity
may even lead EU businesses to move overseasclvii or delay the deployment of low-latency
services to the detriment of EU’s economic growth. Consequently, the provision of cloud and AI
computingservices to EU customers will continue relying on infrastructure located outside of the
EU. Given the preference of many users for keeping their data in the EU, this bottleneck will slow
down the adoption of cloud and AI computing services in the EU.
Today’s stable market share of 15% for European CSPs shows no indication of change, despite a
growing marketclviii. The largest European CSPs may be able to solidify their position as national
players, but their smaller customer base will hinder their capacity to invest, scale up and innovate.
Conversely, hyperscalers will continue to innovate and grow into the AI market, with their
solutions becoming indispensable, particularly for startups and SMEs and in MS where European
CSPs lack a commercial presence. Current trends suggest that US dominance in development and
adoption of AI technologies, with China catching up, will persistclix. European businesses and
public authorities will continue to rely on US AI providers to the detriment of European service
providers struggling to work at the frontier.
Dependence on hyperscale cloud and AI computing service providers, particularly for highly
critical use cases, will continue to expose data to third-country access and carry risks to service
continuity, endangering operational autonomy.
3. WHY SHOULD THE EU ACT?
3.1. Legal basis
Article 114 of the Treaty on the Functioning of the European Union (TFEU) empowers the EU to
adopt measures aimed at improving the functioning of the internal market through the
approximation of the provisions laid down by law, regulation or administrative action in Member
States. These measures can take the form of a Regulation or a Directive. National approaches to
expanding DC capacity risk creating a fragmented landscape on DC deployment and potentially a
regulatory race to the bottom in sustainability and permitting requirements. Diverging public
procurement practices for cloud and AI computing services and diverging sovereignty criteria may
prevent providers from fully benefitting of the internal market. If EU intervention takes the form
of a legislative proposal, it can be based on Article 114 TFEU.
Article 173(3) TFEU is the basis for enhancing the EU’s competitiveness and innovation capacity.
It enables measures to accelerate industry’s adaptation to structural changes; encourage an
environment favourable to initiatives and to the development of undertakings throughout the EU,
particularly small and medium-sized undertakings and favourable to cooperation between
undertakings; and foster better exploitation of the industrial potential of policies of innovation,
research and technological development. The lack of computing capacity in the EU negatively
affects the competitiveness of industry, keeping it from leveraging the full potential of adopting
AI, particularly those that rely on low latency. By increasing the availability of compute capacity,
this initiative aims to strengthen Europe’s competitiveness and innovation capacity. If it takes the
form of a legislative proposal, it can thus also be based on Article 173(3) TFEU.
Should the legislative proposal include elements associated with both improving the functioning
of the internal market as well as addressing the competitiveness of the Union’s industry, the
proposal would take form of a single act, building on the legal basis provided for under Articles
114 and 173(1) TFEU, to ensure a coherent approach to address, in different ways, the need for
strengthening of the Union’s cloud and AI ecosystem.
27
3.2. Subsidiarity: Necessity of EU action
Article 5(3) TFEU stipulates that action at EU level should be taken only when the envisaged
objectives cannot be sufficiently achieved by Member States alone and, due to the scale or effects
of the proposed action, can be better achieved at EU level.
The development of computing capacity in the EU currently takes place along national lines. Each
Member State operates under a distinct framework, with different processes and requirements for
DC deployment, reflecting local conditions and needs. However, as mentioned above, national
policies for DC acceleration risk further fragmentation and race-to-the-bottom with respect to
sustainability. Moreover, an increasing number of low-latency applications require close
computing capacity. More generally, the EU faces an acute shortage of computing capacity, a
problem that risks negatively affecting its competitiveness and requires EU-level action to
maintain a regulatory and investment environment that is easy to navigate for DC operators and
investors, including across borders.
Closing this capacity gap and allowing European businesses and public administrations to
leverage compute capacity while ensuring sustainability requires action at EU level. The
dependence on cloud and AI computing services supplied by non-European providers has the
same root causes across the EU and affects businesses and public administrations in all Member
States. European service providers face difficulties to scale up across the EU, for example due to
different national trustworthiness standards, particularly in public procurement. Divergent national
procurement practices complexify the market for European providers and the underlying situation
of imperfect information is a market failure requiring an EU-level response. Calls for EU-action to
address these challenges were also made in the public consultation86.
3.3. Subsidiarity: Added value of EU action
EU action would have a clear added value in addressing the problem of limited and geographically
concentrated availability of computing capacity. By providing a common approach to accelerating
DC deployment, it would enable the coherent planning and deployment of computing capacity in a
geographically balanced way, while avoiding a race to the bottom and reducing regulatory
complexity for investors and DC operators. The EU is uniquely positioned to ensure that
investment and acceleration policies reflect collective priorities and avoid fragmentation. EU-level
action would ensure that all businesses and public administrations can access sufficient compute
capacity to meet their needs and is a prerequisite for Europe to become an AI continent.
In addressing the dependence on cloud and AI computing services supplied by non-European
providers, EU action would deliver benefits that exceed what Member States could achieve
individually, especially in addressing the underlying market failures of imperfect information.
This would improve the functioning of the internal market and enable cloud and AI computing
service providers to grow beyond their national markets.
86 In replying to this topic, 80% of respondents emphasised the importance of the EU reducing its reliance on non-EU cloud and AI computing
service providers. Public authorities strongly supported coordinated EU-level action with 80% supporting EU-level actions such the establishment of cybersecurity guidelines; the adoption of standards, open specifications, and mechanisms to ensure interoperability; the creation of a mechanism
to federate cloud and AI computing services across public administration within and across MS, the creation of guidelines with standard criteria to
procure cloud and AI computing services and guidelines with standard award criteria. In addition, 72% supported the creation of clear environmental compliance requirement et EU level, and 66% were in favour of unified guidelines at EU level for energy efficiency for computing
infrastructure. Finally, public authorities called for an EU-level definition of cloud sovereignty.
28
4. OBJECTIVES: WHAT IS TO BE ACHIEVED?
4.1. General objectives
The general objective of the intervention is to ensure the functioning of the internal market for
cloud and AI computing services and to secure the conditions necessary for the Union’s
competitiveness and strategic autonomy.
4.2. Specific objectives
Specific objective 1 (SO1): Increase computing capacity deployed in the EU through
innovative and sustainable technologies. By 2030, the EU should at least triple its current DC
capacity, prioritising energy-efficient technologies in at least 80% of new installations. As demand
continues growing, this should be considered an intermediate objective so that by 2035, the
computing capacity in the EU should meet its needs.
Specific objective 2 (SO2): Ensure attractive conditions for the deployment of sustainable
and innovative computing capacity87. While SO1 is aimed at the deployment of capacity, this
SO targets the conditions for investment and deployment. By 2030, operators should be able to
obtain all permits to build and run a DC in less than 18 months throughout the EU, including
access to land, permits for energy access, and connectivity – which are also a major attention point
for investors.
Specific objective 3 (SO3): Decrease the overall reliance on non-European cloud and AI
computing services. By 2035, this intervention should increase the market share of European
cloud and AI computing service providers in the European market to 30%. Strengthening the
Union’s strategic autonomy requires reducing dependencies and ensuring that European users
have credible European alternatives to non-European incumbents. A stronger European supply
base improves the Union’s capacity to act autonomously and enhances long-term resilience,
competitiveness, and security of supply.
Specific objective 4 (SO4): Contribute to the protection of public order by enhancing the
resilience of supply of cloud and AI computing services, in particular in the public sector. By
2035, 100% of the highly critical use cases in the public sector should be operated using sovereign
cloud and AI computing services to ensure data confidentiality, operational autonomy and prevent
harms that could undermine public order. Highly critical use cases are those of a particular
systemic importance and that underpin essential functions or involve the processing of sensitive
data. Ensuring that, for them, data is protected, and service continuity is guaranteed is a key
element of attaining strategic autonomy. That is why these use cases are a priority for the move
towards services whose provision is outside of the reach of third-country policies that could result
in data access or interruptions to service continuity, i.e. sovereign services88.
Figure 5. Illustrative summary of the problems, drivers and objectives associated with this initiative
87 An example would be a data centre using immersion cooling. This technology involves submerging servers in a non-conductive liquid, which is more efficient at dissipating heat than traditional air cooling. This approach not only reduces cooling energy consumption but also allows for higher
server densities, making data centres more compact and efficient, generating more computing power while occupying less space and using fewer
resources. 88 Defining what constitutes a sovereign service is part of this initiative. However, already at this stage of the assessment, it is important to point out
that sovereignty should not be equated with ‘European’.
29
5. WHAT ARE THE AVAILABLE POLICY OPTIONS?
5.1. What is the baseline from which options are assessed?
The baseline scenario assumes that today’s policies and regulations continue. Annex 9 presents the
underlying assumptions.
For the limited availability of computing capacity in the EU (P1), the baseline scenario considers
the projection of EU DC capacity in the absence of additional intervention in the period 2025-
2036. Growth in the EU’s total installed compute capacity, measured in DC IT load, is projected
to reach around 42 GW by 2036, growing yearly by 12% over this period, yet demand for
computing capacity is expected to increase by 13% over the same period, creating a structural
capacity gap of 19 GW as seen in Figure 6 below, across all capacity growth scenarios89.
Figure 6. Evolution of the gap between data centre demand and supply
Source: Technopolis et al. (2025)clx
89 The study presents projections for data centre capacity and demand across the EU27 based on three scenarios (low, central, and high). These
projections illustrate the potential evolution of installed capacity and needs over the next decade, reflecting variations in natural resources, demand
drivers, technological innovation, and market responsiveness. The study finds a structural capacity gap across all the different growth scenarios tested for this assessment, i.e. of 12 GW in the low growth scenario, 19 GW in the central scenario (used as the baseline) and of 23 GW in the high
growth scenario. Please see Annex 4, Section 2.3 for additional information.
30
DC acceleration policies and regulatory framework remain non-harmonised across Europe.Some
Member States would develop national strategies to make specific locations more attractive for
DC operators. Outside of these possible strategies, bottlenecks such as lengthy permitting
processes would persist and capacity expansion is expected to follow existing market trends, i.e.
concentrated around existing hubs. Permitting delays, grid-connection queues and land-use
restrictions would persist, with localised alleviation through national reforms. The Commission
would continue pursuing greater transparency on the environmental performance of DC, notably
under the EEDand the upcoming review of the Taxonomy for Sustainable Finance. Industry may
set additional sustainability targets, e.g. under the Climate Neutral Data Centre Pact. Electricity
grid constraints would become a binding factor in several Member States by the early 2030s as
total DC demand surpasses 200 TWh. Under current conditions, the International Energy Agency
has estimated that around 20 % of announced projects worldwide are expected to experience
significant delay or downsizingclxi. Public investment would remain focused on high-performance
compute infrastructure for the training of large AI models (AI Factories, Gigafactories). Under
policy option 0, the EU’s compute supply increases in absolute terms but lags well behind other
regions, as shown in Figure 7 below. North America and Asia-Pacific expand faster, increasing
their share of global compute. This trend risks constraining AI model training and cloud
workloads within the EU, particularly for SMEs and public sector users.
Figure 7. Forecast of data centre capacity in the EU-27 vs ROW from 2025 to 2030
Source: Technopolis et al. (2025)clxii
For the dependence on cloud and AI computing services supplied by non-European providers
(P2), the baseline scenario considers that the market share of European service providers (15%)
will remain stable despite the opportunities that a growing market can offer. The Policy 0 scenario
reflects continued efforts under the Data Act to ensure cloud switching or parallel use of several
providers. Given the recent adoption of the Data Act, its full effects will take time to materialise.
The standard clauses recommended will allow for an easier switching of cloud services, whose
consequences will be observable as they are progressively adopted. For the interoperability
aspects, the effects are expected to take longer to materialize given the low number of existing
open specifications and harmonized standards addressing the issue, that would become of
mandatory application for providers following the mechanisms envisioned in the Act. The Data
Act empowers the Commission to further standardization requestions and the adoption of open
specifications compliant with the Data Act and that are published at a later stage. Pursuant to the
recently launched investigation, a potential designation of large CSPs as gatekeepers under the
Digital Markets Act may further open the marketclxiii. The revision of the CSA could enhance the
security and resilience of ICT supply chains including for cloud and AI computing services. The
IPCEI-CIS will continue to support the participating European providers in the development of a
multi-tenant cloud-to-edge software paradigm. The activities of the European Alliance on
Industrial Data, Edge and Cloud will continue to enhance cooperation among European providers.
31
The Commission would pursue investment in OSS, such as Simpl, or the Open Internet Stack.
While these may positively affect the uptake of cloud and AI computing services, they should be
balanced against the problems and drivers spelled out in section 2. Without a common
understanding of what a sovereign service entails, providers can promise a sovereign service to
customers without clarity on its substance. This could lead to continuous uncertainty for service
providers and users, particularly in the public sector. The absence of an EU-level cloud
cybersecurity certification scheme is already resulting in national approaches, fragmenting the
internal market. Finally, the ongoing parallel work on the Capital Markets Union strategy should
address the question of availability of private capital more structurally. Against this backdrop, the
baseline considers three scenarios of how the market share and revenues of European service
providers will evolve in the period 2025 - 2036 without additional measures:
Table 3. Baseline scenario for Problem 2
Scenario Market share of EU
providers in 2036
Cumulative revenues of
EU providers (2025 –
2036) (EUR bn)
Cumulative revenues of
non-EU providers (2025 –
2036) (EUR bn)
Baseline (pessimistic scenario)10% 438 33 212
Baseline (flat-share scenario)15% 564 31 956
Baseline (optimistic scenario)17% 611 31.483
The 15% baseline assumes European cloud providers maintain their current market share through
a combination of modest revenue growth from rising customer interest in sovereignty and
specialized use cases, offset by persistent structural barriers including vendor lock-in effects,
hyperscaler bundling strategies, and the integration challenges that prevent easy switching. The
10% pessimistic scenario reflects the resumption of the downward trend observed until 2022: as
US hyperscalers dramatically accelerate investment in cloud and AI services, creating capabilities
European providers cannot match, existing regulatory protections fail to meaningfully reduce
lock-in or enable genuine data portability, while aggressive bundled pricing and occasional
provider failures erode enterprise confidence in European alternatives. By contrast, the 17%
optimistic scenario envisions a more favourable environment where stronger existing regulations
ensure genuine interoperability between cloud platforms, reliable data portability safeguards
reduce switching friction, and clearer security standards boost customer trust, enabling European
providers to convert sovereignty conscious customers such as the public sector and expand into
segments currently dominated by hyperscalers.
5.2. Description of the policy options
This section structures the policy options by problems as identified in section 2. The options for
problem 1 combine measures of different regulatory intensity and different calibrations of EU
versus national delivery. The options for problem 2 are presented along a gradient of intensity of
the intervention. Problem 1 and 2 interplay with each other since they concern successive steps of
the same value chain, and a supply-demand relationship naturally exists: cloud and AI computing
services can only be delivered if the underlying infrastructure exists. Conversely, DC operators
only invest based on the expected uptake of their future capacity. The policy options are
nevertheless dealt with separately as they correspond to very different realities (physical world vs.
dematerialised services), contain policy measures of a different nature and mostly concern
different stakeholders. Problem 1 concerns the provision of physical infrastructure (data centres,
land availability, permitting, grid connection, connectivity, etc.) mainly involving data centre
operators (i.e. colocation providers and, CSPs when they operate their own infrastructure),
utilities, local authorities and investors. Conversely, problem 2 concerns the market for cloud and
AI computing services, which is shaped by very distinct market and competition dynamics. These
two markets are governed by distinct regulatory and economic mechanisms, which call for
32
different sets of measures. This approach has allowed to assess the impacts of measures focusing
on each of the two problems separately, i.e. infrastructure-related constraints vs market dynamics
in the delivery of services, while still recognising their interdependence.
5.2.1. Policy Options to address the limited and geographically concentrated
availability of computing capacity in the EU
Table 4. Overview of the first set of policy options, responding to Problem 1
Policy Option PO1-A Policy Option PO1-B Policy Option PO1-C
Enhancing the existing
collaborative framework
Legislative and financial intervention
enforced nationally
Legislative and financial
intervention enforced at EU-level
PM1 - Cloud Alliance
expansion
PM2 - Data Centre Forum
PM3 - Data Centre guidelines
PM4 - National facilitator
PM5 - Fast-track areas
PM6 - National funding support
PM7 - Deployment targets
PM8 - EU R&D funding
PM9 - EU deployment funding for
strategic projects
PM10 - EU-level fast-track
Figure 8. Illustration of policy options to address the limited availability of computing capacity
PO1-A could be combined with both PO1-B and PO1-C. However, PO1-B and PO-C are designed
as mutually exclusive. PO1-B packages measures that put Member States in the driving seat,
whereas PO1-C places the implementation at the EU-level. See also section 7.7.
Policy Option PO1-A: Enhancing the existing collaborative framework
This policy option strengthens existing collaborative mechanisms between Member States, EU
institutions, industry stakeholders and research bodies to support DC expansion. It addresses the
identified regulatory and coordination failures through soft measures. Participation and adherence
to produced guidelines would remain voluntary.
• Policy Measure 1 (PM1) creates a new working group for DC operators in the existing
Alliance for Industrial Data, Edge and Cloud, enabling exchange of best practices on
deployment, and regular institutionalised dialogue between DC operators, Member State
representatives and European CSPs. The working group would be chaired by a representative
of the DC industry, with the Commission providing the secretariat.
• Policy Measure 2 (PM2) establishes a broader forum for public-private stakeholders (DC
operators, TSOs, connectivity providers, equipment manufacturers, and local authorities),
enabling coordination and dialogue on DC projects and on the integration of innovative
solutions in DCs. It would be convened by the Commission and can build on existing ad hoc
DC grid integration roundtables.
• Policy Measure 3 (PM3) consists of adopting EU-level guidelines for deploying sustainable
DCs. These would go beyond best practices on energy efficiencyclxiv and would offer guidance
on deploying a sustainable DC. They would include recommendations on identifying suitable
33
land for deployment. Leveraging the forum created under PM2, the guidelines would be co-
developed by its participants and subject to periodic review.
This option is a soft way of addressing PD2 (bottlenecks slowing down data centre build-out)
While they do not directly change factors like permitting or access to energy, they intend to make
them less of a bottleneck by establishing direct links between the actors that are crucial for driving
data centre build-out and improving information flows between them (e.g. between data centre
operators and TSOs on grid capacity needs). The exchange of best practices among data centre
operators will help the industry better navigate the deployment environment in the EU and
guidelines will help industry identify, for example, the areas in which data centre deployment may
unfold more quickly due to available grid capacity. Beyond specific projects, the interactions
between relevant parties would feed into guidelines for data centre deployment – a collection of
best practices to be leveraged for faster data centre build-out across the EU. This option would
improve the connection between European CSPs (existing Alliance members) and key data centre
players (future Alliance members under this option). This would improve opportunities for
European CSPs to access data centre capacity or participate directly in the build-out.
The existing Cloud Alliance is a vibrant community of cloud companies which collaborate around
projects of common interest and are generally eager to extend membership up/downstream, i.e. to
DC operators and their equipment manufacturers. As part of the public consultation, 85% of
business respondents (n=52) supported the development of EU guidelines.
Policy Option PO1-B: Legislative and financial intervention enforced nationally
This second policy option consists of legislative and financial intervention implemented at
national level, where the EU-level intervention is limited to a coordination role.
• Policy Measure 4 (PM4) obliges Member States to designate a national facilitator for all DC
projects. Member States would be free to designate as facilitator the entity that suits the task
best within their own national structures, for example a service within their central
administration or an agency. The facilitator would accompany the applicants from start to finish
of the DC project for all authorisations relevant to DC rollout, i.e. planning and building
permits, environmental assessments, water and heat use authorisations, and grid connections.
Where needed, the facilitator would escalate issues for rapid resolution and interact with the
designated facilitator in another Member State if needed for a specific DC.
• Policy Measure 5 (PM5) on the basis of a national data centre strategy, obliges Member States
to identify suitable areas to fast-track the deployment of DCs. The identification process would
follow commonly defined EU-level criteria and ensure that the area displays the necessary
characteristics for DC deployment, e.g. connectivity and guaranteed existing or future grid
availability. Within these areas, Member States would be required to enact – based on their
choice - measures for accelerated DC deployment, beyond administrative acceleration for
environmental screenings and communicate clearly on the applicable procedures and criteria.
This could include a particular end use for the Data Centre such as being part of a broader
digital ecosystem implementing the Apply AI objectives of the Union. The toolbox established
in the new Regulation on speeding-up environmental assessments would be leveraged for
designated areas, to allow them to benefit from the additional favourable provisions in
environmental assessments. In designating the area and designing the related acceleration
measures, Member States would receive support from an EU-level coordination hub.
Established within the Commission, the hub will offer technical assistance to Member States.
The designation of acceleration zones could include an EU-level mechanism to attract smaller
users of collocation capacity. Where such status exists in the national context and beyond
environmental assessments, DCs would receive favourable status in administrative procedures
as projects of highest national significance. In identified suitable sites, Member States would be
34
bound to the target of reaching a maximum timeline of 18 months for all relevant permits
beyond environmental assessments90. Some fast-track areas would be reserved to DCs that are
of a particular added value to the EU, for examples based on their innovation or sustainability
performance (measured in line with the upcoming rating scheme under the EED) or because of
the users they will serve. Accordingly, access requirements in terms of sustainability would be
defined at EU level but would leave freedom to Member States to add criteria.
• Policy Measure 6 (PM6) consists of the possibility of Member States to voluntarily grant
public support in line with applicable State aid rules to particularly innovative and sustainable
DC projects located in the fast-track areas of PM5 (as with PM5, the sustainability performance
would be in line with the EED’s rating scheme). Public support could take the form of tax
incentives and consider rewarding DCs which contribute positively to grid flexibility and
stability.
• Policy Measure 7 (PM7) sets an EU-wide DC capacity target which Member States must
collectively reach by 2035 and an EU-level monitoring of the progress towards reaching the
target. The capacity target would be formulated in MW of DC capacity and would aim to close
the compute capacity gap while monitoring of capacity growth in underserved regions, thereby
fostering a more geographically balanced distribution of such capacity over time. The
monitoring would be integrated into the existing digital decade cycle which would result in an
annual report on progress – both at EU and Member State level.
This policy option would help overcome the identified bottlenecks currently slowing down data
centre deployment (PD2): The identification of suitable sites for DC deployment is a time-
consuming and cumbersome step usually undertaken solely by DC investors who work on the
basis of incomplete or inaccessible data (e.g. there’s no publicly accessible repository of areas
with electricity availability). The national facilitator would help data centre operators navigate
permitting requirements and help relevant authorities process requests as efficiently as possible.
Rather than altering permitting procedures and changing local requirements, this option addresses
regulatory complexity by making it more navigable for data centre operators and ensuring fastest
possible treatment within fast-track areas. The designation of fast-track areas would help data
centre operators identify suitable land. These measures would also aim to improve transparency
and coordination around zoning and grid access to facilitate more contestable access to key inputs.
Through the option for national funding, this option allows the de-risking of investments in
strategic data centre projects, helping overcome access-to-capital issues. This can prove
particularly beneficial for European providers to gain access to data centre capacity or participate
in the build-out themselves, which can boost the scale and scope of services provided by
Europeans (PD1).
Respondent views on Policy Option PO1-B. Feedback from DC operators from the public
consultation (n=30 respondents) on measures under this policy option:
Measure Overall support
One-stop-shop service to simplify infrastructure permitting procedures 83%
Expedited approval mechanisms and clear conditions for strategic or critical
projects
97%
Public-private partnerships for large-scale data centres 63%
Tax incentives for using sustainable technologies 73%
There has been noticeable support in the Call for Evidence (CfE) for measures within PO1-B.
Most companies and business associations favour simplification of permitting requirements,
90 Annex 4, section 8 presents the relevant permitting steps and their relative duration in 12 Member States. The necessary improvements to grid
availability will be driven by the Grids package (see section 5.3).
35
introducing a national facilitator for DC projects, and fast-track mechanisms. Some respondents
suggested faster grid connection and the creation of special DC deployment areas established at
national level.
Policy Option PO1-C: Legislative and financial intervention enforced at EU-level
This third policy option introduces binding acceleration measures enacted at EU-level and
supported with EU funding.
• Policy Measure 8 (PM8) identifies the key EU-level R&D funding challenges to the
development of energy- and resource-efficient and secure DC technologies, including advanced
cooling, renewable energy and storage integration, and AI-based optimisation tools91. While a
fictitious amount is set for modelling purposes, the initiative would not mention a specific
amount, thus not pre-empting the MFF discussions.
• Policy Measure 9 (PM9) creates the possibility, without prejudice to the outcome of the
negotiations on the next MFF proposal, for strategic DC deployment projects to be supported
by Union programmes, funds and financial instruments, in accordance with the objectives set
out in the regulations establishing those funds and programmes, in particular those
implementing the EU’s vision for an AI continent / Apply AI. The Cloud and AI Development
Act would specify binding EU-level criteria in accordance with which strategic projects would
be identified by the EC. As with PM8, a fictitious amount is used for modelling, while the
initiative would only be declarative.
• Policy Measure 10 (PM10) empowers the Commission to designate suitable areas for
accelerated DC deployment following EU-level criteria and in consultation with Member States
experts within an EU DC Acceleration Board. This Board would be created as a dedicated
committee (comitology). Permit granting for DC projects would follow EU-wide rules and
require approval from the Board. For these areas, Member States would be required to enact a
set of acceleration measures prescribed at EU level. The toolbox established in the new
Regulation on speeding-up environmental assessments would be leveraged for DC projects
deployed in these areas, to benefit from the additional favourable provisions in environmental
assessments.
By centralising the designation of areas and processes for accelerated deployment under an EU
expert body, this policy option would tackle the bottlenecks to DC expansion (PD2). Direct
funding for the development of sustainable DC technologies and support to DC integrating such
technologies would increase sustainable capacity. In terms of substance, this option has roughly
the same thrust as PO1-B, but with EU-level decision-making and the use of EU funds for R&D
and the deployment of strategic data centres. While open to all data centre operators, this option
will also benefit European cloud and AI service providers where they themselves seek to build
data centres or rent newly created co-location capacity. The measures thus also contribute to
addressing the lack of scale and scope of EU cloud service providers (PD1) whose ability to scale,
as described in section 2.2.1., is – among other factors – restricted by their limited infrastructure
presence and the high costs of expanding it.
In response to the CfE, several companies, business associations and a Member State expressed
support for R&D funding at national and EU level. Respondentsgenerally showed strong support
for measures that would advance the energy efficiency performance of data centres. In the
questionnaire (n=243), 70% of respondents supported funding for R&D of energy-efficient
technologies.
91 Such as AI solutions for optimising the operation of DCs.
36
5.2.2. Policy Options to address the dependence on cloud and AI computing services
supplied by non-European providers
Table 5. Overview of the second set of policy options responding to problem 2
Policy Option PO2-A Policy Option PO2-B Policy Option PO2-C
Supporting measures to increase
transparency and visibility of
sovereign cloud and AI computing
services
Voluntary framework for advancing
sovereign cloud and AI computing
services
EU-coordinated procurement and
support framework for sovereign
cloud and AI computing services
PM11 – Creating EU-level
harmonised criteria for sovereign
cloud and AI computing services
PM12 – EU guidelines on the
requirements to be fulfilled by
sovereign cloud and AI computing
services
PM13 – Annual conference on EU
Digital Sovereignty
PM14 – Interoperability flanking
measures
PM15 – Voluntary sovereign risk
assessments for the use of cloud &
AI computing services in the public
sector
PM16 – Voluntary award criteria
PM17 – Public sector cloud
federation and EU broker
PM18 – Vendor-neutral cloud and AI
training programme
PM19 - Mandatory award criteria
PM20 – Open Source use in the
public sector
PM21 – Mandatory sovereign risk
assessments for the use of cloud and
AI computing services
PM22 – Joint EU-level procurement
of cloud and AI
PM23 – SME cloud and AI support
scheme
PM24 – Cloud and AI toolbox
PO2-A packages measures which require EU action but without binding effects. PO2-A can be
implemented together with both PO2-B and PO2-C. PO2-B represents an increasing level of EU
involvement. PO2-C strengthens this further and covers binding measures. PO2-B and PO2-C are
not per se mutually exclusive but some measures in PO2-C represent a different gradient of
intervention from similar measures in PO2-B. Figure 9 illustrates how individual measures from
each PO build on each other, making them mutually exclusive. The relationship between the
options below and the identified problem drivers is discussed jointly at the end of this section.
Policy Option PO2-A: Supporting measures to increase transparency and visibility of
sovereign cloud and AI computing services
This policy option aims at establishing a common understanding of the notion of sovereignty for
cloud and AI computing services and increasing their visibility on the market. It is complemented
by a flanking measure to ensure the participation of European providers in the development of
interoperability standards.
• Policy Measure 11 (PM11) establishes a harmonised Union-level framework for sovereign
cloud and AI computing services. AI systems (which are products and not services) are not
concerned by the measure, as they are already subject to the AI Act. Acknowledging that
different use cases require varying degrees of ‘sovereignty’, the framework provides for four
levels of sovereignty assurance. Services not meeting any of the conditions would not be
classified with any sovereignty level but would obviously remain available in the market.
To be considered ‘sovereign level 1’, a service must meet the following cumulative criteria:
(i) the service provider must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure, personnel and
assets located in the Union (meaning the EEA); and
(iii) customer data, including metadata and telemetry data, is in the EU unless the
customer explicitly requires otherwise; and
(iv) the service provider demonstrates that it complies with state-of-the-art cybersecurity
standards; and
(v) if technical and operational support is outsourced to third-party providers outside of
the Union, necessary measure are put in place to ensure that would not compromise
the provider’s operational autonomy; and
37
(vi) there is full transparencey around the use of subcontractors, for which the cloud
service provider assesses that they meet Union legal obligations; and
(vii) where the cloud service provider is subject to the control of a third country or a third
country entity, it must be able to prove that the laws and government practices in that
country do not require the provider to tell that country's authorities about software
vulnerabilities before those vulnerabilities have been publicly discovered
These requirements would allow service providers with a parent company headquartered
outside of the Union to be considered as ‘sovereign level 1’.
To be considered ‘sovereign level 2’, a service must meet the following cumulative criteria:
(i) the service provider and subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure, personnel and
assets located in the Union; and
(iii) provide available personnel complying with additional personnel screening and Union
citizenship requirements, if the customer determines that imposing these additional
requirements is necessary; and
(iv) the service provider must be controlled by a legal entity in the Union. Alternatively,
if the service is controlled by a third-country legal entity, it must demonstrate that
it has in place the necessary technical, legal and organisational measures necessary to
prevent third-country governmental access and transfer of data stored in the Union, to
prevent or refuse any request from a third-country government, ensure that the control
of the third-country or third-country entity is not exercised in a manner that restricts
the provider’s ability to deliver the service, and to prevent the service disruption
and/or degradation of the service by a third-country government92; and
(v) the data generated by using the audited service shall not be re-used for the training or
fine-tuning of an AI system operated by an entity outside the EEA and in any case
are not transferred outside the Union; and
(vi) the customer data, including metadata and telemetry data, remain in the Union unless
the customer explicitly requests otherwise; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at least
at level ‘substantial’ under the European Cybersecurity Certification Scheme for
Cloud Services (EUCS)93; and
(viii) the service provider must demonstrate a high degree of control over the software
components that underpins the service. This notably implies that there exists a list of
identified dependencies related to the provision of the service, and where the software
components are provided by a third-country entity, the relevant code of the security
relevant components of the service stack can be audited, and there exists a migration
plan in the event a vendor fails or a third-country imposes restrictions; and
(ix) if the subcontractors are from a third country or a third country entity, appropriate
measures in place to demonstrate the absence of control; and
92 In the absence of a harmonised framework, non-EU service providers attempting to prevent third-country governmental access and transfer of
data stored in the Union and to prevent or refuse any request from a third-country government are using a diverse technical, legal and organisational measures. This include technical architecture with segregated physical infrastructure, ensuring that the encryption keys are not accessible to the
provider or are held exclusively by the customer, adding specific clauses in their EU employees’ contract that forbid them from taking instructions
from outside of the EU, setting up independent boards to review extra-territorial data access requests, etc. 93 As part of the ‘One Europe, one market’ roadmap agreed by the Parliament, the Council and the Commission, the co-legislator have agreed to
finalise negotiation for this initiative ed by Q4 2027. Adding one year for the measures to take effect, this implies CADA entering into force in
early 2029. EUCS technical work is finalized and has been adopted by CEN-CENELEC Technical Specifications. The candidate scheme has therefore reached an advanced stage of development, which now needs to be transformed into an Implementing Act adopted under the
Cybersecurity Act, a process much shorter than CADA’s interinstitutional negotiations.
38
(x) operational and technical support, including outsourcing, are initiated and performed
exclusively within the Union; and
(xi) where the cloud service provider is subject to the control of a third country or a third
country entity, it must be able to prove that the laws and government practices in that
country do not require the provider to tell that country's authorities about software
vulnerabilities before those vulnerabilities have been publicly discovered
These requirements would allow service providers controlled by a third-country or third-
country entity to be considered as ‘sovereign level 2’, but on the basis of some organisational
efforts. Service providers owned and controlled by a legal entity in the Union would face less
difficulties in complying with these criteria.
To be considered ‘sovereign level 3’, a service must meet the following cumulative criteria:
(i) the service provider and subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure and assets located in
the Union; and
(iii) members of the board, executive team and personnel operating the service are Union
nationals, located in the Union, and are security cleared where appropriate; and
(iv) the service provider must be owned and controlled by a Union legal entity and the
subcontractors are not subject to the control of a third country or a third-country
entity. A cloud computing service subject to the control of a third country or a legal
entity established in a third-country can still be audited against the audit criteria
where the third country has implemented specific safeguards that ensure that there is
no risk of unauthorised access to Union data or possible disruption of service quality
or continuity; and
(v) the data generated by using the audited service shall not be re-used for the training or
fine-tuning of an AI system operated by an entity outside the Union and in any case
are not transferred outside of the Union, and
(vi) the customer data, including metadata and telemetry data, remain in the Union unless
the customer explicitly requests otherwise; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at least
at level ‘substantial’ under the European Cybersecurity Certification Scheme for
Cloud Services (EUCS); and
(viii) the service provider must demonstrate a high degree of control over the software
components that underpin the service (software stack). This notably implies that there
exists a list of identified dependencies related to the provision of the service, and
where the software components are provided by a third-country entity, the relevant
code of the security relevant components of the service stack can be audited, and
there exists a migration plan in the event a vendor fails or a third-country imposes
restrictions; and
(ix) operational and technical support, including outsourcing, are initiated and performed
exclusively within the Union and by Union citizens, and by third parties that are not
subject to the control of a third country or third country entity; and
(x) where the cloud service provider is subject to the control of a third country or a third
country entity, it must be able to prove that the laws and government practices in that
country do not require the provider to tell that country's authorities about software
vulnerabilities before those vulnerabilities have been publicly discovered.
39
These requirements would not allow service providers whose parent company is
headquartered outside of the Union to be considered as ‘sovereign level 3’.
To be considered ‘sovereign level 4’, a service must meet the following cumulative criteria:
(i) the service provider, and subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure and assets located in
the Union; and
(iii) members of the board, executive team and personnel operating the service are Union
nationals, located in the Union, and are security cleared where appropriate; and
(iv) the service provider must be owned and controlled by a Union legal entity and the
subcontractors involved in the provision of the service are located in the Union,
owned and controlled by a Union legal entity; and
(v) the data generated by using the audited service shall not be re-used for the training or
fine-tuning of an AI system operated by a third-country legal entity and in any
case are not transferred outside of the Union; and
(vi) the customer data, including metadata and telemetry data, remain exclusively in the
Union; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at least
at level ‘high’ under the European Cybersecurity Certification Scheme for Cloud
Services (EUCS); and
(viii) the service provider must demonstrate effective control over the software
components that underpin the service (software stack) by demonstrating that a third
country or a third country entity does not have excessive control over the software
lifecycle. This notably implies that the relevant code of the service stack can be
audited and that effective control of the code exists by a Union legal entity; and
(ix) operational and technical support, including outsourcing, are initiated and performed
exclusively within the Union and by Union citizens, and by third parties that are not
subject to the control of a third country or third country entity; and
(x) where the cloud service provider is subject to the control of a third country or a third
country entity, it must be able to prove that the laws and government practices in that
country do not require the provider to tell that country's authorities about software
vulnerabilities before those vulnerabilities have been publicly discovered
These requirements would not allow providers headquartered outside of the Union to be
considered as ‘sovereign level 4’.
These criteria above deal with sovereignty. Cybersecurity and sovereignty are closely related and
are complementary. However, they do not focus on the same aspects nor pursue identical
objectives. The Commission, through its own tendering for EUR 180 m of sovereign cloud
services, was able to test the feasibility of such definitions established in different levels and could
verify that it can be implemented.
40
Types of providers whose services could currently reach sovereignty Levels 1-4
Level 1: US hyperscalers generally all have offerings that would allow them to qualify
under Level 1. They offer dedicated EU-based cloud services with strong cybersecurity
credentials which rely exclusively on (data) ‘regions’ located in the EU, including for
redundancy.
Level 2: Several US hyperscalers have partnered with EU companies to provide
additional sovereignty assurances – these partnerships are often called “sovereign joint
ventures”. While the implementation models differ, such partnerships generally rely on
majority EU ownership, operation by EU personnel, data centre region limited to a
particular Member State, cloud software stack provided by a hyperscaler, with security
layer controlled by the EU partner, auditable core software, with some reaching
certification under the SecNumCloud scheme. Non-European companies, outside of
joint ventures, could also qualify provided they demonstrate the absence of 3rd country
interference.
Level 3: EU providers would easily satisfy the control criteria. They would have to pass
the necessary cybersecurity controls to achieve the future EUCS certification at
assurance level Substantial or High. The other difficulty would be to demonstrate a
‘very high’ control over their software supply chain, something demonstrated in the
recently awarded contract by the Commission for sovereign cloud services.
Level 4: This level differs from level 3 in stricter cybersecurity certification and full
control over the software supply chain, something that some emerging EU offerings
propose, at least for some type of cloud services.
41
• Policy Measure 12 (PM12) foresees the adoption of guidelines on the harmonised EU-level
criteria for sovereign cloud and AI computing services defined under PM11 to help
implementation. These guidelines would be adopted by the Commission, addressing
expectations towards providers and users about what constitutes a sovereign service.
• Policy Measure 13 (PM13) addresses information asymmetries through awareness raising for
the sovereignty common understanding and other measures. In particular, an annual week-long
EU-organised conference on digital sovereignty, bringing together academic institutions,
researchers, and overall public and private stakeholders to foster collaboration in cloud and AI
innovation and the development of sovereign cloud and AI computing services. The EU-
organised awareness raising efforts would centralise the currently scattered efforts of private
sector entities to enhance the visibility of sovereign services and create a forum for providers to
connect with prospective buyers from the private and public sector.
• Policy Measure 14 (PM14) sets up measures ensuring the effectiveness of the interoperability
provisions of the Data Act: A Commission-led coordination group with Member States and
industry to drive progress on the development of interoperability standards. This would serve as
a preparatory step to the standardisation requests which the Commission can issue under the
Data Act. In addition, the measure creates a hook for possible financial support, in the context
of the next MFF (but without prejudging the outcome of such negotiations), for EU industry
participation in EU standardisation organisations to ensure that the perspective of smaller
European service providers is represented when these organisations work on a standardisation
request issued under the Data Act.
In response to the CfE, many European companies favoured a harmonised sovereignty criteria at
EU level. Public authorities showed strong support for the development of operational sovereignty
criteria. Some companies and business associations, especially located outside the EU, stressed
that such criteria should not exclude providers merely based on headquarters location.
Policy Option PO2-B: Voluntary framework for advancing sovereign cloud and AI
computing services
This option contains measures which go beyond defining and making more visible sovereign
services. They create a legislative framework for identifying such services, leaving its use
voluntary. The measures also target the broader uptake of a more diverse set cloud and AI
computing services in the public and private sector.
• Policy Measure 15 (PM15) would recommend Member States to carry at least one
sovereignty risk assessment and repeat it at least every four years but more frequent if deemed
necessary. The purpose of the sovereignty risk assessment is to identify which public sector use
cases within a Member State require the use of which sovereignty level as described under
PM11. The sovereignty risk assessment would assess, inter alia, the risks induced by the access
to such data by a third-country authority or third-country legal entity; or the risk of possible
service disruption due to dependence on a single or limited number of third-country services
providers.
On the basis of dedicated discussions conducted with 3 different public authorities representing
about 200 NIS2-Annex 1 contracting authorities, operating at regional, national and European
level (out of an estimate of 6 400 NIS2 entities across the EU), this assessment assumes that the
matching of sovereignty levels to public sector demand follows the following pattern: 70% of
use cases would require a sovereignty level 1; 20% for level 2; 9% for level 3; and 1% for level
4. Even though the scheme is novel and does not correspond to existing frameworks, this
assessment fits with broad orders of magnitude that can be inferred from existing analyses
42
conducted in several Member States that have introduced risk assessments for their public
sector clouds, such as France, Poland94 or Italy95.
This approximation is anchored in the idea that a layered and progressively more demanding
criteria is designed to tackle a progressively smaller amount of use cases. For instance, a water
supply public company might have separate IT systems to deal with non-critical use cases, but
fewer to manage critical systems.
Table 6. Illustrative distribution of criticality of IT systems within a fictitious public sector water company
Non-critical use cases
(candidate for level 1)
Critical use case
(candidate for levels 2-4)
• Procurement
• Stock management
• Inventory
• Helpdesk
• Billing system
• Accounting system
• CRM
• Workforce management:
o Payroll
o Timesheets
o Training
o Holidays
o Recruitment
o …
• …
• Water monitoring system
• Incident response system
• Early warning system
In concrete terms, France’s SecNumCloud, the closest scheme to this sovereignty framework,
was developed acknowledging that there is demand for even higher levels of sovereignty for
the most sensitive use cases, estimated at 10% of all public sector use cases (which here
correspond to sovereignty level 3 and level 4). Conversely, out of the remaining use cases (90%
of all public sector use cases), the scheme was designed to address approximately only ⅕ to ¼
of them (that is approximately 18% to 22.5% of all public sector use cases). While the
individual criteria under SecNumCloud are not the same, the compound effect of level 2
criteria under CADA allows the same type of providers (having implemented the same type of
mitigating measures) to qualify for this level. The proportion of 20% (rounded up mean value
of 18% and 22.5%) was therefore retained as a first approximation for the proportion of all
public sector use cases where conformity with level 2 would be required.
Critical use cases, defined as the use cases whose disruption would affect operational autonomy
or public order, correspond to use cases covered by level 2, 3 and 4, so 30%. The risk
assessment would have to consider the reality of the supply market to avoid unrealistic
outcomes, such as mandating the use of services that don’t exist (yet) in the market. The
measure targets critical use cases, not individual public authorities, but it is likely that Member
States would focus on the public sector entities of high criticality (as defined in NIS2 annex 1),
which amount to 6 400 throughout the EU, an amount retained in this assessment where such
data point is needed. This figure results from the amount of NIS 2 entities in Europe (160
94 See Cloud in Government Services 95 See Strategia Cloud Italia
43
00096), from which 20% are essential entities under Annex I97, from which 20% are assumed to
be public entities: 160 000 * 20% * 20% = 6 400 NIS2 Annex 1 public sector entities.
To facilitate appropriate and coherent sovereignty risk assessments, the European Commission
would develop guidelines for Member States to conduct such assessments and provide a sample
risk assessment methodology (note that these guidelines concern the conduct of risk
assessments and differ from PM12, which consist in explaining the different levels of
sovereignty). For Member States to have up-to-date information about the market conditions of
cloud and AI sovereign solutions, the Commission would also produce market monitoring
reports that will point Member States to possible gaps in the coverage of some services.
The Member State would be recommended to reflect the outcome of the risk assessment in
applicable public tenders, unless duly justified.
While PM11 only puts forward the definition of sovereignty levels, PM15 goes further by
putting forward a framework through which the respective levels of sovereignty can be
assessed. Verifying compliance against sovereignty level 1 would be based on self-assessments
conducted by the service provider. Cloud computing service providers qualifying as SMEs
would not be required to undergo the validation by the national competent authority.Verifying
compliance of service against sovereignty level 2-3-4 would be done through third party’s
auditors and verified by national competent authorities designated by Member States.. The
competent authorities will then verify the audit report, opinion and evidence and will provide
the decision that would allow the service provider to participate in procurement procedures
across the Union that are limited to the respective assurance level. Other Member States shall
be notified of the draft decision, to which they can object within 60 days. If continued
objections occur, the Commission will adopt a binding decision.
The competent authorities should also register the audit approval in a Union repository,
maintained by the Commission. The repository of sovereign cloud and AI computing services
will be a public list of audited sovereign cloud and AI computing services that verifiably
comply with the sovereignty requirements. The benefits are for providers and users alike:
providers will enhance their visibility and users their market research.
To cater for market evolutions, the sovereignty criteria of all levels and evaluation
methodology would be modifiable by comitology. This evaluation methodology would help
third party auditors in their assessment of the service and ensure full harmonisation in the way
different auditors conduct their assessment and for Member States to ensure that the procedures
have been followed.
The outcome of the audit would be valid across the Union, with a recourse possibility for
providers and Member States. Other Member States and the European Commission shall also
have the possibility to request clarifications about certifications granted by other MS. The
assessment would be renewable following the same evaluation methodology.
This measure is primarily designed to contribute to the protection of public order by enhancing
the resilience in the public sector, which is SO4. Nevertheless, European providers would face
less costs and efforts to meet sovereignty conditions. When it comes to meeting the criteria to
96 See SWD published as part of the proposal for the Digital Omnibus 97 This is an extrapolation to the EU of France's NIS2 available data
44
demonstrate sovereignty level 1 and 2, EU providers can more readily substantiate that they are
not affected by third-country policies affecting data access or limiting service continuity. As
well, level 3 and level 4 sovereignty can only be served by service providers owned and
controlled by EU entities. This implies that PM15 will also contribute to decreasing the overall
reliance on non-European cloud and AI computing services, which is SO3.
• Policy Measure 16 (PM16), with a view to support the EU added value provided by public
procurement of cloud and AI computing services, this measure establishes a set of voluntary
non-price award criteria for the public procurement of cloud and AI computing services. The
criteria should be of ancillary and non-decisive value in view of the overall tender and are to be
included, voluntarily, by contracting authorities in the procurement of cloud and AI computing
services that are not off the shelf, standardised or commercially available services. These
criteria will earn additional points to tenderers that demonstrate:
(i) That the tenderer contributes to reinforcing the digital technology supply chain in the
Union, including the use of software or hardware designed or manufactured in the
Union
(ii) That the tenderer has integrated Union technologies, including the uptake of research
and development results stemming from EU-funded research and development
programs;
(iii) That the innovation required to deliver the service being procured is conducted in
the Union or in a third country that contributes to the development of a European cloud
and AI ecosystem;
(iv) That the service is delivered, to the greatest extent feasible with regard to market
availability or technical requirements, through critical [computing, storage and
networking] hardware components designed in the Union or manufactured in the
Union, or both, or, where this is not feasible, through hardware components from a
country or countries that contribute to strengthening security of supply and developing
a European cloud and AI ecosystem;
• Policy Measure 17 (PM17) establishes a voluntary public sector cloud federation of EU
institutions, bodies and agencies, Member States and other public sector bodies. Members of
the federation would be allowed to share the DC services, cloud and AI computing services
which they own and control. This measure requires the Commission to create a platform
providing a brokering service. The Commission and the Member States would adopt the
required non-functional service level agreements (e.g. uptime) and the harmonised technical
specifications.
• Policy Measure 18 (PM18) creates a vendor-neutral cloud and AI training programme to
upskill workers in the single market, particularly addressing the interconnection between these
services and their underlying technologies. This implies developing curricula and certification
mechanism and funding the institutions delivering the training. The initiative would be
declarative, without financial commitments.
Among public authorities that replied to the public consultations, 77% support mechanisms to
allow the federation of cloud and AI computing services between public administrations within
and across Member States. In terms of EU-level action, 53% of total respondents support the
inclusion of a criterion ensuring sovereignty, autonomy, resilience and availability to help public
administrations in their procurement process. This measure was supported by 77% of public
authorities. 85% were in favour of guidelines with standard criteria for procuring cloud and AI
computing services. Several respondentsexpressed support for skills training programs in both the
public and private sectors.
45
Policy Option PO2-C: EU-coordinated procurement and support framework for sovereign
cloud and AI computing services
This option leverages PO2-B (and by extension one element from PO2-A, see Figure 9) to foster
the uptake of cloud and AI computing services in the public and private sector, with a higher level
of EU intervention and mandatory requirements for the public sector.
• Policy Measure 19 (PM19): First, with a view to reducing critical dependencies, this measure
makes the use of the non-price award criteria under PM16 mandatory. When procuring cloud
and AI computing services for which the EU has been assessed as dependent, contracting
authorities would be obliged to apply the award criteria established under PM16.
Second, Member States would have to draft an action plan to explain how they intend to use
public procurement, notably through pre-commercial procurement and procurement of
innovation instruments, to increase the uptake of highly innovative cloud and AI computing
services from small-cap companies, notably start-ups and scale-ups. As a follow-up, the uptake
of such services, would be monitored through a centralised, Commission-led monitoring
exercise. This would avoid the need for separate reporting obligations and create synergies with
other measures within this Policy Option that are aimed at assessing the market presence of EU
providers (cf. PM21). PM19 will contribute to decreasing the overall reliance on non-European
cloud and AI computing services, which is SO3.
• Policy Measure 20 (PM20) aims to encourage the share and, reuse of open source assets by
the public sector. Under this measure, public authorities are encouraged to promote the use of
open standards and available open source solutions where these demonstrate comparative merit
to proprietary solutions and reuse digital assets, including code, configurations, and
documentation, developed by other public administrations rather than commissioning similar
solutions from scratch, thereby eliminating duplication of efforts and redundant expenditure on
functionally similar solutions; and to share software into a common repository, ensuring that
investments made by public administrations in the EU contributes to a pool of collectively
maintained sovereign solutions.
Therefore, first, this policy measure requires Member States to take the necessary measures to
encourage the use of open standards and open source software by public sector organisations
taking into consideration aspects such as functionalities, total cost, user-centricity,
cybersecurity or other relevant justified objective criteria.
Second, public organisations will have to consider making the software for which they hold
rights publicly available on an open source repository connected to the EU OSS catalogue
maintained by the Commission. The released software should put in place the appropriate
safeguards to protect security or intellectual property rights. To ensure the quality and security
of publicly released code, open source repositories will have clear governance and operational
mechanisms, such as quality assurance (e.g. DevOps) and the continuous assessment of the
cybersecurity posture of the components (e.g. static and dynamic security testing)
complemented by a Software Bill of Materials. The release of software as open source shall not
be made in the case in which the cost of doing it would be disproportionate, there is a risk to
the security of information systems of the Union or there is public order restriction.
Third, this measure includes the creation and maintenance of Open Source Programme Offices
(OSPOs) within the public sector to support the implementation and management of open
source software. To facilitate the cooperation among these OSPOs, the Commission will
establish a Network of OSPOs.
• Policy Measure 21 (PM21) consists of two elements. First, it turns what is only a
Recommendation under policy measure 15 into a binding requirement for Member States.
46
Member States shall carry at least one sovereignty risk assessment and repeat it at least every
four years but more frequent if deemed necessary. The purpose of the sovereignty risk
assessment is to identify which public sector use cases within a Member State require the use of
which sovereignty level as described under PM11. The sovereignty risk assessment would
assess, inter alia, the risks induced by the access to such data by a third-country authority or
third-country legal entity; or the risk of possible service disruption due to dependence on a
single or limited number of third-country services providers. On the basis of dedicated
discussions conducted with 3 different public authorities representing about 200 contracting
authorities, this assessment is based on the finding that the matching of sovereignty levels to
public sector demand follows the following pattern: 70% of use cases would require a
sovereignty level 1; 20% for level 2; 9% for level 3; and 1% for level 4. Even though the
scheme is novel and does not correspond to existing frameworks, this assessment fits with
broad orders of magnitude that can be inferred from existing analyses conducted in several
Member States that have introduced risk assessments for their public sector clouds, such as
France, Poland98 or Italy99. This approximation is anchored in the idea that a layered and
progressively more demanding criteria is designed to tackle a progressively smaller amount of
use cases. For instance, a water supply public company might have separate IT systems to deal
with non-critical use cases (e.g. procurement, stock management, inventory, helpdesk, billing
system, workforce management, fleet management, etc), but fewer to manage critical systems
(e.g. water monitoring system). For instance, France’s SecNumCloud, the closest scheme to
this sovereignty framework was developed acknowledging that there’s demand for even higher
levels of sovereignty, estimated at 10% (which here correspond to level 3 and level 4), and that
for the remaining 90%, SecNumCloud was designed to cover 1/5 to 1/4 of the use cases, so
somewhere in between 90% * 1/5 = 18% and 90% * 1/4 = 22.5%. While the criteria for
SecNumCloud and level 2 differ, this figure is nevertheless retained as a first approximation.
Critical use cases, defined as the use cases whose disruption would affect operational autonomy
or public order, correspond to use cases covered by level 2, 3 and 4. The risk assessment would
have to consider the reality of the supply market to avoid unrealistic outcomes, such as
mandating the use of services that don’t exist (yet) in the market.
The measure targets critical use cases, not individual public authorities, but it is likely that
Member States would focus on the public sector entities of high criticality (as defined in NIS2
annex 1), which amount to 6 400 throughout the EU, an amount retained in this assessment
where such data point is needed. This figure results from the amount of NIS 2 entities in
Europe (160 000100), from which 20% are essential entities under Annex I101, from which 20%
are assumed to be public entities: 160 000 * 20% * 20% = 6 400 NIS2 Annex 1 public sector
entities.
To facilitate appropriate and coherent sovereignty risk assessments, the European Commission
would develop guidelines for Member States to conduct such assessments and provide a sample
risk assessment methodology (note that these guidelines concern the conduct of risk
assessments and differ from PM12, which consist in explaining the different levels of
sovereignty). For Member States to have up-to-date information about the market conditions of
cloud and AI sovereign solutions, the Commission would also produce market monitoring
reports that will point Member States to possible gaps in the coverage of some services.
98 See Cloud in Government Services 99 See Strategia Cloud Italia 100 See SWD published as part of the proposal for the Digital Omnibus 101 This is an extrapolation to the EU of France's NIS2 available data
47
The Member States would have to determine which public authorities are required to procure
specific levels of sovereignty and make this mandatory at national level, ensuring that
procurement aligns with the risk assessment, unless duly justified.
While PM11 only puts forward the definition of sovereignty levels, PM15 goes further by
putting forward a framework through which the respective levels of sovereignty can be
assessed. Cloud computing service providers qualifying as SMEs would not be required to
undergo the validation by the national competent authority. Verifying compliance against
sovereignty level 1 would be based on self-assessments conducted by the service provider
itself, while assessing the compliance of the service against sovereignty levels 2-3-4 would be
performed through independent third party’s auditors and verified by national competent
authorities designated by the Member States. The competent authorities will then verify the
evidence provided by the service providers, the audit report and opinion and will provide a
decision: acceptance, rejection or request for further information. If positive, the competent
authority of the establishment of the service provider shall notify other MS for objections,
which the evaluating competent authority will have to take into consideration for the final
decision. This decision will allow the service provider to participate in procurement procedures
across the Union. In the case of continued objections, the Commission will assess it and adopt a
binding decision to settle the dispute. If the evaluating competent authority determines that
there is not enough information to take a decision, it shall request the provider for additional
information. Finally, in the case of a rejection, the cloud service provider will have the
possibility to recourse, which the evaluating competent authority will have to take into account
for the final decision.
Competent authorities should also register the audit approval in a Union repository, maintained
by the Commission, as well as on the Digital Wallet of the service provider. The repository of
sovereign cloud and AI computing services will be a public list of audited sovereign cloud and
AI computing services that verifiably comply with the sovereignty requirements. The benefits
are for providers and users alike: providers will enhance their visibility and users their market
research.
To cater for market evolutions, the sovereignty criteria of all levels and evaluation
methodology would be modifiable by comitology. This evaluation methodology would help
third party auditors in their assessment of the service and ensure full harmonisation in the way
different auditors conduct their assessment and for Member States to ensure that the procedures
have been followed.
This measure is primarily designed to contribute to the protection of public order by enhancing
the resilience in the public sector, which is SO4. Nevertheless, European providers would face
less costs and efforts to meet sovereignty conditions. When it comes to meeting the criteria to
demonstrate sovereignty level 1 and 2, EU providers can more readily substantiate that they are
not affected by third-country policies affecting data access or limiting service continuity. As
well, level 3 and level 4 sovereignty can only be served by service providers owned and
controlled by EU entities. This implies that PM21 will also contribute to decreasing the overall
reliance on non-European cloud and AI computing services, which is SO3.
Second, private sector essential entities listed under Annex I of NIS 2 are encouraged to
integrate into the cybersecurity risk assessment they already conduct, the assessment of the
48
risks stemming from their use of cloud and AI computing services. This includes an analysis of
the laws applicable to the computing service and the extraterritorial reach of such laws; the
risks associated to the possible unauthorised third country government access to and transfer of
data; the risks associated to service continuity and quality; the operational dependency and
possible loss of autonomy.
• Policy Measure 22 (PM22) establishes a voluntary framework for EU-level public
procurement of cloud and AI computing services for EU institutions, bodies and agencies, and
Member States public sector organisations. The intention of the joint procurement framework is
to enable participants to procure with a common voice. The joint procurement framework
would complement Member States’ existing procurement practices. This measure is about
creating a framework enabling joint procurement, but the participating contracting authorities
are left with the power to decide the details of what they ultimately want to procure. This
measure builds on the common platform under PM17. It will allow aggregation of demand for
innovative solutions, allowing the EU to do what the US has done with a lot of success in the
past, i.e. award large contracts which support early innovation and uptake (see PD1) and
mitigate the lack of stable contracts holding European providers back from scaling up (see
PD3).
• Policy Measure 23 (PM23) creates a support scheme for SMEs to support through consultancy
the adoption of cloud and AI computing services by SMEs, including European CSPs. While
technically independent, this measure presents synergies with PM18. As with other measures
that imply funding, this does not pre-empt the next MFF.
• Policy Measure 24 (PM24) would imply that the Commission sets up an online toolbox for the
private sector service providers to easily identify software tools, in particular open source ones,
with the specific aim to help them in complementing their service offering. This toolbox would
create visibility for lesser known tools and serve as a final repository for the outcome of
projects funded by European programmes.
Overall, all thecategories of respondents showed strong support for measures that would promote
open source, including to achieve sovereignty. Most EU citizens, public authorities and academic
institutions would favour an obligation to release the code developed with public money in open
source repositories, as well as a common EU open source licensing schema. Public authorities
showed strong support for a marketplace for cloud and AI computing services. In terms of EU-
level action to prevent unlawful access to data, several public authorities favoured common
criteria to help them during the procurement process to identify highly critical use cases. Several
companies also supported this. Over 60 position papers received through the CfE stressed the
importance of public procurement, including joint procurement and the strategic role the public
sector could play as an anchor client.
Figure 9. Illustrations of Policy Options to address the dependence on cloud and AI computing services
supplied by non-European providers
49
In relation to the identified problem drivers, the above-mentioned measures are expected to
intervene as follows: PM11 & PM12 would allow users of cloud and AI computing services to
reliably identify the level of sovereignty of a service. PM15 and PM21 add a sovereignty risk
assessment for the public sector to identify its sovereignty needs. These measures directly address
PD4 (absence of clarity around the concept of sovereign cloud and AI computing services), with
growing degrees of intensity. As European providers of cloud and AI computing services can be
considered to be in a good starting position for meeting the benchmark of sovereignty, and
because of the public sector’s particular interest in relying on sovereign services, these measures
also address PD3 (limited public sector uptake of cloud and AI computing services supplied by
European providers) – with PM19 and PM21 being the most direct. PD3 refers in particular to the
lack of stable contracts for European CSPs, which would be mitigated through PM15 and, more
directly, through PM21 and PM19. PM13, PM18, PM23 and PM24 enhance the visibility and
usability of services provided by a more diverse set of cloud and AI computing service providers,
including European ones. They make it more likely for users, including the public sector, to
choose these solutions (meaning a contribution to tackling PD3) and contribute to addressing the
factors summarised under PD1 (lack of scale and scope of EU cloud service providers). PM14
contributes to interoperability, thus addressing PD1 (enhanced interoperability is a way of arriving
at more integrated service offerings and lower the threshold for users to combine multiple
offerings from European providers). PM16 and PM19, allows for the recognition of an EU-added
value in public procurement and PM15 and PM21 will shed light on sovereignty vulnerabilities in
the public sector, likely to result in a growing turn towards sovereign, including European services
and thus addressing PD1 (more public contracts can be expected to help European providers grow
the scale and scope of their service offer). PM17 and PM22 will boost public sector uptake,
including to the benefit of European providers. They can thus be seen as addressing PD3. PM22
would allow aggregation of demand for innovative solutions, allowing the EU to do what the US
has done with a lot of success in the past, i.e. award large contracts which support early innovation
and uptake (see PD1). PM20 boosts the public sector’s use of/contribution to open source, creating
opportunities for smaller players, because open source allows for a faster development of new
solutions and allows new ideas to challenge existing market silos. It thus contributes to both PD3
and PD1. PM24 addresses the information imbalance currently affecting the cloud market by
allowing for better visibility of smaller providers and comparability across provider ecosystems.
50
5.2.3. Relationship between policy options and policy measures
Table 7. Overview of the Policy Options and links with the Problems, Problem Drivers, and Specific Objectives
Problem
driver(s) Problem Policy
Option Policy Measures
Specific
objectives*
PD1 PD2 SO1 SO2
✓
P1: Limited
and
geographically
concentrated
availability of
computing
capacity in the
EU
1A:
Enhancing
the existing
collaborative
framework
PM1: Expanding the Alliance for Industrial Data, Edge and Cloud with a workgroup on
data centres and extending membership to relevant players ✓
✓✓ PM2: Creating a forum for exchanges between relevant public and private stakeholders
involved in the buildout of data centres ✓
✓✓ PM3: Adopting guidelines on building sustainable data centres in the EU✓ ✓
✓✓ 1B:
Legislative
and financial
intervention
enforced
nationally
PM4: National facilitator for data centre projects ✓
✓✓ PM5: Mechanism for Member States to identify areas for fast-track data centre
deployment✓ ✓
✓✓ PM6: Possibility forpublic support for data centres✓ ✓
✓ PM7: Set deployment targets and monitor progress✓
✓✓ 1C:
Legislative
and financial
intervention
enforced at
EU level
PM8: EU R&D funding✓
✓✓ PM9: EU deployment funding for strategic projects✓ ✓
✓✓ PM10: EU-levelidentification of areas for fast-track data centre deployment✓ ✓
* SO1: Increase computing capacity deployed in the EU through innovative and sustainable technologies
SO2: Ensure attractive conditions for the deployment of sustainable and innovative computing capacity
PD1: Lack of scale and scope of European cloud and AI computing service providers
PD2: Bottlenecks slowing down data centre expansion
51
Problem
driver(s) Problem Policy
Option Policy Measures
Specific
objectives*
PD1 PD3 PD4 SO3 SO4
✓
P2:
Dependence on
cloud and AI
computing
services
supplied by
non-European
providers
2A: measures
to increase
transparency
and visibility
of sovereign
cloud and AI
computing
services
PM11: Creating an EU-level harmonized criteria for sovereign cloud and
AI computing services✓ ✓
✓ PM12: EU guidelines on the requirements to be fulfilled by sovereign
cloud and AI computing service providers ✓ ✓
✓✓ PM13: Awareness raising on EU digital sovereignty ✓ ✓
✓ PM14: Measures for effective interoperability of cloud and AI computing
services✓
✓✓✓ 2B: Voluntary
framework
for advancing
sovereign
cloud and AI
computing
services
PM15: Voluntary sovereign risk assessments for the use of cloud and AI
computing services in the public sector✓ ✓
✓✓ PM16: Voluntary award criteria rewarding a European added value✓ ✓
✓ PM17: Public sector cloud federation and EU broker ✓
✓✓ PM18: Creating vendor-neutral cloud and AI training✓
✓✓ 2C: EU-
coordinated
procurement
and support
framework
for sovereign
cloud and AI
computing
services
PM19: Mandatory specific award criteria for the procurement of cloud
and AI computing services✓
✓✓ PM20: Boosting open source use in and by public administrations✓ ✓
✓✓✓ PM21: Mandatory sovereignty risk assessment for the procurement of cloud
and AI computing services ✓ ✓
✓✓ PM22: Joint EU-level public procurement of cloud and AI computing
services✓ ✓
✓ PM23: SME cloud and AI support scheme✓
✓ PM24: Cloud and AI toolbox✓
*SO3: Decrease the overall reliance on non-European cloud and AI computing service providers
SO4: Contribute to the protection of public order by enhancing the resilience of supply of cloud and AI computing services, in particular in the public sector
PD1: Lack of scale and scope of European cloud and AI computing service providers
PD3: Limited public sector uptake of cloud and AI computing services supplied by European providers
52
PD4: Absence of clarity around the concept of sovereign cloud and AI computing services
5.2.4. Relationship between Problem Drivers and Policy Measures
Table 8. Aspects of the Problem Drivers addressed by the Policy Measures
PD1: Lack of scale and scope
of EU CSPs
PD2: Bottlenecks slowing
down data centre expansion
PD3: Limited public sector
uptake of cloud and AI
computing services supplied
by European providers and
diverging procurement
practices
PD4: Absence of clarity
around the concept of
sovereign cloud and AI
computing services
PM1 - Cloud Alliance
expansion
Improved networking between
colocation operators and
European CSPs
PM2 - Data Centre Forum Identification of suitable land;
improved administrative
awareness for individual DC
projects; improved awareness
for electricity grid needs
PM3 - Data Centre
guidelines
Easier access to best practices
by smaller scale operators
Identification of suitable land;
best practices on fast
deployment
PM4 - National facilitator Being accompanied through
administrative processes is of
particular benefit to smaller,
including European, CSPs
Acceleration of administrative
processes; making regulatory
environment easier to navigate
PM5 - Fast-track areas Smaller scale operators stand
to benefit proportionally more
from accelerated
administrative processes
Identification of suitable land;
acceleration of administrative
processes; making regulatory
environment easier to navigate
PM6 - National funding
support
Lowering the infrastructure
entry barrier
Access to capital
PM7 - Deployment targets Recognition of strategic
importance
PM8 - EU R&D funding Lowering the
infrastructure/innovation entry
barrier
Access to capital
53
PD1: Lack of scale and scope
of EU CSPs
PD2: Bottlenecks slowing
down data centre expansion
PD3: Limited public sector
uptake of cloud and AI
computing services supplied
by European providers and
diverging procurement
practices
PD4: Absence of clarity
around the concept of
sovereign cloud and AI
computing services
PM9 - EU deployment
funding for strategic projects
Lowering the infrastructure
entry barrier
Access to capital
PM10 - EU-level fast-track Smaller scale operators stand
to benefit proportionally more
from accelerated
administrative processes
Identification of suitable land;
acceleration of administrative
processes; making regulatory
environment easier to navigate
PM11 - EU-level harmonized
criteria for sovereign cloud
and AI computing services
Improved identification of
sovereign services
PM12 – EU guidelines on
requirements to be fulfilled
by sovereign cloud and AI
computing services
Improved identification of
sovereign services
PM13 – Awareness raising
on EU Digital Sovereignty
Enhanced visibility and drive
uptake of sovereign, including
European, CSPs
Improved identification and
visibility of sovereign services
PM14 – Interoperability
flanking measures
Improved opportunities for
building integrated offers
PM15 – Voluntary sovereign
risk assessments for the use
of cloud and AI computing
services in the public sector
Improved commercial
opportunities for providers of
sovereign services, including
Europeans
Recommended uptake of
sovereign services, including
those provided by European
CSPs, for highly critical use
cases
Improved identification and
visibility of sovereign services
PM16 – Voluntary award
criteria
Improved commercial
opportunities for providers
from the EU (and countries
with which the EU has no
dependency)
Facilitated access to EU
providers (and from countries
with which the EU has no
dependency)
PM17 – Public sector cloud Boosts uptake
54
PD1: Lack of scale and scope
of EU CSPs
PD2: Bottlenecks slowing
down data centre expansion
PD3: Limited public sector
uptake of cloud and AI
computing services supplied
by European providers and
diverging procurement
practices
PD4: Absence of clarity
around the concept of
sovereign cloud and AI
computing services
federation and EU broker
PM18 – Vendor-neutral
cloud and AI training
programme
Increased likelihood of uptake
of services provided by smaller
CSPs
Upskilled workforce increases
likelihood of switching
PM19 – Mandatory award
criteria
Improved commercial
opportunities for providers
from the EU and countries
with which the EU has no
dependency
Facilitated access to EU
providers (and from countries
with which the EU has no
dependency)
PM20 – Open Source use in
the public sector
Creates opportunities for
smaller providers
Boosts uptake
PM21 – Mandatory
sovereign risk assessments
for the use of cloud and AI
computing services
Improved commercial
opportunities for providers of
sovereign services, including
Europeans
Mandatory uptake of sovereign
services, including those
provided by European CSPs
Improved identification and
visibility of sovereign services
PM22 – Joint EU-level
procurement of cloud and AI
Enhanced possibilities for
large contracts supporting
early innovation and uptake
Boosts uptake
PM23 – SME cloud and AI
support scheme
Enhanced commercial
opportunities for SME
providers, including
Europeans
PM24 – Cloud and AI
toolbox
Enhanced visibility and
comparability, including for
European providers
55
5.3. Options discarded at an early stage
• Mandating the digitalisation of permitting processes for DCs throughout the EU. While the
Inter-institutional Better Regulation encourages to consider the digitalisation of processes
when defining policies, digitalising the permitting process is intuitively perceived as
something better tackled in a horizontal measure covering the whole spectrum of industrial
installations rather than through a sectoral basis.
• Introducing measures on tackling the lack of grid capacity: As the issue with the lack of
available grids capacity in certain locations is not unique to DCs, and grid access is regulated
by energy legislation, option introducing sector specific measures to tackle lack of grid
capacity was discarded. Ensuring grids will be in place and ready to uptake future loads is
dealt by on a horizontal level under the European Grids package.
• Options to directly improve the availability of private capital have been discarded. First,
because the policy measures under PO1-A/B/C, by simplifying and accelerating the
deployment of DCs, are already conducive to an improved investment landscape.
• Creation of a national central and public repository where Member States can include all
relevant information on areas designated for DC deployment. This option is discarded as it
would favour real estate speculation and overall bring little benefit for DC investors.
• Common Procurement Vocabulary (CPV) i.e. standardised tender vocabulary at EU-level for
procuring cloud and AI computing services for a more precise monitoring of public sector
capacity and demand. The existing CPV codes do not include a specific entry for cloud and AI
computing services and are instead classified as ICT services or ICT infrastructure. This
option is discarded as adding such a code would require a full re-think of the codes related to
the procurement of ICT in general, something that is beyond the scope of this intervention.
• The set up of an agency to implement several of the PMs laid down in this assessment, such as
the operating the coordination hub from PM5, operating the broker from PM17, conducting
the procurements from PM22 or running the marketplace from PM24. This option was
discarded at an early stage as perceived as institutionally difficult to deliver.
• In the context of PM15/21, reverse the logic of the sovereignty framework where it would fall
on contracting authorities to trigger sovereignty audits in the context of public tenders and
only if a similar service is not already audited (as currently the case under FedRAMP). This
option has been discarded as it would limit the number of services under levels 2, 3 or 4, at
least initially, and create an artificial barrier for providers to have their services assessed
against the sovereignty framework. It also delays procurement procedures where a new service
needs to go through the whole evaluation process.
5.4. Possible combination of options
As discussed above, the first set of PMs are grouped by their regulatory nature and governance
level, i.e. non-regulatory vs regulatory and national vs EU action. PO1-A focuses only on soft,
non-regulatory instruments, whereas PO1-B and PO1-C are of a regulatory nature, enforced at
national and EU level respectively. As such, some combinations of measures are mutually
exclusive, e.g. PM4 and PM5 (national regulation under PO1-B) cannot coexist with PM10 (EU-
level regulation under PO1-C) as they address at different levels the bottlenecks to expand DC
capacity in the EU. However, other measures under PO1-B and PO1-C can be complementary, for
example potential public support for DCs (PM6) could align with EU-level instruments, such as
those under PM8 and PM9, creating a multi-level financing framework to addresses a critical
dependency from the EU. Similarly, all or selected elements of PO1-A (e.g. guidelines, forum)
could be combined with either PO1-B or PO1-C to enhance regulatory action with soft support.
56
The deployment targets to monitor the expansion of capacity in the EU under PM7 could also be
adopted independently of PO1-B, either in combination with PO1-A, PO1-C or in the context of
the forthcoming review of the Digital Decade Policy Programme.
By contrast, PO2-A/B/C follow a deliberately incremental approach to intervention along
sovereignty and federation of public resources (see section 5 and annex 4 for details). The related
PMs are not designed to be combined with one another, but rather reflect the degree of
intervention, with the higher levels of intervention building on the mechanisms established in the
preceding one. As best visualised in the graph at the end of section 5.2.2, PM15 builds on PM11;
in turn, PM21 builds on PM15. PM19 builds on PM16. Similarly, PM22 builds on PM17. Finally,
PM23 has clear synergies with PM18. The remaining PMs, i.e. PM12, PM13 and PM14 under
PO2-A are stand-alone measures that could be retained independently
6. WHAT ARE THE IMPACTS OF THE POLICY OPTIONS?
This section summarises the main expected economic, social and environmental impacts of each
Policy Option (PO) compared to the baseline. The assessment draws on multiple data sources,
including stakeholder consultations (interviews and surveys) and desk research in the context of
the supporting study. To the extent possible, the impacts are quantified based on available
assessments or modelling. Where dedicated modelling/monetisation was not possible due to the
lack of data, a qualitative assessment was performed, drawing on existing studies and input from
stakeholders, to ensure transparency on the full range of potential effects. A sensitivity analysis is
provided in section 7.5 to indicate the potential range of error and uncertainty for selected key
estimates. Where central values were not sufficiently informative, or relied on an excessive
number of assumptions, illustrative ranges are provided to inform of the expected consequences
and impacts of the proposed measures. Annexes 3, 4 and 12 provide further details on the
methodological approach, detailed tables and estimates.
In addition, it is worth mentioning that this assessment found practical experience in the public
procurement for sovereign cloud services conducted by the European Commission. The tender
was launched in October 2025, as part of Commission’s efforts to strengthen the digital
sovereignty posture of the European Union Interinstitutional Bodies and Agencies. Even though
they exist in different contexts, the sovereign non price award criteria used in the tender has some
resemblance to the proposed sovereignty framework in PM15/PM21102.
6.1. Economic impact
This section describes first the expected direct, quantifiable economic impact of the policy options
on key identified stakeholders, i.e. industry, including SMEs, public authorities, and the European
Commission. As part of the analysis, an assessment of wider economic effects is then provided,
including impacts on innovation and technological sovereignty. This impact is assessed mainly
based on literature, desk research and stakeholder evidence.
6.1.1. Impact on industry, including SMEs and private sector essential entities
The impact of the policy options on industry was assessed based on the quantifiable expected
costs and benefits for key business stakeholders. First, improving the availability of computing
capacity in PO1-A, PO1-B, PO1-C, directly impacts DC operators and other enterprises building
DCs. Second, reducing the dependence on cloud and AI computing services provided by non-EU
102 In the Commission tender, the non-price award criteria included a number of ‘sovereignty’ criteria against which tenders were scored by the
evaluation committee. In PM15 and PM21, the harmonised sovereignty criteria are assessed by a third party auditor and subsequently verified by a public authority. This leads the service to obtaining a label which is necessary to participate to call for tenders where the contracting authority does
not need to look again into the sovereignty aspects of the service (but can still request information).
57
cloud and AI service providers in PO2-A, PO2-B, PO2-C, is expected to have a direct impact on
cloud and AI computing service providers, on the public sector and on private sector
essential entities in accordance to Annex I of NIS2. This section outlines the quantifiable and
expected direct costs and benefits for these stakeholders in present value over the next 10 years,
while section 6.1.5 presents the anticipated wider economic effects that the proposed measures
could have. The detailed explanations for the assumptions behind these estimates can be found in
Annex 4, sections 2 and 3103.
Adjustment and administrative costs for data centre operators. PO1-A would only foresee
adjustment costs for DC operators’ stemming from their participation in the new Alliance
Working Group (PM1), forum for exchanges (PM2) and from their potential uptake and
discussion of the guidelines on building sustainable DCs (PM3). These costs have been estimated
as ranging between EUR 8-18 m. Under PO1-B,400 estimated DC operators and CSPs active in
building data centres are expected to face one-off administrative costs to comply with the
conditions to access fast-track areas (PM5), i.e. preparing the application file, completing
standardised templates and submitting evidence of compliance with areas conditions. The total
costs for this activity are estimated between EUR 1-3 m at present value over 10-years. Operators
are also expected to face adjustment costs to adapt new projects for DC build out and comply with
the related requirements. These adjustment costs are estimated to be approximately between EUR
6 and 26 m for these stakeholders over the assessment period. The participation in a potential
funding scheme under PM6 would also generate administrative costs to account for the staff time
spent on preparing applications to benefit from funding. Under PM7, businesses would face minor
administrative costs to respond to surveys and/or provide additional information for reporting on
DC capacity. Total costs under this option for data centre operators are thus estimated between
EUR 7 and 31 m. Under PO1-C,similar activities would occur at EU level. Companies
participating in EU funding programmes (PM8, PM9) would face administrative costs for
preparing and managing proposals. 20 proposals every two years have been estimated for such
purposes, leading to costs ranging between EUR 1 and 3 m. As in PO1-B, providers would also be
expected to face both administrative and adjustment costs to access the areas and benefit from
accelerated DC deployment under PM10. These costs have been estimated to range between EUR
5 and 25 m, catering for some uncertainty in the effort dedicated to such administrative and
adjustment activities. The total value of costs for DC operators and CSPs active in building data
centres are thus estimated to range between EUR 7 and 27 m under this option (discounted values
over the 10-year assessment period).
Adjustment and administrative costs for cloud and AI computing service providers. Under PO2-
A, cloud and AI computing service providers would face adjustment costs, estimated as additional
effort to discuss the new guidelines (PM12), participating to the annual conference (PM13) and in
setting up and participating to the coordination group for standards development (PM14). These
activities are expected to generate total costs for providers ranging between EUR 16 and 19 m.
Most of these costs would derive from PM13, where 300 participants have been accounted for.
Under PO2-B, recurrent adjustment costs for providers would emerge mainly from complying
voluntarily with the sovereignty risk assessment framework (PM15). These are estimated at
service level because the audit procedure would apply to individual cloud and AI computing
services, i.e. providers incur costs for each service that undergoes an audit. These costs emerge
103 For readability purposes, direct administrative and adjustment costs and corresponding direct savings or economic benefits are merged, in the
following sections but are further detailed in Annex 4 and broken down by cost type (one-off or recurrent). The policy options are assumed to be implemented from 2027 onwards, with the quantitative assessment covering the 2027-2036 period for the EU and all figures expressed in real 2025
euros.
58
from the one-off effort required to meet the necessary legal, organisational and technical
conditions to reach the sovereignty assurance levels 2-4, undergo third-party audit, and pay audit
and renewal fees. Recurrent audit-related administrative costs are also expected in subsequent
years. Moreover, EU-based cloud and AI computing service providers are expected to incur one-
off adjustment costs to comply with the non-price award criteria (PM16), e.g. rewarding EU-based
R&D&I for innovative services, by altering their systems and processes to be able to participate in
public procurement procedures. Additional administrative costs have been estimated as providers
prepare their bids. Summing these costs leads to total costs ranging from EUR 67 m to EUR 160
m. Under PO2-C, the measures are estimated to generate similar administrative and adjustment
costs for providers as PO2-B, but more services (600 compared to 150 in PO2-B) are expected to
undergo third-party audits to participate in public procurement procedures (PM21). This leads to
higher costs for cloud and AI service providers across the EU that are estimated to range between
EUR 233 and EUR 484 m. As above, these adjustment costs are personnel related, i.e. the effort
needed to modify the providers’ policies and procedures and adjust the legal and organisational
safeguards to be able to meet the criteria under sovereignty assurance levels 2-4.
On top of this, private sector essential entities identified under Annex 1 of NIS2 would have to
incur administrative costs to include non-technical risks in their risk assessments, i.e. adding
sovereignty elements on their existing risk assessment activities. These costs could range between
EUR 486 m and 2.6 bn, i.e. corresponding to the effort required per entity (from a minimum of
20 to a maximum of 110 days) and taking into account 25 600 private entities operating in sectors
of high criticality across the EU.
Focus box #1: costs for service providers to develop sovereign services
Assessing the cost and benefits for providers to provide sovereign services is a complex task that
involves many parameters and differs greatly from provider to provider. As well, in the absence of
an established market for sovereign services, data sources are rather anchored in providers’
business plans, not in ex post analysis of established businesses. Such data is confidential to
companies, and the below considerations are based on targeted discussions with stakeholders that
requested to remain anonymous. A first consideration is that the consulted companies
unanimously indicated that, in developing the business plan for these new services, they count on
new large critical use cases that are today not in the cloud; in other words, they see sovereign
services to generate a new source of income, but not to substitute existing.
Cost wise, new costs notably include the amortisation of all one-off adjustment costs such as the
additional cost induced by using EU-located infrastructure, the additional compliance costs
induced by the audits, the additional costs of being certified under EUCS; and proper recurrent
costs such as the higher salaries of employing EU workforce.
As an illustration, speaking under the condition of anonymity for business secrecy reasons, an EU
service provider specialised in sovereign services speaks of an overall investment of EUR 1.5 bn,
including hardware, for a broad range of IaaS and PaaS services (for an unspecified computing
capacity).
Conversely, another EU service provider with an established range of non-sovereign services
speaks of an overall investment in the range of EUR 20-40 m to adjust existing hardware and
software to the stricter norms that a sovereign service entails, with plans to invest progressively
should the market develops.
The benefits for service providers to develop sovereign services are covered under section 6.1.5.
59
Economic benefits and cost savings for data centre operators. PO1-A would be expected to
generate direct cost savings, via improved information and administrative burden reductions.
Although voluntary, the new guidelines (PM3) would contribute to increase clarity, simplify
procedures and reduce the incidence of mistakes, saving time for operators during the pre-
operation phase of building a new DC, modelled as a central 10-day time saving. When multiplied
by the expected number of facilities that could benefit from this time and consequent cost saving,
the guidelines would generate discounted savings ranging between EUR 1 and 3.5 m over the
assessment period. PO1-B would bring the largest direct economic impact for DC operators,
modelled through a Net Present Value approach which compared the baseline permitting duration
with the accelerated scenario to estimate the economic value of bringing constructions and
commercial operations of a DC facility forward in time. Based on the evidence collected, the
project facilitator for DC roll-out (PM4) combined with access to fast-track areas (PM5) are
expected to reduce uncertainty on investment locations by excluding non-viable sites and decrease
time to market for DC projects by on average 14 months. Under the baseline scenario, the total
time to build a DC, i.e. from planning to entry into operation, was estimated at 32 months on
average across the EU. This time is expected to decrease by 6 months from 2027 onwards due to
the project facilitator and by an additional 8 months due to streamlined access to fast-track
areas104. The estimated NPV gain is also associated with the expected reduction in PUE foreseen
under this measure, as only the most sustainable DCs would be able to benefit from the fast-
tracking. Therefore, the expected improvements in energy efficiency, modelled as a declining
PUE, would also contribute to reducing operating costs over time. Over the 10-year assessment,
this option could result in discounted economic benefits ranging between EUR 8 and 27 bn,
considering the uncertainty related to these time savings and their compounded impact on project
value. Under PO1-C, similar direct economic benefits are foreseen for operators thanks to the EU-
level identification of fast-track areas (PM10). Based on discussions with industry and following
the validation by the sector in the targeted workshop, these savings are expected to be less
consistent than under PO1-B, with an EU-level management of fast-track areas expected to
produce time-savings of 3 months. This would in turn still generate consistent economic benefits
that would range between EUR 5 and 12 bn over the 10-year period.
Figure 10 shows the potential evolution of DC supply under each policy option and the expected
reduction in the gap against projected demand in the baseline scenario105.
104 Based on study findings and validated by the sector in the targeted workshop. 105 Annex 4, Section 2.3.3 further explains how the capacity trajectories have been used as the quantitative basis for estimating how many new data
centre could be built under each option and the share of this capacity that is expected to benefit from the policy intervention over time.
60
Figure 10. Forecasted evolution of data centre capacity (GW) under each PO, with respect to the baseline
Source: Technopolis et al. (2025)106clxv
Cost savings for cloud and AI computing service providers. Under PO2-A, administrative cost
savings for providers would mainly stem from the use of the guidelines (PM12), saving time for
operators during procurement, modelled as a modest 0.5 days saved per tender procedure, leading
to total costs savings that could range between EUR 2 and 15 m (10-year, NPV). Under PO2-B,
providers would face recurrent cost savings from the introduction of a clear and predictable
sovereignty risk assessment framework across several Member States (PM15). Instead of
responding to different requirements, controls or evidence requests from individual public
authorities, service providers would benefit from a common audit procedure and recognised
assurance levels with savings ranging fromEUR 5 to 16 m per service, under an assumption that
the provider operates in 3 to 10 markets. This would reduce duplication and simplify procurement
procedures, while reducing costs related to tender documentation. Over time, this is expected to
lower compliance costs and make it easier for providers to offer the same audited services to a
wider public sector market. This is expected to lead to total benefits for providers ranging between
EUR 200 and 700 m, when discounted across 10 years. Under PO2-C, direct costs savings would
stem from the same mechanism explained above, but extended to all MS, and an increased number
of services being audited by providers under PM21, compared to PM15, leading to total expected
benefits ranging between EUR 2 and 5.1 bn, including those savings coming from savings of 1 to
3 working days per bid when drafting specifications by using standard non-specific award criteria
(PM19). In the case of PO2-C, savings in the audit process would range from EUR 8 to 25 m per
service, assuming each provider run its business in 5 to 15 member states.
Impact on SMEs. Though large businesses dominate the DC, cloud and AI markets, many SMEs
in the EU still play a relevant role, particularly providers of cloud and AI services. Thus, the
initiative is considered “relevant” for SMEs. Annex 6 contains the SME test. Given the lack of
quantitative data, the impact on SMEs is assessed qualitatively below.
PO1-A would generate value for SMEs with improved access to information and clearer
frameworks, notably through the guidelines. Early visibility and clarity on applicable requirements
106 The capacity includes hyperscalers and colocation providers, but not enterprise DCs in the private or public sector.
48
54
60
0
10
20
30
40
50
60
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036
G W
Data centre supply evolution forecast (GW) under each Policy Option, with respect to the baseline
Baseline
PO1A
PO1C
PO1B
Demand
estimate
61
and collective voicing of SMEs would reduce costs linked to information asymmetries and
coordination failures. As smaller operators, SMEs are disproportionately affected by complex
permitting procedures and would benefit from administrative simplification. In PO1-B
streamlined procedures are expected to reduce legal and consultancy expenses and would
accelerate timelines, lowering entry barriers, reducing administrative costs and opening
opportunities for SMEs in the DC value chain. The impacts of PO1-C on SMEs align to those of
PO1-B, though EU-level approval might be perceived as more “distant”, raising language
considerations of such mechanism. In addition, public funding for R&D and innovation would
enhance SME participation in advanced projects without requiring large upfront capital. Public
funding for strategic deployment projects would improve SMEs’ access to contracts for large-
scale projects, either directly or within a consortium. PO2 would impact SMEs as providers of
cloud and AI computing services, but also as consumers of these services. Under PO2-A, the non-
prescriptive EU criteria of sovereign cloud and AI computing services and the guidelines would
reduce administrative burdens on SMEs while increasing clarity and transparency for sovereign
services.The annual conference on EU digital sovereignty would allow SMEs to showcase their
services and raise concerns collectively at lower costs. SMEs would also be the greatest
beneficiaries of the interoperability measures as it simplifies the re-use of existing technology they
lack. Under PO2-B, the sovereignty audit procedure with validity throughout the EU would
reduce verification costs, especially for SMEs offering their services in critical sectors to public
administrations in different Member States, albeit facing more burdensome compliance costs,
voluntary public procurement criteria favouring specific characteristics could boost SME
participation. Additionally, a vendor-neutral training would provide SMEs with a highly qualified
workforce, capable of using and developing advanced cloud and AI solutions. PO2-C would
entail the greatest direct and indirect net benefits for SMEs. The validity of audit reports across the
EU would allow SMEs’ services would allow them to extend their commercial offers otherwise
hindered. Beyond, remaining costs include funding programmes participation, which is also
expected to enhance SMEs’ competitiveness. Open source access would reduce entry barriers,
where solutions are evaluated on their merits. Encouraging open source would not disadvantage
proprietary solutions where these genuinely represent the option on the best-value for money, but
this would need to be properly demonstrated. Similarly, for companies, notably SMEs whose
business models are built around the delivery of open source solutions and related services, this
measure would provide them with the opportunity to present their solutions on equal terms
removing barriers that have historically limited open source participation in public procurement
procedures.Similarly, financial support would include improved funding rates for SMEs, and the
service toolbox would help smaller providers find complementary solutions to their service
offering. Ultimately, this would foster SMEs’ access to innovation, seeking to increase their
competitiveness and reduce their upfront costs. The SME cloud and AI adoption scheme would
bring direct economic gains for SMEs applying cloud and AI technologies through improved
competitiveness, and indirect gains for IT SMEs offering consultancy, reinforcing the ecosystem
to drive a broader SME digital transformation.
SMEs’ views on Policy Measures. Feedback from the public consultation shows the following:
Table 9. SMEs and large companies views on Policy Measures
Measure Support from
SMEs*
Support from large
companies**
Criterion ensuring sovereignty, autonomy, resilience for administrations
procuring cloud and AI computing services 88% 31%
Supporting an open source software ecosystem 82% 55%
Guidelines with standard criteria to procure cloud and AI computing
services for public administrations 82% 39%
62
Measure Support from
SMEs*
Support from large
companies**
Tax incentives for using sustainable technologies 76% 39%
Funding for R&D of energy-efficient technologies, standardised energy
efficiency benchmarks; addressing energy availability for data centres 71% 59%
Expedited approval mechanisms and clear conditions for strategic projects 70% 81%
*17 SMEs replied to the public consultation questionnaire ** 49 Large companies replied to the public consultation questionnaire
6.1.2. Impact on public authorities
Costs for public authorities. In PO1-A, costs for authorities are estimated to be modest and would
be limited to their participation in the forum (PM2) as well as to their contribution to drafting the
guidelines (PM3), leading to total expected costs ranging between EUR 1 and 2 m over 10 years,
assuming one authority representative per MS would participate in each activity. Under PO1-B,
authorities are expected to face adjustment and administrative costs to establish and operate the
project facilitator (PM4)107 and identify areas for fast-track deployment as part of their national
strategies for data centre deployment (PM5). This would lead to an estimated cumulative cost over
10-years ranging between 4 and 8 m per national authority. Additional costs are estimated for
setting up and managing a potential support programme for strategic DC roll-out (PM6). Finally,
authorities would face minor enforcement costs for periodic checks to verify data on compute
capacity as well as the implementation of national data centre strategies under PM7. The option is
expected to generate total costs for 27 national authorities ranging between EUR 106 m and 236
m over ten years. In PO1-C, authorities are expected to incur adjustment costs to adapt to the
EU-level DC fast track platform (PM10), including the appointment of national representatives to
the new board and periodic provision of data for identifying suitable fast-track areas. This is
expected to generate costs ranging between EUR 84 and 90 m over ten years. Under PO2-A, the
measures would generate modest adjustment costs for national authorities. The EU-level criteria of
sovereign cloud and AI computing services (PM11) are expected to generate adjustment costs
related to the review of national procurement templates, while guidelines (PM12) are expected to
generate costs in relation to their adoption and prior consultation process. This would lead to total
expected costs for authorities ranging from EUR 14 to 15 m in discounted terms over 10 years.
Under PO2-B, authorities would face increasing costs regarding the sovereignty risk assessment
scheme (PM15), the award criteria (PM16), participation in the public sector cloud federation
(PM17). First, authorities would face one-off adjustment costs to carry out the sovereignty risk
assessment to map sovereignty assurance levels to cloud and AI computing services used in the
public sector. They would face recurrent costs for their periodic renewals and for revision of the
audit reports (PM15). These are relatively modest and estimated in a range of EUR 25 to 99 m.
Under this measure, contracting authorities could consider using the sovereignty framework for
their procurement decisions. See the box below for the assessment of the potential costs for public
authorities related to porting and migrating cloud services. Under this option, since the measure is
not mandatory, a smaller group of public authorities has been considered to use the sovereignty
framework. The intervention could accelerate the porting of a limited subset of critical cloud
applications to sovereign levels 3-4. Based on this, the anticipated porting is assumed to concern
only between 1% and 6% of the relevant application base, resulting in discounted costs ranging
from EUR 620 m to EUR 4 bn, over three years from 2030. Minor costs are estimated for
107 This estimate represents a worst case scenario in which authorities would need to establish a dedicated team from scratch, whereas in practice Member States may rely on existing resources under the Gigabit Infrastructure Act, including by designating a single information point established
under Regulation (EU) 2024/1309, with the relevant functions, procedures and mechanisms applying accordingly.
63
authorities voluntarily updating their procurement templates with the new procurement criteria
(PM16). The platform for public sector cloud federation is expected to generate the costs ranging
between EUR 170 to 260 m in discounted terms. Summing all these activities leads to total costs
under this option to range between EUR 820 m and 4.4 bn. PO2-C foresees costs previously
described but scaled upwards as well as new ones. For the former, elements include both the one-
off and recurring costs for the sovereignty risk assessment scheme, reflecting the higher number of
public authorities that would perform them compared with PM15. Under PM15, around 150
audited services are expected to be verified by 2036, while under the mandatory approach in
PM21 this figure is projected to rise to 600 in the same period, which is broadly consistent with
benchmarks such as FedRAMP. Also, in the case of PM15, given the voluntary nature of the
scheme, only a limited number of Member States would perform the risk assessment and the
consequent verification of audited services. Under this option, the mandatory nature of PM21
assumes that all MS would prepare the personnel and the infrastructure to run the scheme. Hence,
the changes of PM15 and PM21 would need to be seen in combination: increase in the number of
authorities adapting their processes to perform the sovereignty risk assessment and an increase in
the number of audit reports to be verified. Therefore, under this option, the cost for authorities
assumes that the policy intervention would accelerate the porting of the limited subset of critical
applications across the full population of 6 400 public entities considered in scope. This would
result in total discounted costs of around EUR 3 to 15 bn, over the same period. Under PM19,
authorities would also face some costs to update their procurement templates with the new
procurement criteria, alongside minor effort to draft an action plan or light touch strategic note
outlining how they intend to use public procurement to increase the uptake of highly innovative
cloud and AI computing services. PM20, which aims to boost open source in public
administrations is expected to produce several costs for public authorities, i.e. related to (i) the
creation and maintenance of the Open Source Programme Offices (OSPOs), (ii) setting up open
source repositories and governance mechanisms defining contribution, review, and acceptance
procedures, (iii) the continuous maintenance and set-up of additional repositories, (iv) the
activities needed to release code as open source and (v) the comparative assessments of
proprietary and open source solutions for their procurement procedures. While currently public
authorities may not have this knowledge in-house, the initial comparative assessments and the set-
up of the repositories may require certain investment in staff training and in the development of
training evaluation capacity in the contracting authorities. However, these administrative costs are
expected to be mitigated thanks to the deployment of the Open Source Programme Offices
(OSPO). OSPOs, as part of their role in their technical, legal, procedural and strategic – related
tasks, may provide contracting authorities with standardised evaluation frameworks, scoring
methodologies or advisory support thereby sharing expertise costs across the public sector rather
than requiring each authority to develop this capacity independently. Finally, in PM22 for the
public sector cloud federation, national public authorities will have the same costs as described for
PM17 above, as it builds on top of it. These measures would result in total costs for public
authorities amounting to a range of EUR 4 to 18 bn for this option.
Focus box #2: costs for authorities to port and migrate services
This box makes the distinction in between:
• Cloud porting, i.e. moving a service from one cloud provider to another cloud provider
• Cloud migrating, i.e. moving a service from on-premises to cloud
None of the proposed policy measures obliges public authorities to port an existing cloud service
from one provider to another, or to migrate on-premises service to the cloud or to port. These
decisions are made on a case-by-case basis by public authorities alone, considering the authority's
64
specific needs, existing systems, and procurement cycle.
Examining costs is however relevant since the results of Member States' sovereignty risk
assessments are expected to induce cloud porting decisions108. As well, the expected increase in
trust resulting from the existence of a sovereignty framework is expected to increase or anticipate
cloud migration decisions. These costs are not immediate consequences of the intervention's entry
into force, but rather potential future expenses that may arise during the normal course of
business. These costs are thus presented for illustrative purposes.
Porting an existing workload from one cloud provider to another (cloud-to-sovereign cloud)
Porting refers to moving an existing cloud-based service from one cloud provider to another, in
this case to a sovereign cloud service. This concerns services that are already cloudified but may
need to be moved/ported if sovereignty risk assessments conclude so. It is worth noting that this is
something that anyway happens naturally as existing contracts come to an end and are re-tendered.
To port an existing service, costs include a feasibility assessment, strategy and planning, target
environment setup, migration, testing and validation, cutover, DNS switching, and recurrent
activities linked to operating the service in the new environment, including performance
optimisation and maintenance. These activities mainly involve personnel costs and parallel
running costs.
It is important to underline that the cost of porting is comparable when systems are ported from
one cloud providers to another or to a sovereign cloud environment. In other words: the fact that
the end cloud is sovereign does not per se affect the cost of porting.
The porting of a single system
To estimate the cost of porting an IT system, applications can be grouped into small, medium-
sized and large, based on their size, complexity, business case and infrastructure needs. Annex 12
provides an in-depth analysis of the cost categories and their magnitudes, which can be grouped in
the following categories:
Application type Indicative effort Estimated cost
Small application 50 days EUR 20 000 – 50 000
Medium-sized application 400 days EUR 200 000
Large application 1000 days EUR 500 000
Real-life use case 1: a porting of a high usage application to a sovereign PaaS
One real-life example concerns the porting of an existing application from a cloud to a sovereign
Platform-as-a-Service environment.
• In the current situation, the service costs approximately EUR 2 m per year with a large
commercial cloud provider.
• The target scenario is porting the application to a sovereign PaaS environment.
• The application is a monolithic Laravel application with a large-scale MySQL database
and several large Elasticsearch clusters.
• It does not process or store personal data, but it has a high usage profile, with sustained
traffic and high-intensity connections.
• Its criticality is mainly reputational, as an incident could have a high public impact,
108It must be noted that porting can also be triggered naturally as a consequence of regular procurement decisions after the expiration of cloud service contracts.
65
although the service is not considered real-time critical from a business continuity
perspective.
The estimated transition effort is around 2 FTEs over six months, broken down as follows:
Workstream Estimated cost
Application adaptation, migration and validation 1 FTE
Infrastructure provisioning, platform adaptation and automation 0.5 FTE
Security and compliance alignment 0.5 FTE
Total 2 FTEs over 6 months
This example shows that the initial migration requires a structured investment to adapt the
application, infrastructure and security posture to a sovereign environment. However, part of this
effort can be reusable for future migrations, particularly infrastructure patterns, deployment
practices and security baselines. The learning curve is therefore front-loaded, meaning that
subsequent migrations of similar systems could benefit from the developed experience and skills.
The actual effort may vary based on the application’s technological stack, its dependence on
cloud-specific services and its overall architecture complexity.
Real-life use case 2: transition of an application based on a high-volume database to a sovereign
PaaS
Another real-life example of a transition of an existing application to a sovereign Platform-as-a-
Service environment:
• The application, based on a few tenths of large scale microservices managing a very high-
volume document database (1.5 bn records), is hosted by a large commercial cloud
provider and is migrated to a sovereign PaaS environment.
• The application is monolithic with a large-scale relational database, WebLogic application
server and proprietary monitoring tools.
• It does not store or process personal data, just public documents, but it has a high usage
profile, with more than 100 m visits and around 20 m documents consulted every year.
• Its criticality is mainly reputational, as an incident could have high public visibility,
although the service is not considered real-time critical from a business continuity
perspective.
• The chosen migration strategy was re-platforming, i.e. migrating the application and the
database, and moving from proprietary solutions to open-source solutions (e.g. Apache
Tomcat, Postgre SQL, Virtuoso) which entails refactoring code to eliminate dependencies
to the deprecated technologies and to align it with the new technological choices, while
maintaining the same performance.
• The estimated transition effort is around 8 000 hours, or 4.5 FTEs over a year.
What would sovereign cloud porting cost for a typical public authority?
Extrapolating the individual costs of porting different types of applications to the full set of public
authorities potentially affected by the policy measures 15 and 21 would require considering
several parameters. These include:
• The outcome of the risk assessment with the sovereignty assurance level applicable to the
concerned applications
• The number and complexity of IT systems managed by each public authority
66
• The current level of cloudification of each authority
• Whether the public authority’s data are already hosted in the EU under their existing cloud
arrangement.
On this basis, three illustrative cases are presented to show how costs can vary across different
types of public authorities. These scenarios provide a benchmark for individual authorities, while
the aggregated cost across public entities possibly in scope under PO2-B and PO2-C is detailed
under Annex 12. The estimates cover the additional anticipated porting of a limited subset of
critical applications to levels 3-4 where such porting could be directly accelerated by the
intervention.
Small public authority
A small public authority, e.g. a mid-size municipality, manages 200 IT systems, with limited use
of cloud services. Around 30% of the systems, i.e. 60, would need to be ported to a sovereign
cloud environment. Most of these systems would be small applications (70%), e.g. standard
administrative tools, with only limited number of medium (29%) or large platforms (1%). In this
scenario, following the risk assessment outcomes, all IT systems (60) are subject to sovereignty
requirements, as the authority never considered sovereignty criteria in previous tenders. Most
applications (70%) fall under lower sovereignty assurance levels (level 1) while 20% need to be
moved to sovereignty level 2 and 10% at least sovereignty level 3. Based on this, the cost of
porting for the authority is expected to reach around EUR 4.7-6.0 m over five years,
corresponding to around EUR 1 m per year.
Medium public authority
A medium sized public authority, e.g. a national agency, manages around 800 IT systems, with a
medium intensity cloud-user profile. Around 30% of its systems are cloudified, i.e. 240, and could
be subject to higher sovereignty requirements, i.e. level 2 and above. Based on the outcome of the
risk assessment, only 30% of these IT systems would require porting to a sovereign cloud
environment, i.e. around 72 systems. Most of these systems are small (60%) or medium
applications (37%), with a limited number of large and complex platforms (3%). On this basis, the
cost of porting for this authority would reach around EUR 7.3-8.6 m over five years, or around
EUR 1.6 m per year.
Large public authority
A large public authority, e.g. a large ministry, manages 1500 IT systems, with intensive use of
cloud services, e.g. mission-critical platforms used by citizens, and a higher number of systems
subject to sovereignty requirements. Compared to the medium case, it would manage a larger and
more complex portfolio, including a larger share of medium (45%) and large (5%) applications.
For this scenario, only critical systems falling under Level 2 and above are expected to require
porting to a sovereign cloud environment as the authority may already be hosting most of its IT
solutions with a cloud service falling under level 1 or on-premises. The cost of porting would
reach around EUR 16.6-18.6 m over five years, i.e. around EUR 3.5 m per year.
It is important to distinguish this cost of porting from the price of cloud services, which can be
affected by a premium, something discussed under section 2.3.4.
Migration of legacy applications to the cloud
Most public sector use cases are not cloud-based today with 70% estimated to be on-premises.
This can be further split, in between 30% that are on premises and could be migrated to the cloud,
and a leftover of 40% of on-premises that are legacy applications being phased out (sometimes
67
long periods of time), or that are too small to deserve particular attention.
To migrate services from on-premises infrastructure to the cloud, public authorities face
adjustment costs associated mainly with staff time and migration activities. The effort to migrate a
legacy-to-cloud application are harder to estimate than the ones for porting because the range of
situations is much broader. Annex 12 details them further, leading to an estimate 1.5x the effort
compared to porting a service from cloud-to-cloud.
These migration costs need to be put in perspective with the benefits of cloudification. Literature
points to potential benefits such as lower infrastructure costs, greater scalability and flexibility,
improved service quality, and access to more advanced security and analytical capabilities. These
potential benefits are discussed further in Section 6.1.5.
Cost savings for national public authorities. PO1-A is expected to realise direct cost savings for
authorities in terms of administrative burden reduction from harmonised EU guidelines that would
limit divergent interpretations of permitting requirements (PM3), leading to savings that would
range between EUR 0.1 and 0.6 m, modelled as a 2-day saving per new project adopting the
guidelines. Under PO1-B, recurrent savings would stem from administrative simplification for DC
buildout (PM4), combined with the central mechanism to identify fast-track areas (PM5), which
would avoid duplication and back and forth interactions with economic operators, leading to total
expected administrative cost savings under this option ranging between EUR 166 and 277 m. In
PO1-C,savings would result from a centralised EU-level structure reducing national efforts
(PM10). Consequently, total cost savings for 27 public authorities could be between 6 and 19 m,
based on the variability in the actual effort saved by using a central EU-based database and
avoiding repeated clarifications with operators on siting rules.Under PO2-A, cost savings are
expected to arise from clearer information used in procurement processes. However, the scale of
this effect appears to be very limited to be quantified. Under PO2-B,the sovereignty scheme and
repository (PM15) would help streamline compliance verification by providing information in a
single place. Participation in the public sector cloud and AI computing service federation (PM17)
is expected to generate considerable resource savings through an optimised use of compute
capacity. Finally, the use of standard award criteria (PM16) is also expected to save authorities
time in procurement processes, leading to total expected savings ranging between EUR 8 and 17
bn. Under PO2-C,cost savings derive from standardisation and automation at scale. Mandating
sovereignty risk assessments (PM21) are expected to create a more systematic and trustful manner
to assess needs for cloud and AI computing services and reduce their time for procurement thanks
to the comfort of having compliance with sovereignty requirements verified ex-ante. Today, there
are in several Member States (ex: France, Italy, Finland) some schemes and methodologies used at
national level for the overall security and resilience of cloud computing services. Whereas a
voluntary approach may not lead to convergence between them, a compulsory system as proposed
in this measure, combined with Commission’s support and oversight, will have more benefits in
terms of synergies between Member States, and will also complement other measures proposed to
consolidate efforts. Notably, participation in the public sector cloud and AI computing service
federation alongside voluntary EU-level joint procurement (PM22) are expected to generate
significant savings through an optimised use of compute capacity, a more cost-efficient
procurement and improved economies of scale. Greater use of open source (PM20) is expected to
reduce proprietary license expenditures, lowering licensing costs and total cost of ownership for
IT systemsclxvi. These measures could altogether lead to considerable cost savings for authorities
ranging between EUR 21 and 61 bn over 10 years. A significant part of these savings come from
the EU-level joint procurement that starts managing 2% of the total cloud and AI procurement
68
volume of national contracting authorities reaching 20% at the end of the 10-year period and
moving from 10% savings in first year to a range of 20-40% at the end of the period as economies
of scale increase.
6.1.3. Impact on the European Commission
Administrative and adjustment costs. PO1-A is expected to generate one-off and recurrent
administrative costs, regarding the development of guidelines (PM3), the establishment of the new
Alliance Working Group (PM1) and DC forum (PM2), leading to modest additional costs with
respect to the baseline of around EUR 0.5 m. PO1-B would entail one-off adjustment costs and
recurrent ones for establishing and subsequently managing the coordination hub supporting
Member States in identifying fast-track areas (PM5), as well as adjustment effort to monitor and
coordinate the monitoring and reporting on computing capacity (PM7), leading to expected total
discounted costs over 10-years of around EUR 1 m. PO1-C wouldgenerate the highest economic
burden for the Commission, with adjustment costs linked to setting-up and operating the 7-year
funding programme for R&D and innovation ecosystems (PM8) and for funding the deployment
of strategic projects (PM9). Finally, the EU-level identification of fast-track areas (PM10) would
entail one-off adjustment costs to set up the mechanism and procuring a supporting digital tool,
along with recurrent costs to manage the mechanism to identify areas for fast-tracked DC
deployment and biennial adjustment costs for outsourced studies and expert meetings, leading to
total costs of EUR 17 and 19 m. PO2-A is expected to generate one-off adjustment costs for the
Commission from drafting and consultation for the development of an EU-criteria and guidelines
on the notion of sovereign cloud and AI computing services (PM12). The annual week-long
conference on digital sovereignty (PM13) would also entail recurrent adjustment costs for
preparation time and event budget. Finally, the increased standardisation efforts on interoperability
(PM14) are expected to generate modest one-off adjustment costs for procuring ad hoc studies and
setting up the coordination group, and recurrent costs for participating and supporting discussions.
This would lead to total costs of around EUR 7 m. Under PO2-B, costs are expected to arise from
establishing and managing the repository of sovereign services (PM15), setting up the public
sector cloud federation platform (PM17), and develop the training programme (PM18). These
measures are expected to amount to discounted total costs of approximately EUR 130 m. In PO2-
C,costs for the mandatory sovereignty risk assessment (PM21) would mirror those listed under
PO2-B and comprise the procedures for setting up the repository. EU-level procurement (PM22)
would generate adjustment costs related to hosting and maintaining the platform, and recurrent
ones for operating it for joint procurement. The development, set-up and administration of the
adoption scheme for SMEs (PM23) is expected to generate the highest adjustment costs under this
option. Finally, the online toolbox for service providers to easily identify software tools to help
them in complementing their service offering (PM24) is expected to generate costs related to the
development and management of the toolbox. The measures are altogether expected to generate
costs for the Commission amounting to in- between EUR 470 and 760 m at present value over the
next 10 years.
6.1.4. Impact on innovation and technological sovereignty
PO1-A would not expand the supply chain or balance dependencies on non-EU providers.
Innovation effects would also be limited. PO1-B would be very impactful in terms of
technological sovereignty and innovation. The funding of strategic projects would allow national
public authorities to steer the development of infrastructure to sectors relevant to sovereignty and
to DCs that deploy novel technologies for resource efficiency. PO1-C would have a higher
innovation impact than PO1-B. Funding the development and deployment of novel technologies
69
would give an edge to the European DC ecosystem. PO2-A would have moderate impact on
technological sovereignty. Soft harmonisation measures like defining a sovereign cloud and AI
computing services, guidelines, and establishing an annual conference on digital sovereignty
would indeed create a common understanding of sovereignty but would have limited impact. The
development of harmonised interoperability standards would enhance the uptake of sovereign
technologies. PO2-B would have greater impact than PO2-A on technological sovereignty in the
Union, as the voluntary sovereignty risk assessment would allow public authorities to procure
cloud and AI computing services that comply with the sovereignty criteria. The federation would
ensure that authorities benefit from de facto sovereign cloud and AI resources shared by other
public bodies and would enhance resilience by allowing access to capacity located in other
Member States. The voluntary award criteria on R&D investment and supply chain reinforcement
are expected to encourage the development of an EU located innovation ecosystem. PO2-C,
which builds on PO2-B, is expected to have the greatest impact in terms of innovation and
technological sovereignty. The sovereignty risk assessments performed by the public and private
sectors will improve the overall sovereignty level of the EU economy and will provide additional
opportunities and incentives to procure from European providers that are more immune to risks
linked notably to data access or throttling of service quality. This in turn will encourage
investment to improve innovation and technological sovereignty in the EU. The establishment of a
framework for joint procurement would allow public authorities to increase their autonomy and
resilience while funding innovative solutions in Europe. The non-price award criteria would allow
contracting authorities to procure cloud and AI computing services to favour investment in
innovative cloud and AI computing services, while reducing critical dependencies. The open
source principles would also play a role in the increase of the Union’s operational autonomy,
particularly in the public sector. Finally, the toolbox for cloud and AI computing service would
support European cloud and AI computing service providers in complementing their service
offerings and increasing their market visibility, by addressing fragmentation and lack of
discoverability in the cloud market. Therefore, the toolbox is expected to contribute to reducing
dependency on a limited number of dominant solutions.
6.1.5. Wider economic effects
Beyond the direct costs and benefits quantified above, the proposed measures may generate wider
economic effects through two main channels: first, support the deployment of data centres, with a
focus on strategic ones prioritising innovation and sustainability, and second, by increasing the
adoption of sovereign cloud and AI computing services. These effects are difficult to quantify
because they relate to resilience, autonomy, avoided disruption and long-term competitiveness.
However, given their economic relevance, the main evidence collected is summarised here as part
of the overall assessment.
Compute Capacity. The overall macroeconomic impact of constructing new data centres is best
analysed qualitatively. The evidence base remains fragmented in part because official statistics
often fail to categorise data centres as a distinct sector, and in part because several studies rely on
project-specific case studies, surveys or modelling that cannot be scaled to the whole European
context without large multipliers. Existing literature suggests that the construction and
deployment of new data centres have a positive impact on the economy in the short to
medium term. The development of new data centre projects generates demand for various goods
and services, including construction, engineering, electrical equipment and professional services.
These in turn can have ripple effects on local economies through related supply-chains and
induced household spending. However, the extent of these effects is heavily influenced by local
procurement patterns, labour market conditions, and degree to which these expenditures are
70
retained domestically. Studies indicate that investing in additional data centre capacity can
stimulate economic activity during the construction phase, but the magnitude of this effect
depends on the specific context109. Some have attempted to quantify the contribution of data
centres to GDP or value added. For example, in the UK, it has been estimated that the data centre
sector generates around GBP 4.7 bn in annual Gross Value Added. Similarly, a 2025 study by
Copenhagen Economics estimates that in Portugal the data centre sector contributed EUR 311 m
to GDP between 2022 and 2024, with potential cumulative contributions ranging from EUR 6.1 bn
to EUR 26.2 bn between 2025 and 2030, depending on the investment scenario110.
When looking at international trade, expanding EU DC capacity is also expected to affect the EU’s
external trade balance by increasing imports of semiconductors and ICT components. Today,
Europe has a structural import dependency on semiconductors and the most advanced AI chips
used in DCs are designed and fabricated outside the EU, mainly in the US, Taiwan, South Korea
and Chinaclxvii. Each new GW of installed capacity requires several millions of components,
representing between EUR 3 to 4 bn of ICT hardware expenditure per GWclxviii. Applying this ratio
to the EU’s projected 2025-2030 buildout, total chip and server imports could reach EUR 47 to 83
bn over this period. This dependency is expected to further catalyse foreign direct investment
(FDI) and technology upgrades within the EU, with technology firms potentially contributing to
the development of clusters around the growing DC sectorclxix. At EU level, in 2024, investors
announced FDI projects worth around EUR 60 bnclxx.
From a broader economic perspective, the development of additional compute capacity serves as
an important enabler for cloud computing, AI and other data-intensive activities. The
primary economic benefit is likely to arise from the downstream effects on productivity and
innovation, rather than the construction and operation of the data centres themselves. This in turn
can have a positive impact of the economy by supporting growth and regional digital
development. According to the OECD, cloud services play a key role in the value chain for many
digital technologies used by businesses and governments, and can enhance innovation and
productivity, particularly for SMEs that would otherwise face high fixed costs in acquiring
advanced ICT capacityclxxi. Scalable and secure compute and cloud-based services can empower
enterprises and public authorities to access advanced technologies without having to build the
infrastructure in-house, something which requires time, a particular skill set and large upfront
investments. Better access to digital infrastructure can improve the conditions for technology
diffusion and digitalisation across sectors, ultimately giving rise to data-driven business models,
including AI-enabled services. In the EU context, this is consistent with the policy objective of
ensuring that strategic data centre capacity supports public interest use cases, such as healthcare,
public administration, research, security, critical infrastructure and AI-enabled public services.
The ultimate benefit of fostering this deployment would reside not in the creation of more data
centres, but in the availability of strategically located and appropriately governed compute
capacity for economically and socially important workloads. As noted above in Section 2.2.1,
the availability of local data centre access can also have an impact on latency, reliability and cost
conditions, which can encourage greater cloud adoptionclxxii. Furthermore, local data centres can
enable more advanced AI applications by reducing latency and improving internet traffic, while
109 See: The local economic impact of a proposed data centre campus in London and Virginia Joint Legislative Audit and Review Commission
Report to the Governor and the General Assembly of Virginia on data centers 110 The study focuses on immediate impacts in terms of investments and jobs and secondary effects on businesses supplying DCs. It does not
account for broader spillover effects of DCs on innovation and productivity across sectors or broader economy-wide effects, such as potential
impacts on price changes or shifts in consumer behaviour. Available here. The report of the European Data Centre Association projects the European colocation DC sector’s GDP contribution to increase from EUR 30 bn in 2023 to EUR 83.8 bn by 2030. Available here. See also: Does
GDP growth minus AI capex equal zero?
71
insufficient domestic compute capacity can create or deepen dependence on external providers.
With respect to the impact of AI, recent literature highlights its potential to generate economic
benefits, particularly through productivity gains, innovation, scientific discovery and new business
models. However, the evidence is strongest for task-level productivity gains and more uncertain at
enterprise, sector and macroeconomic level.
Cloud and AI computing services. The second channel of wider effects relates to the increased
adoption of sovereign cloud and AI computing services. Benefits include enhanced autonomy,
improved resilience, reduced dependency, stronger legal and operational control, and better
conditions for AI-enabled innovation. Even though several of these effects are hard to monetise,
they are economically relevant as cloud and AI computing services increasingly support public
administration, healthcare, research, education, justice and law enforcement, cybersecurity and
industrial competitiveness. If public authorities depend heavily on a small number of external
providers or non-EU jurisdictions, they may face legal, operational, geopolitical and continuity
risks. Sovereign cloud and AI services can reduce these risks by providing more diverse supply
options, clearer governance, and ensuring stronger alignment with EU and Member State
requirements.
A key benefit is therefore greater resilience and continuity of public services. Dependence on a
limited number of cloud infrastructures can create systemic risks, as disruption in a major provider
can affect many services at the same time. The October 2025 AWS outage, for example, disrupted
businesses and digital services globally, highlighting this vulnerabilityclxxiii. The World Economic
Forum estimated that the 2024 global CrowdStrike outage caused around USD 5 billion in losses,
disrupted airlines, banks, healthcare providers, retail payment systems and ATMs worldwide,
underlining these system-wide economic consequencesclxxiv. These examples do not imply that
sovereign providers are immune from disruption. Rather, they show the economic relevance of
resilience, provider diversity, and reduced single-provider dependency. Indeed, sovereign cloud
services strengthen operational control. Sovereignty does not only mean storing data in the EU.
It also concerns who operates the service, who can access systems, which jurisdiction applies, and
whether public authorities retain effective control over sensitive workloads. These are particularly
relevant for strategic public sector use cases, such as hospitals, emergency services, tax systems,
justice systems or critical infrastructure, where continuity and trust outweigh marginal cost
optimisation. In addition, greater availability of credible EU alternatives may have competition
and cost benefits. Moving part of its services to EU providers would give public administration the
capacity to improve their negotiating power and contribute to decreasing this market
concentration.
Another relevant potential wider economic effect concerns the domestic value added through
import substitution. A higher market share for EU providers would shift part of existing and
future cloud and AI spending towards services produced within the Union. This would increase
the share of wages, profits, reinvestment and tax revenues retained in the EU rather than accruing
them abroad. This, in turn, could stimulate additional investment in cloud and AI infrastructure
within the EU, enabling European providers to scale up their offerings and strengthen their
competitiveness. An illustrative calculation shows the possible order of magnitude.
Considering that the EU-27 public cloud revenues could reach around EUR 320 bn by 2030111,
that the public sector represents around 14% of demand, and that level 3 and 4 sovereignty levels
correspond to 10% of these needs (see description of policy measure 21 here above), this means a
111 Public Cloud - EU-27 | Statista Market Forecast
72
market of EUR 320 bn * 14% * 10% = EUR 4.48 bn only addressable by EU controlled providers
by 2030112.
On top of this, it is likely that the signalling effect of European providers being selected for highly
sensitive applications will give them increased opportunities in the market, and it can therefore be
expected that European providers would increase their market shares also for lower sovereignty
levels. Similarly, in the private sector, as regards sensitive sectors subject to a new obligation to
pay attention to sovereignty issues, it can be expected that a number of entities will consider
giving new business to European entities, which will have less difficulty in demonstrating their
sovereignty levels. Sovereign cloud and AI adoption may support productivity gains, although
these are still uncertain. McKinsey estimates that Europe could unlock up to EUR 480 billion
annually by 2030 in a “European digital sovereignty” scenario, where AI adoption is high and
European providers capture most of the value. In a high adoption but externally dependent
scenario, the estimated impact is lower, at around EUR 375 billion. In lower-adoption scenarios,
the estimated impact falls to around EUR 80 to 100 billionclxxv.
Increased availability of sovereign cloud and AI services may also increase trust among public
authorities and thereby support greater cloudification as a side effect of the intervention.
While for public administrations, the costs of porting or migration are one-off, several benefits
may accumulate over time. These include a reduced need to own and manage infrastructure,
greater scalability, faster deployment of new functionalities, more flexible cost structures,
improved service quality and easier access to advanced analytics and AI tools. Cloud migration is
often part of a broader modernisation effort: once the project is completed, the public authority
may be freed from managing some underlying IT infrastructure and may be able to add new
functionalities more easily. Some literature and case-study evidence point to total cost of
ownership savings in the range of 20–50% for cloudification projects, although these
estimates remain highly context-specificclxxvi.
At the same time, these benefits must be balanced against the potential price premium (mark-up)
of sovereign services. Any additional costs linked to providing sovereign services - compliance
costs linked to audits, EUCS certification costs, the use of EU-based infrastructure, higher labour
costs for EU-based staff - would be passed on to customers through higher prices.
6.2. Social impact
Employment. DCs are highly automated facilities, and the effect on employment is greater during
their construction phase than their operational phases. As reported by stakeholders, construction is
typically undertaken by large multi-national construction companies with participation of local
teams for the initial phases and non-local specialised teams for fit-out phases. For a 100 MW site,
around 1 500 – 2 000 construction employees would typically be on site per day. During the
operational phase, approximately 6-8 full time jobs are created for every 1 MW of capacity, with
economies of scale appearing as DCs measure beyond 50 MW. Some interviewees also
highlighted that they provide offices for some of their larger customers, converting DCs into co-
working hubs. When it comes to providing cloud and AI computing services, the EU’s shortage in
ICT skills can hamper the development of cloud and AI ecosystems, with a limited availability of
experts creating a significant gap in an already limited specialist workforce. For AI, only a small
proportion of specialists are actively involved in the industryclxxvii, with many remaining in
academia. This results in a low transferability of results to the market and, where transfer is
112 In lower market growth (250bn) and lower share by the public sector (10%), this would reach EUR 250 bn * 10% * 10% = EUR 2.5 bn by 2030.
In higher market growth (390bn) and higher share by the public sector (16%), this would reach EUR 390 bn * 16% * 10% = EUR 6.24 bn by 2030.
73
successful, companies lack the means to scale up or retain talent. Beyond these direct impacts,
cloud and AI computing services play an important role in the digitalisation of other sectors,
which has a complex effect on employmentclxxviii. With respect to the contribution of technological
sovereignty, 71% of citizens from the three biggest Member States believe that sovereignty can
help create and protect local jobsclxxix.
Citizens. With regards to DCs, there are documented instances of public resistance against their
construction due to environmental concerns, e.g. energy consumption and water security and
potential consequences on related pricesclxxx. However, citizens increasingly use digital services in
their daily livesclxxxi, where cloud and AI computing services and their underlying infrastructure
play a vital enabling role. AI, while nascent, is already pervasiveclxxxii. Citizens stand to benefit
from accelerated deployment of nearby compute capacity for low-latency applications like
automated driving, decentralised energy grids or assisted surgeries or living. These growing social
benefits should help increase the long-term public acceptance of the facilities that power these
technologies. Concerning sovereign cloud and AI computing services, a strong majority of citizens
from the three biggest Member States support greater sovereignty even if this increases costs and
argue that governments should lead the investment and development to achieve this objective. The
support drops, however, should this result into a lower spending on public servicesclxxxiii.
Some POs have particular social impacts. Under PO1-A, the public guidelines and the yearly
forum with the involvement of local municipalities, would help increase community trust and
engagement. Under PO1-B,the social acceptance of DCs is expected to be the greatest, as
decisions on DC build-out give a greater role to local authorities and are thus taken closer to
citizens. Acceptance can be expected to be particularly high when DCs demonstrably contribute to
local communities, for example through waste heat reused in local energy systems. Social
acceptance should also be a core attention point of the coordination hub which would disseminate
best practices on deployment. Conversely, PO1-C is expected to have a detrimental social impact
as EU-level decision making on DC deployment is perceived as further away from the citizen.
PM18, which belongs to PO2-B and is then picked up in PO2-C,would have direct social impacts
since it focusses on the development of ICT skills where Europe is lagging, resulting in higher
employability of citizens. For PO1-A/B/C, the possible effects on consumer electricity prices have
not been considered as this would require too many unverifiable assumptions (see next section).
6.3. Environmental impact
DCs in the EU currently are projected to consume around 99 TWh of electricity in 2025,
equivalent to roughly 3 % of total EU power generation. Over the next decade, rising capacity will
be the primary driver of increased electricity use, though its effect will differ depending on the
pace of expansion and the success of energy-efficiency. Electricity consumption and its associated
CO₂ emissions represent the dominant environmental impact of DCs and are therefore used as a
key proxy for quantifying the environmental impact of additional data centre capacity. Beyond
electricity consumption and associated CO₂ emissions, data centres generate environmental
impacts across several additional dimensions. Water consumption, embodied emissions in
equipment and buildings, refrigerant leakage, and local environmental externalities are also
relevant measures of environmental impact113.
The European Climate Law (ECL) consistency check was performed by comparing the projected
electricity use and associated CO2e emissions under each policy scenario with the EU’s declining
113 These impacts could not be systematically quantified at EU level due to lack of harmonised data. Looking ahead and considering the share of the emissions generated for DC construction materials (concrete, steel as well as other materials as semiconductors), the life cycle carbon footprint
approach should be another relevant method to measure the environmental impact of each DC project.
74
grid-emission trajectory toward a 55% GHG emissions reduction target by 2030 and climate-
neutrality by 2050. The first subsection presents the quantification of environmental impact in
terms of increased energy consumption and CO₂ emissions under the different policy options, with
considerations on the interaction between data centre deployment and impacts on the electricity
system. Implications of increased data centre capacity on water consumption are then discussed in
section 6.3.2.
6.3.1. Impact on electricity use and CO₂ emissions
Under the baseline scenario, total capacity is expected to expand to 46.3 GW114 in 2036 with an
expected modest PUE improvement from 1.29 down to 1.23115. Annual electricity use is expected
to rise from 99 TWh in 2025 to around 314 TWh in 2036, i.e. an average increase of 11% per
year. The cumulative electricity demand over 2025–2036 would amount to 2 571 TWh. Assuming
an evolution in the EU grid carbon intensity from 0.25 kg CO₂e/kWh today to 0.16 kg CO₂e/kWh in
2036clxxxiv, this would correspond to 50 Mt CO₂e in 2036. The environmental impact of PO1-
A/B/C is assessed against the increase in electricity demand (in TWh) and CO2 emissions resulting
from additional DC deployment under the different scenarios116.
Table 10. EU-27 data centre capacity in 2036: electricity and CO2 impacts vs baseline
Scenario 2036 Capacity
(GW)
PUE
2036
Electricity Use
2036 (TWh)
Cumulative 2025–
2036 (TWh)
CO₂e /GW
2036
(Mt/GW)
Cumulative CO₂e
2025–2036
(Mt)
Baseline 46.3 1.23 314 2 571 1.18 495
PO1-A 52.2 1.18 339 2 701 1.13 518
PO1-B 65.9 1.12 408 3 028 1.09 576
PO1-C 58.7 1.05 340 2 737 1.01 525
In PO1-A, the guidelines would encourage early integration of resource-efficient features, the
adoption of renewables and best practices in energy management, while the coordination forum
would facilitate dialogue with energy providers. These measures are expected to improve the PUE
of new DCs by around 4% by 2036 compared to the baseline, reaching 1.18. Total installed
capacity would increase to 52.2 GW and electricity demand would reach 339 TWh in 2036, or
about 8% higher than the baseline and cumulative consumption over 2025-2036 would total 2 701
TWh, or 130 TWh more than the baseline. Increased expansion of capacity under this option
would push up cumulative carbon emissions by 23 Mt with respect to the baseline. In PO1-B,
owing to fast-track area identification, Member States would have a stronger hand at linking DC
development to sustainable energy availability and avoiding environmentally sensitive sites. EU-
level discussions with national authorities are also expected to favour a fast uptake across Member
States of sustainability best practices in DC build-out. Altogether, this policy option is expected to
deliver an improvement in PUE of 9% by 2036. In this scenario, DC capacity is expected to
increase fastest, reaching 65.9 GW by 2036, while annual electricity use would reach 408 TWh by
2036, the highest among all scenarios. Cumulative 2025–2036 demand would exceed 3 000 TWh,
approximately 30% above the baseline. Even with continuous efficiency gains, the speed of
expansion would push up total energy and resource requirements. Cumulative CO₂ emissions are
expected to be 82 Mt higher than the baseline over the decade. However, average CO2e
emissions/GW decline 12% faster than in the baseline scenario due to the expected improvements
in PUE. With PO1-C, EU R&D funding (PM8) is expected to support the development and
114 Data centre capacity includes private sector, i.e. colocation and hyperscalers, and public sector capacity. 115 For additional information concerning expected PUE changed under the different Policy Options, please see Annex 4, Section 2.3.5. 116 The baseline assumes that electrical grid limitations remain an issue in certain primary markets such as Ireland and the Netherlands. However, it
considers that regulators and operators in several markets have begun investing in grid modernisation and demand management.
75
gradual uptake of sustainable innovations, especially after an initial phase of testing and
demonstration. These may include AI-driven energy management, advanced cooling systems,
solutions for waste heat reuse and renewable and storage integration. Strategic funding for
deployment (PM9) would also be used to incentivise projects incorporating highly sustainable
features, making eligibility conditional on sustainability criteria. These two measures focusing on
targeted grants or technical support will be designed to stimulate the development, testing and
market uptake of advanced solutions with the objective of considerably reducing PUE and WUE
levels over a 10-year horizon. This type of intervention could help smaller colocation providers
reduce PUE as they would aim to facilitate upfront R&D investments, pilots and CAPEX into
innovative technologies that small operators struggle to finance compared to hyperscalersclxxxv.
EU-level identification of fast-track areas (PM10) would ensure that facilities are deployed where
they are least environmentally damaging at continent level. Capacity under this option is expected
to reach 58.7 GW by 2036, with the strongest efficiency gains: PUE falls to 1.05 (–15% with
respect to the baseline). Despite 27% higher capacity than the baseline, total electricity use
remains close to PO1-A at 340 TWh, so is the cumulative consumption over 2025–2036. Due to
cleaner electricity sourcing, stronger PUE reductions and general forecasted decarbonisation of the
grid mix, average CO₂e emissions/GW are expected to decline 21% faster than in the baseline
scenario117. Across all scenarios, electricity use at least triples between 2025 and 2036. However,
with policy measures under PO1-C, stronger efficiency standards and renewable integration are
expected to contain DC power demand to around 13% of EU electricity generation.
Finally, under PO2-B and PO2-C, the cross-border re-use of cloud and AI capacity among
Member States through the federation would be inducive of environmental savings: a rough
estimation points to 5% energy efficiency improvement for 10% increase in server utilisation
rateclxxxvi.
The resulting emissions would increase across all scenarios due to growing DC electricity
demand. However, CO2e emissions per GW of added capacity would decline much faster under
PO1-C than under the baseline, owing to the stronger reduction in PUE and gradual adoption of
sustainable energy integration measures. The baseline pathway is not compatible with long-term
climate neutrality goals, as it leads to increased energy demand without integrating sustainability
requirements. Policy intervention is essential to uphold consistency with the ECL and ensure that
possible national DC acceleration policies do not result in a race-to-the-bottom in terms of
sustainability and minimise environmental impacts and grid strainclxxxvii.
As mentioned under section 2.2.2., grid capacity limitations are also real and increasing barriers
for the development of data centre capacity. The International Energy Agency highlighted that
waiting times for grid connection in key data centre hubs range from two to ten years due to
capacity limits and congestion, thus slowing down project development and diverting investment
to regions with available grid capacityclxxxviii. In this context, it is important to note the positive
contribution which DCs – if properly leveraged – can make to grid stability, notably by offering
flexibility services118. The co-location of energy generation on a DC site can help reduce grid
stress and DC investments have the potential to also boost investments in grid infrastructure.
6.3.2. Impact on water consumption
Data centres are also becoming a significant and rapidly growing source of water consumption,
with implications for regional water security and climate resilience. Depending on the cooling
117 Carbon footprint is mostly driven by scope 2 energy consumption. To reduce climate impact, low-carbon energy use should be incentivised. 118 Notably, due to their storage capabilities, data centres present a unique opportunity to enhance power system flexibility. See for example: Data
centres as a source of flexibility for power systems - ScienceDirect.
76
technology used, their cooling systems can require large volumes of freshwater, often millions of
litres per day for hyperscale facilities, placing pressure on local supplies, especially in drought-
prone or water-stressed areas. As demand for cloud services and AI model training accelerates,
water use is expected to rise unless mitigated through efficiency measures, site-specific resource
planning, and low-water or water-free cooling technologies. Estimates suggest that data centres
across the EU consumed around 76 bn litres of water in 2025. At global level, the International
Energy Agency estimates that water consumption by data centres amounts to approximately 560
bn litres per yearclxxxix, implying that the EU-27 accounts for around 14% of global data centre
water consumption. When assessed against broader metrics related to water use, data centres
represent a relatively small share of overall water pressure. Total freshwater abstraction in the EU
has decreased by 19% between 2000 and 2022, corresponding to a compound annual growth rate
of -0.8%, according to the European Environment Agencycxc. Extrapolating this rate to 2025
suggests that, out of an estimated 192,000 million m³ of freshwater abstracted annually in the EU-
27, data centres accounted for 0.04%, i.e. remaining well below the levels observed in
international counterparts. As data centre capacity is expected to grow in the next years, overall
water consumption by data centres is also foreseen to increase. However, this trend could be
alleviated by an industry shift towards more sustainable and resource-efficient operations. The
adoption of advanced cooling technologies and improved practices is expected to consistently
reduce water intensity over time. A reduction in Water Usage Effectiveness (WUE) from current
levels of 1.8 to 0.6 litres per kWh over the next decade, in line with the Climate Neutral Data
Centre Pact’s target of WUE levels below 0.4 litres per kWh by 2040, and ongoing efficiency
improvements by hyperscalers, could substantially limit water demand. Under these assumptions,
water consumption growth could be reduced by around 8 percentage points in CAGR, resulting in
total water consumption that is approximately half of what would be observed in the absence of
comparable efficiency oriented measures.
Against the sector’s continued growth, measures that incentivise resource-efficient investments,
technological improvements and a balanced distribution of new data centre capacity across
Member States, with particular attention to water-stressed regions, are key to reduce the
environmental impact also in terms of water consumption. As noted above, overseeing the water
use of data centres is crucial not only for environmental sustainability but also for social equity
and economic stability. Given the local nature of water systems, the increase in data centre
capacity can impact water supply reliability and influence drought resilience in specific areas. The
analysis highlights the need for governance mechanisms that ensure new infrastructure aligns with
sustainable water management, transparent reporting, and equitable access to shared water
resources.
6.3.3. Other key metrics concerning environmental footprint
Research has shown that focusing uniquely on electricity, CO2 and water usage overlooks key
environmental impact categories linked to building new data centre capacity, in particular as grids
are expected to become more sustainable in the future. Studies have highlighted the importance of
embodied impacts, i.e. environmental impacts associated with construction materials, mechanical
and electrical equipment and IT hardware maintenance, which can become a relevant portion of
the overall data centre footprint over timecxci. A recent peer-reviewed paper similarly highlights
the need for comprehensive life cycle assessments of cloud infrastructure to measure
environmental impactcxcii. Several equipment and IT hardware pieces needed in data centres,
especially if in constant use, may need replacing upon reaching end-of-life. The proper collection
and treatment of these waste electrical and electronic equipment (WEEE) from data centres is
77
necessary to protect human health and the environment, in particular as regards depollution and
hazardous substances, as well as the recovery of materials from recycling119.
Another environmental factor in DC development remains sustainable cooling. While powering
the cooling is an integral part of overall energy consumption, using low global warming (GWP)
potential refrigerants remains important. Refrigerants used in cooling systems can pose significant
climate challenges due to leaks of high-global warming potential fluorinated gases (F-gases). The
European Environment Agency outlines why fluorinated greenhouse gases are a key focus for
mitigation and tracks EU actions in this area120. Moreover, at EU level, the recast F-gas regulation
combined with the ICT Taxonomy provide guidance on the types of sustainable refrigerants to be
used for cooling EU DCs121.
Lastly, environmental assessments and Scope-3 frameworkscxciii bring attention to other impactful
areas, e.g. supply-chain emissions and waste from equipment and infrastructure (Scope 3),
emphasising the need to consider other parameters beyond electricity usage for accurately
monitoring the environmental footprint of data centres in the future.
7. HOW DO THE OPTIONS COMPARE?
This chapter evaluates the policy options in terms of their effectiveness, efficiency, coherence,
subsidiarity and proportionality. It brings together the results of the preceding impact analysis to
examine how each option performs against these criteria and highlight their relative strengths and
weaknesses. It then presents the results of the sensitivity analysis conducted, using best- and
worst-case scenarios to derive confidence intervals for testing the robustness of the cost-benefit
results under varying assumptions. The final section then focuses on the comparison of the
options, presenting their overall performance across the different criteria.
7.1. Effectiveness
The analysis of effectiveness examines the extent to which the policy options under consideration
are expected to contribute to the achievement of the general and specific objectives of this
initiative. As outlined in section 5.2., the first set of options primarily addresses SO1 and SO2,
while the second set has been designed to better address SO3 and SO4. The table below illustrates
the relationship between the policy objectives and the assessment criteria, which served as initial
benchmarks for evaluating the potential effectiveness of the options.
The assessment draws primarily on evidence gathered through literature review and desk research,
complemented by interviews, the CATI survey, and validation workshops. While this section
presents the analysis mostly in qualitative terms, the same evidence base, including survey results,
was integrated into a multi-criteria decision analysis (see Annex 4, section 6) to evaluate the
effectiveness of the proposed measures, alongside other assessment criteria.
Table 11. Links between objectives and assessment criteria
General objective Specific objectives Assessment criteria
Ensure the
functioning of the
internal market
SO1 – Increase computing
capacity in the EU through
innovative and sustainable
• Expected increase in EU installed computing capacity
(MW)
• Expected improved PUE of new data centres
119 As announced in the Clean Industrial Deal (CID) Communication adopted 26.02.2025, the Commission is preparing a review of the WEEE
Directive as a pillar of the upcoming Circular Economy Act (CEA) proposal intended for later in 2026, in particular with a view to improve areas identified in the Evaluation of the WEEE Directive published 02.07.2025, including WEEE collection, treatment of WEEE, and recovery of critical
raw materials (CRMs) embedded in various WEEE.120 See: EU progress under the hydrofluorocarbon phase out set out in the EU F-gas Regulation | Hydrofluorocarbon phase out in Europe | European Environment Agency (EEA) 121 See Regulation (EU) 2024/573: https://eur-lex.europa.eu/legal-content/EN/ALL/?uri=CELEX:32024R0573
78
General objective Specific objectives Assessment criteria
for cloud and AI
computing
services and
secure the
conditions
necessary for the
competitiveness of
the Union’s
industry and the
resilience of its
public sector as
users of such
services
technologies • Increase in the number of pilots launched and innovative
technologies, e.g. immersion cooling for energy efficiency,
adopted in new data centre projects.
SO2 – Ensure attractive
conditions for the deployment
of sustainable and innovative
computing capacity
• Expected reduction in the time needed for data centre
permitting and total administrative proceedings.
• Expected reduction in underutilised land (share of identified
viable sites used).
• Share of new capacity deployed outside existing
hubs/underserved areas.
• Additional investment in data centre capacity driven by
better investment conditions for data centre projects.
SO3 – Decrease the overall
reliance on non-European
cloud and AI computing
services
• Expected increase in the market share of European cloud
and AI computing service providers serving highly critical
use cases
• Expected reduction in dependency concentration on non-
EU providers
SO4 – Enhance the resilience
of supply of cloud and AI
computing services, in
particular in the public sector
• Expected increase in the number of public sector contracts
supplied by providers meeting the sovereign services at
level 2, 3 or 4 scored under level 2, level 3 or 4
• Expected increase in the (re-)use and sharing of open source
solutions
• Expected increase of federated resources across the public
sector and use of the joint procurement of cloud and AI
services
Regarding SO1 (increase computing capacity in the EU through innovative and sustainable
technologies), PO1-A would provide a limited effect. An expanded working group and structured
forum are expected to reduce coordination and information frictions mainly linked to inconsistent
permitting and zoning procedures across Member States. The guidelines focused on sustainability
would be designed to translate “innovative and sustainable” technologies into best practices for
data centres. These would aim to complement the existing reporting obligations under the EED
with concrete instruments to improve compliance with sustainability requirements and allow an
earlier integration of efficiency-related considerations in project design. At the same time, being
voluntary instruments based on ad hoc participation or adoption by businesses, these measures are
expected to indirectly and slightly increase computing capacity in the EU and reduce PUE with
respect to the status quo scenario. While the option is expected to improve the quality of new
capacity, its concrete contribution to additional MW or innovative technologies for energy-
efficiency is expected to be moderate. PO1-B is expected to be more effective at increasing EU-
based computing capacity, with the strongest direct impact on additional data centre deployment.
This is expected as the result of more favourable growth conditions, driven by simplification
measures, faster permitting and potential public support by Member States. The national facilitator
(PM4) and nationally defined strategies, including fast-track areas (PM5) are expected to reduce
permitting timelines and time needed for data centres to connect to the grid, while ensuring that
projects also locate where grid capacity, land and water constraints make sustainable expansion of
capacity technically feasible. This would foster lower PUE levels across the EU as access to the
zones would also be linked to sustainability performance. Without this linkage, adverse
consequences linked to excess grid stress or carbon-intensive expansion would undermine SO1.
The regular monitoring of compute capacity and national strategies would ensure that the
intervention is effective in closing the demand-supply gap, without under or over supplying
capacity across the EU. PO1-C is expected to be less effective than PO1-B, due to the addition of
79
a decision step at EU-level over what is currently decided at national or regional level. PM10
consolidates the 27 national permitting and fast-tracking systems for data centre deployment into a
single EU-level decision framework. This centralisation is expected to create different types of
inefficiencies for data centre deployment. The first one would be related to the number of projects
that would be processed in parallel. While under PO1-B access to fast-track areas can be
authorised in parallel across MS, under PO1-C projects must pass through a common EU-level
decision layer, which is likely to limit the throughput of projects that can be supported. The lower
effectiveness of this option was also strongly raised by stakeholders during workshops and
interviews, where several operators and public authorities expressed their concerns that an EU-
coordinated process would slow down procedures rather than accelerate deployment. They
specifically mentioned that top-level identification of fast-track areas for projects would be less
able to reflect geographical and local specificities, e.g. related to grid conditions or community
impacts. While under PO1-B,the social acceptance of DCs is expected to be the greatest, as
decisions on build-out would give a greater role to local authorities and are thus taken closer to
citizens, EU-level decision making on DC deployment will likely be perceived as taken far away
from citizens. This option is expected to be mostly effective in promoting sustainable technologies
through EU R&D funding (PM8) and EU deployment funding (PM9). These measures are
expected to improve technology readiness, e.g. in advanced cooling or energy management, while
also de-risking first-of-a-kind projects of a strategic interest. Therefore, this option is expected to
perform better than the other options and the baseline on long-term sustainable computing
capacity, while below them in terms of short to medium-term delivery of additional data centres.
All POs dealing with the dependence on cloud and AI computing services provided by non-EU
providers (PO2-A, PO2-B and PO2-C) are expected to have an indirect impact on DC capacity in
the EU, particularly PO2-C as one of the criteria to identify sovereign services across all levels
includes the need for infrastructure to be located in the EU. Therefore, if the demand for sovereign
services increases, the infrastructure in the EU will also have to increase.
Looking at SO2 (ensure attractive conditions for the deployment of sustainable and
innovative computing capacity), PO1-A would provide a relatively low-cost way to address the
bottlenecks that affect DC build-out in the EU. However, it is expected to have a limited effect on
addressing permitting delays, access to resources and capital for strategic projects. Under this
scenario, attractive conditions for data centre deployment are built through structured dialogue and
guidelines, which contribute to reduce uncertainty for businesses and investors during site
selection and project design. The reduced transaction costs allow operators to face fewer project
redesigns, while authorities also rely on shared benchmarks. This is expected to somehow reduce
permitting timelines by a few weeks and lower regulatory risks but would not consistently change
the economics of new data centre projects. Similarly, this option is expected to lead to
improvements in established markets, without unlocking significant capacity deployment in other
regions or alternative locations, other than through natural market dynamics. PO1-B shows a
significant potential to reduce bottlenecks for DC buildout, thus ensuring more attractive
investment conditions. The introduction of fast-track areas would address inefficient procedures,
uncertainty in approval processes and delays that operators cite as critical barriers. The systematic
mapping of suitable sites and removal of zoning uncertainty are expected to make land
investment-ready and reduce underutilised yet suitable sites. In parallel, the creation of a project
facilitator aims to accelerate administrative processes and is expected to decrease the time for
80
building a new DC facility by at least 6 months122. This improved timeline reliability is translated
as higher project Net Present Value and Internal Rates of Return (IRR) for investors, with an
expected increase in private capital mobilisation. Under this option, projects’ IRR would rise by
over 80 basis points with respect to the baseline, increasing from 9.86% to 10.70%. This shift is
expected to enhance investor appetite, as confirmed by interviews with investors, who highlighted
that this shift in IRR can reposition assets within investor target bands, improving competitiveness
against alternative infrastructure investments and affecting bankability123. This is particularly
relevant for mid-sized European data centres (typically 5–25 MW), which are aligned with
institutional investment ticket sizes but face tighter margins and higher relative development risk
than hyperscaler projects. Where needed, national funding support (PM6) can help lower capital
and operating costs, improving project bankability and investment decisions, especially for
smaller providers, thus lowering their entry barriers in the market. Sweden has provided a useful
example of how tax incentives could be poorly managed if not linked to specific energy objectives
and subsequent evaluationcxciv. Monitoring deployment is also expected to increase policy
credibility and transparent information for investors. By targeting the drivers of geographic
clustering (i.e. divergent frameworks, path dependency, uninternalized externalities) and lowering
entry barriers in underutilised regions, this option is expected to unlock investment in secondary
markets, e.g. Italy or Portugal, and in developing regions, e.g. Bulgaria124, leading to a rebalancing
of capacity across the EU. PO1-C, although effectiveat addressing fragmentation and promoting
harmonisation through EU-level coordination, is expected to be less effective in achieving relevant
reductions in permitting duration and financing conditions of projects. On one hand, R&D funding
and deployment funding would be a relevant political signal towards the development of
sustainable digital infrastructure. EU-level coordination would also help minimise regulatory
hurdles for cross-border investors. With respect to promoting a more geographically balanced
distribution of capacity and increase land optimisation, EU-level decision making is, in principle,
best placed to steer deployment towards an optimal territorial allocation of computing
infrastructure. However, its effectiveness in reaching this objective is contingent on local
engagement, which may be harder to secure than under PO1-B. Industry consultations highlighted
that the creation of additional governance layers could slow down decision-making processes,
especially if still requiring national or local review. Thus, this option is expected to be less
effective than PO1-B in addressing barriers to deployment.
For SO3 (decrease the overall reliance on non-European cloud and AI computing services),
PO2-A is expected to have limited effectiveness relative to the baseline. Harmonised criteria,
guidelines and a dedicated conference, are expected to increase clarity and promote a more
coherent understanding of the concept of sovereignty. This may reduce information asymmetries
and improve comparability of service offerings, especially in public procurement. However, these
measures are unlikely, on their own, to shift demand away from non-European cloud and AI
computing services. In the absence of clear and robust assessment mechanisms to verify the
122 This has been confirmed through interviews with stakeholders and validated during a final workshop in the context of the supporting study. The
impact has been assessed on the basis of an overall 18-month permitting period for data centre deployment. However, any reduction in this timeline, including for example through the introduction of as 12-month deadline would further increase Europe’s attractiveness as a location for data centre
investment 123 Interviews with investors confirmed that data centre investments are evaluated across a wide risk-return spectrum, broadly consistent with
market benchmarks positioning core infrastructure targeting around 7-9% IRR and core-plus assets around 10-13%. Expected returns depend
strongly on asset maturity and risk profile: stabilised, fully built platforms with secured power and long-term contracts are treated as lower-risk assets, while development-stage projects face higher execution risk related to permitting, power availability, equipment procurement, and
commercialisation. Equity investors typically target “single digit returns plus a risk premium,” with materially higher expectations for projects
exposed to development, energy, or commercial risk. 124 During stakeholder interviews, and as mentioned above, Bulgaria was identified as lacking a clear classification of data centres in its permitting
system, creating long delays and deterring investment in the country.
81
harmonised criteria, sovereignty claims would remain self-declared and the notion of sovereignty
risks remaining a marketing tool with little or no effect in trust building. Ensuring the
effectiveness of the interoperability provisions of the Data Act would contribute to creating
opportunities for EU providers to build integrated offers. However, the Data Act’s effects remain
to be seen, given its recent adoption. The main consequences of this option would reside in the
gradual standardisation of the services, coupled with measures to promote interoperability, rather
than a structural change in procurement or deployment choices, which would alter existing market
dynamics. PO2-B is expected to be moderately effective in meeting this objective. A voluntary
sovereignty framework combined with voluntary award criteria that give weights to factors such
as the resilience of the EU cloud supply chain are expected to strengthen trust in cloud and AI
computing technologies, especially for the public sector. Through spillovers, parts of the private
sector could also be impacted, notably in sectors where sovereignty considerations are
strategically important. By enabling the federation of computing capacity, as opposed to procuring
from external providers, the option would produce more concrete substitution effects than PO2-A
or the baseline. The option is expected to broaden the range of eligible procurement models for
public authorities, e.g. in terms of tender design, vendor evaluation, without mandating a universal
switch to sovereign solutions. The vendor-neutral cloud and AI computing services training
programme is also expected to reduce the reliance on few, vendor-specific training and
certification programmes. The overall effectiveness of the option in decreasing the reliance on
non-European cloud and AI computing services would depend on authorities’ administrative
capacity, procurement design and consequent market responsiveness. It could reduce some
exposure to third country dependencies if contracting authorities choose services with stronger EU
control features. This PO’s weakness lies in the voluntary nature of the proposed measures, which
could result in uneven adoption across Member States. PO2-C is expected to be the most effective
option in reducing the reliance on non-EU cloud and AI services. The combination of joint
procurement mechanism, the mandatory sovereignty risk assessment and mandatory award
criteria, along with the promotion of open source solutions is expected to have more substantial
impact on market outcomes compared to other options and the baseline. Unlike softer or voluntary
measures, this package embeds an approach to sovereignty into market access and purchasing
decisions, going beyond mere clarification or incentives. First, the mandatory sovereignty risk
assessment for procuring cloud and AI services plays a crucial role in embedding sovereignty and
dependency-related considerations into procurement decisions. This measure ensures that
contracting authorities procure services considering the sovereignty implications of decisions,
leading to more informed choices. By bringing clarity with respect to the use cases for which
cloud and AI computing services shall be procured under specific sovereign levels, the approach
contributes to reducing the reliance on non-EU cloud and AI services. In fact, levels 3-4, which
are estimated to cover 10% of the public sector’s needs, would need to be served by EU providers
to address and ensure the protection of public order in the public sector. Most procurement cases
would fall under intermediate levels of sovereignty, where non-EU providers can participate.
However, EU providers are likely to face fewer difficulties in complying with the sovereignty
requirements, giving them a competitive edge (the reasoning for these numbers is presented under
the descriptions of policy measures PM15 and PM21 in section 5.2.2, and the coverage of the
necessary range of services by EU providers in section 2.3.1). Moreover, with a view to reducing
critical dependencies, the non-price award criteria would allow public authorities to procure
cloud and AI computing services with a higher level of local added value. The overall monitoring
framework managed by the Commission would support Member States in assessing the market
presence of EU providers. By advancing an EU-coordinated procurement and support framework
for sovereign services, the option is also expected to increase trust in cloud and AI computing
82
services and accelerate the cloudification of on-premises solutions of the most critical use cases,
while creating market opportunities for sovereign European cloud providers. This approach is also
expected to create broader spillover effects beyond the public sector. Notably, the mandatory
nature of the sovereignty framework as well as the proposed extended risk assessment for essential
entities of the private sector listed under Annex I NIS 2 to address sovereignty-related risks are
expected to create a wider spillover effect compared to that of PO2-B. The main challenge that
could hinder the adoption of sovereign services in the private sector could be a “sovereignty-
premium price” whose range remains subject to be established (see discussion under section
2.3.4). However, several companies in critical sectors have already publicly manifested their
interest in those solutions and are moving towards that direction125. This would be amplified
through the SME support scheme fostering cloud and AI adoption among SMEs, and the creation
of toolbox for the integration of EU providers solutions, which is expected to increase the
visibility of EU sovereign services. Building on this, the joint procurement mechanism allows
public sector organisations to pool their purchasing power, creating a larger and more stable
market for inter alia EU providers. This mechanism is also expected to attract more investment,
reduce costs and improve the overall competitiveness of EU providers, making them comparable
alternatives to non-EU providers126. Moreover, the promotion of open source solutions within
public administrations would further reinforce this objective by improving fairness and
transparency, reducing vendor lock-in, stimulating competition among providers, and opening the
market to alternative solutions and to new entrants. This measure encourages innovation and local
development, creating opportunities for European companies to thrive in the cloud and AI market,
and contributing to the growth of the open source services sector in Europe. Overall, the measures
included in policy option PO2-C work together to reduce the reliance on non-EU cloud and AI
services, promote European sovereignty, and foster a robust and competitive European cloud and
AI ecosystem. By addressing the risks associated with dependency on non-EU providers with a
granular, layered sovereignty framework and promoting the development of European solutions,
this option is expected to have a lasting impact on the market, creating a more sustainable and
resilient foundation for the public sector's digital transformation.
For SO4 (enhance the resilience of supply of cloud and AI computing services, in particular
in the public sector), PO2-A is expected to have a very limited effect due to the soft nature of the
measures. While common criteria and guidance may improve awareness of relevant risks and
encourage the use of more coherent definitions by public authorities, they would not materially
increase switching capacity, supply redundancy or operational continuity in the event of
disruption. The option thus addresses only part of the information problem but not the structural
sources of fragility and trust. This would result in having a marginal effect on reducing exposure
to service disruption, or non-EU dependencies, especially in the public sector. PO2-B is expected
to be moderately effective in improving the resilience of supply of cloud and AI computing
services. The federation would contribute to ensuring that specific use cases are served by services
with increased EU control. Federation is especially relevant from a resilience perspective as it
supports the diversification of dependencies, and reduces the risks associated with reliance on a
single external provider. The training programme on vendor–agnostic technologies and voluntary
award criteria are also expected to expand the market for resilient and locally managed cloud and
AI computing services. Nonetheless, as mentioned above, the voluntary implementation of these
measures could limit their consistent implementation across the EU. This would result in uneven
125 See for instance Airbus who recently announced their activities to migrate their critical workloads to a European cloud provider. 126 The outcome of the recent DIGIT tender for cloud services is testament to this. Commission advances cloud sovereignty through strategic
procurement
83
resilience gains: authorities with higher procurement maturity would likely benefit more, while
others may not implement the framework in a way that changes their risk exposure. PO2-C is
expected to have the biggest effect on resilience. The establishment of an actionable and
comprehensive sovereignty assessment framework provides national authorities with a more
robust mechanism to procure and uptake cloud and AI computing services, protecting critical use
cases where public order is at stake. The framework allows for the diversification of dependencies,
and services from non-European providers would still be able to qualify for 90% of the public
sector’s needs. The use of harmonised public procurement award criteria would also contribute to
the public sector’s resilience for specific services since procured services would have a higher EU
added value. Moreover, the joint procurement mechanism is expected to simplify the procurement
of cloud and AI computing services and allow public authorities – especially the smaller ones – to
achieve better commercial and contractual terms. Finally, the promotion of open source
technologies would further strengthen resilience by reducing dependencies on proprietary systems.
Altogether, these measures would directly reduce exposure to non-EU dependencies and single
provider concentration risks, strengthening the resilience and autonomy of cloud and AI services,
particularly in the public sector.
With respect to the general objective of promoting competitiveness while strengthening
strategic autonomy, the options differ in effectiveness and in the balance they strike between the
two dimensions. In terms of promoting competitiveness, the first set of options differs in the
expected ability to accelerate data centre deployment and reduce investment barriers. PO1-A
would have a limited but positive effect by improving coordination and sharing good practices,
although its voluntary nature means that it may not significantly reduce permitting, infrastructure
or investment bottlenecks. PO1-B would be the most effective option, as national-level legislative
and financial measures could be tailored to local conditions, including designated areas, fast-track
procedures, public support for projects and capacity monitoring. This would directly support faster
deployment and investment certainty. PO1-C could also support competitiveness through EU-
level funding and fast-tracking, but may be less responsive to local permitting, energy and land-
use constraints. Strategic autonomy would be strengthened under all options to the extent that
additional capacity is built in the EU, with the greatest practical effect expected under PO1-B.
PO2-A is expected to make a modest contribution to competitiveness by reducing information
frictions and clarifying some notions, but it is unlikely to improve the competitive position of EU-
based providers or to alter the dependency structure of the market. Its contribution to strategic
autonomy would therefore also be limited. In practice, it would preserve the status quo in market
structure while improving transparency. As a result, it is unlikely to achieve the general objective.
PO2-B is based on a more comprehensive set of measures that may impact competitiveness and
strategic autonomy. Its contribution to competitiveness would come from stronger demand for
sovereignty-audited services, greater contestability through federation and interoperability, and
reduced lock-in. Its contribution to strategic autonomy would come from lower dependency in
sensitive use cases and a gradual expansion of EU-controlled capacity. However, the option may
not be strong enough to overcome incumbent scale advantages, so both competitiveness and
autonomy gains may remain incomplete. PO2-C is likely to deliver the strongest gains in strategic
autonomy. It is the option most likely to expand the market position, scale and credibility of
sovereignty-qualified providers, and therefore the most likely to change the structure of the market
in favour of strategic autonomy. Over time, this could support competitiveness if stronger EU
demand fosters scale, innovation capacity, local value creation, and broader ecosystems around
interoperable and open solutions. In terms of achieving autonomy, this is the option which most
effectively prevents unlawful access to European data through non-European laws that have an
extraterritorial reach, but in a proportionate way. In fact, MS can decide which use cases should
84
fall under higher levels of sovereignty assurance, based on their considerations. To support the
Single Market, the Commission would issue EU-level guidance for MS to conduct their risk
sovereignty assessments. While not ensuring a unique outcome, this would be a signal for
authorities to converge towards a unique approach. Doing this would allow providers to operate
more easily across the EU, thanks to common criteria and approaches to which services can be
served under different sovereignty levels. Competitiveness is also preserved because the
sovereignty framework is devised in a way to better align MS with the available offer from
European operators thanks to the market monitoring reports provided by the Commission. These
would allow Member States to integrate a market reality check in their risk assessments, avoiding
situations such as tendering at Level 3 where competition is limited or no solution exists. These
market monitoring reports would also act as a powerful guidance for EU providers to prioritise the
development of their offering, thus incentivising the emergence of a competitive European
alternative. In terms of long run competitiveness, the use of clear sovereignty levels for different
use cases based on a risk sovereignty assessment is also expected to reinforce trust and the uptake
of cloud and AI technology across the EU economy and society127.
The overall effectiveness assessment with respect to each specific objective is presented below
using a symbolic scoring, using ranking indicators that go from “o” no relevance to “very
effective”, with respect to the baseline.
Table 12. Effectiveness of the Policy Options against the Specific Objectives
SO1: Increased
computing capacity
through innovative
and sustainable
technologies
SO2: Attractive
conditions for the
deployment of
computing capacity
SO3: Reduced
reliance on non-EU
providers
SO4: Enhanced
resilience of supply of
cloud and AI
computing services
Policy option 1A + ++ o o
Policy option 1B +++ +++ o o
Policy option 1C ++ ++ o o
Policy option 2A + o + +
Policy option 2B + o ++ ++
Policy option 2C ++ o +++ +++
Legend: o no relevance; + limited effectiveness; ++ effective; +++ very effective
7.2. Efficiency
The table below presents a qualitative summary of costs and benefits of the Policy Options borne
by the main stakeholder groups analysed under section 6.1., i.e. data centre operators, cloud and
AI service providers, essential entities of the private sector, public authorities and the European
Commission, mostly calculated using the standard cost model and net present value framework128.
Given the uncertainty around the estimates, the results are presented using “+” and “-”, which
reflect indicative ranges rather than precise values. The objective is to present the likely direction
and scale of each option compared to the baseline. Positive signs indicate benefits, while negative
ones indicate costs, while the number reflects the order of magnitude of the quantified impact. The
127 This issue was recently illustrated by the postponement of Finland’s electoral management system cloudification, previously awarded to AWS, and France’s switch from US solutions to open source equivalents to serve public sector video conferencing needs are illustrations of this broad
trend. 128 The table does not include the impact on SMEs, which is discussed only qualitatively in the previous section. Additional information on quantified costs and benefits for SMEs under PM23 can be found in Annex 4 and are reported in the efficiency overview of the preferred package
below.
85
table does not incorporate non-market social and environmental externalities. These effects are
discussed in the respective sections 6.2, 6.3 and further below as a relevant part of this impact
assessment.
Table 13. Summary of costs and benefits of the policy options – indicative ranges based on central estimates of
NPV for 2027-2036 and qualitative estimates, compared to the baseline
Difference to the Baseline
PO1-A PO1-B PO1-C PO2-A PO2-B PO2-C
Data centre operators
Benefits + +++ +++ N/A N/A N/A
Costs − − − N/A N/A N/A
Cloud & AI service
providers
Benefits N/A N/A N/A + ++ ++
Costs N/A N/A N/A − − −−
Essential entities of the
private sector
Costs N/A N/A N/A N/A N/A −−
Public authorities
Benefits + ++ + N/A ++ ++++
Costs − −− − − −− −−
European Commission
Costs − − − − −− −−
Wider economic effects + ++++ +++ + ++++ ++++
Total benefits + +++ +++ + +++ ++++
Total costs − −− −− − −− −−
Net benefits − +++ +++ − +++ ++++
Legend: N/A not applicable; + small benefit <€100m; ++ moderate benefit €100m-5bn; +++ large benefit €5-20bn;
++++ very large benefit >€20bn; - small cost <€ 100m; -- moderate cost €100m-5bn; --- large cost €5-20bn; ----
very large cost >€20bn
Overall, all the options except for PO-2A and PO1-A are expected to generate positive net benefits
relative to the baseline. For these options the expected benefits outweigh the expected costs. PO2-
C is expected to generate the highest net benefits among all options. Although it also entails
higher quantified costs, these remain lower than the expected benefits. PO1-B, PO1-C and PO2-B
also show strong efficiency, with large, expected benefits relative to the expected costs.
7.3. Coherence
In terms of external coherence, the proposed POs are consistent with existing initiatives. They
seek to close remaining gaps with respect to reaching the objectives. See annex 7 for details.
PO1-A/B/C complement and leverage other initiatives: PO1-A expands the existing Alliance on
Industrial Data, Edge and Cloud. PO1-B/C builds on the future Regulation on accelerating and
streamlining environmental assessments, which will simplify and speed up environmental
screenings and assessments. It allows for sectoral legislation to reference a toolbox with additional
favourable provisions for strategic sectors or categories, which CADA will do for DC projects
built in acceleration areas129. These additional DC-specific support measures are necessary to
rapidly close the capacity gap and can only be delivered in a dedicated instrument. Similarly, PO1-
B/C uses the rating scheme for DC sustainability under the EED to identify which DCs are
sustainable and can benefit from acceleration measures. In the same vein, PO1-B/C complements
129 PO1B/C takes the creation of a single point of contact for environmental assessments for granted and complements it with a facilitator (PM4) who would accompany the DC operator in this and other administrative stages (environmental assessments but also other permitting requirements
related to zoning, land allocation and building).
86
the Grids package by ensuring DC location considers grid availability, information is exchanged
sufficiently in advance to feed into grid planning and hence ensure timely connection of DCs.
Finally, none of the PO1s will overlap with the Industrial Accelerator Act which focusses on the
industrial manufacturing sector to which DCs do not belong.
PO2-B/C will complement the Public Procurement Directives 2014/23, 2014/24 and 2014/25 by
ensuring that the public procurement of cloud and AI computing services is covered by a specific
sectoral approach which caters for its specificities: For example, for reasons of strategic
autonomy, it is of utmost importance to ensure that highly critical public sector use cases of cloud
and AI computing services rely on services which are shielded from third-country data access and
possible interference with service continuity (PO2-B/C). This is different from a blanket ‘Buy
European’ approach, which may be considered as part of the ongoing revision of the Public
Procurement Directives. Due to today’s strong reliance on non-European CSPs, a blanket ‘Buy
European’ approach would not be suitable for public procurement. Instead, for reasons of public
order protection, PO2-B/C put forward ways of mitigating incrementally existing dependencies
through award criteria rewarding, for example, the integration hard- or software from outside the
countries of dependencies. On the side of contracting authorities, this approach implies that they
will have to comply with the horizontal requirements laid down in the Public Procurement
Directives as well as the lex specialis of CADA, especially under PO2-C. In view of the ongoing
review of the Public Procurement Directives, the preparatory work on the impact assessment
supporting the formulation of a Public Procurement Act includes a framework regulating the use
of ‘EU preference’ provisions in sectorial legislations. The measures foreseen in the impact
assessment supporting the Public Procurement Act would underpin the approach taken in CADA
both in terms of tools (award criteria) and justification (demonstrable relation to public order and
proportionate approach to risk mitigation). To prepare the grounds for a faster uptake of cloud
services by the public sector, CADA will be accompanied by a Recommendation on a single EU-
wide cloud and AI policy for public administrations and public procurementtranslating the
Regulation’s approach to public procurement into ready-made tender specifications.
While the Data Act opens the path to a possible reduction of dependencies on non-EU providers
by enabling switching, it does not directly incentivise the development of more sovereign cloud
and AI computing services, something that PO2-A/B/C will do with different intensities. These
options indirectly build on the Data Act’s right to switch and multi-cloud, which all cloud service
providers must enable and which all cloud users, including public administrations, can make use
of, for example to switch to a sovereign service. In laying down and using a harmonized criteria
for sovereign cloud and AI computing services, PO2-A/B/C build on an existing obligation of
cloud service providers under the Data Act: To take technical, organisational and legal measures
to prevent international and third-country governmental access and transfer of non-personal data
held in the Union, outside of recognised international law enforcement cooperation, where this
would conflict with Union or national law. The harmonized criteria for a sovereign service
employed across PO2-A, B and C goes further by also capturing immunity to third-country
policies affecting service continuity. This initiative thus does not create a new compliance burden
for providers but rather enables those providers wishing for their services to reach a sovereignty
label to build on existing compliance work under the Data Act. The AI Act already sets
requirements for AI systems and general-purpose AI models ensuring a high-level protection of
safety, health and fundamental rights, thus harmonising rules for the internal market. As detailed
in annex 8, PO2-A/B/C and more generally CADA, focusses on cloud and AI computing services
exclusively.
The Commission proposal for the Cybersecurity Act 2 empowers the Commission to impose
prohibitions and mitigation measures by means of adopting implementing acts (1) prohibiting
87
specific types of NIS2 entities from using, installing or integrating ICT components from high-risk
suppliers in key ICT assets; and (2) requiring such entities to apply targeted ICT supply chain
mitigation measures, including supplier transparency obligations, restrictions on transfers or
remote processing from third countries, third-party-audited technical safeguards, limits on
outsourcing or supplier contracting, requirements for personnel vetting by national authorities, or
diversification of supply. CADA relies on the CSA 2 for the exclusion of high-risk cloud and AI
vendors from EU critical sectors. At the same time, CADA goes further than the CSA 2 and
introduces a targeted approach to mitigating sovereignty risks specific to the provision and use of
cloud and AI computing services. In defining what constitutes a sovereign service under PM11,
PM15 and PM21, CADA leverages the future cybersecurity certification scheme for cloud
services (EUCS), where the sovereignty levels under CADA require a gradual conformity against
the EUCS assurance levels. There is no overlap between the requirements of the CADA
sovereignty levels and the applicable requirements and those assessed under the EUCS, which
covers exclusively technical cybersecurity aspects.
Figure 11. External coherence with most relevant ongoing legislative initiatives
In terms of internal coherence, all POs are designed to be coherent with the objectives of the EU
and address the identified problem drivers (see also section 5.2.3). The measures under each PO
are compatible with each other, see Annex 7. Sovereignty related measures respond to the needs of
ensuring public order while the measures targeting critical dependencies focus on EU added value
measures, notably in the context of innovation procurement. The promotion of open source
solutions aims to build a foundation for auditable, open and interoperable services and products.
The publicly auditable code provides a level of transparency and verifiability that is not possible
in proprietary solutions, enabling authorities to independently assess security properties and verify
the absence of undisclosed data flows or access mechanisms. Open source licensing removes the
legal mechanisms that make vendor lock-in so difficult to escape in practice. The leverage that
proprietary vendors have for the maintenance of the software is diminished with open source,
reducing therefore the dependencies. The collaborative and distributed development models of
open source create a form of supply chain resilience that is different from the proprietary
solutions: open source communities distribute the maintenance and burden across a wide base of
contributors, none of which can unilaterally determine the future of the project.
88
7.4. Subsidiarity and proportionality
As highlighted in sections 3.2 and 3.3, there is a clear need for action at EU level to address the
identified problems and their drivers. All policy options respect the subsidiarity principle. In
PO1-A, the Commission uses its convening powers to enhance the existing collaborative
framework between Member States to stimulate compute capacity and remove deployment
barriers. PO1-B contains EU-level rules but is designed with subsidiarity at its core since it entails
decisions and financial incentives enforced at national level. PO1-C entails the most centralised
EU action, with decisions and financial intervention enforced at EU-level, at the expense of less
subsidiarity. In PO2-A, the EU’s involvement is limited to be a mere convenor of national actors,
hence highly respectful of subsidiarity. PO2-B requires more EU-level intervention, following a
traditional Single Market logic through an EU-defined and Member State enforced sovereignty
risk assessment mechanism, third-party audits, voluntary award criteria used for procurement of
cloud and AI computing services in the public sector, and the vendor-neutral cloud and AI training
programme. PO2-C adds further EU-level intervention, but in domains where no Member State
can act alone, in particular as regards joint procurement. Moreover, it leaves to national authorities
the ability to determine the sovereignty risk assessment outcome, while only producing guidelines
to support a uniform implementation across MS.
All policy options are assessed to be proportionate, as EU-level action is limited to what is
necessary to advance the EU’s capacity, capability and autonomy in cloud and AI computing.
PO1-A represents the lowest level of intervention and is less effective, while PO1-B entails the
necessary and most effective EU-level intervention to address bottlenecks in the deployment of
DCs by requiring that Member States implement national procedures. PO1-C is more ambitious
and would entail additional administrative implications in terms of coordination with Member
States. PO2-A introduces transparency measures that have limited prospects of changing current
trends. It is proportionate to its ambition, as it imposes low compliance costs but its capacity to
effectively achieve the intervention’s objectives is limited. PO2-B further improves market
openness with additional compliance mechanisms that would increase impacts and the expected
benefits but may not be sufficient due to their voluntary nature. PO2-C is the most ambitious
option as it proposes structural changes, through a tailored approach in existing public
procurement practices and organised joint efforts. It is the most proportionate option given the
magnitude of the problem and the intervention’s goal to secure the necessary conditions for the
Union’s competitiveness and strategic autonomy. It strikes a balance between the need for
effective sovereignty and the need to minimise unnecessary burdens. By establishing a granular
framework with four levels of sovereignty, this option provides a nuanced and targeted approach
to addressing the needs of both the public and private sectors. This tiered approach is justified, as
it reflects the varying degrees of sensitivity and criticality of different use cases, and ensures that
restrictions are proportionate to the risks involved. The addition of attestations of qualification to
the Business Wallet of service providers and the establishment of a publicly available repository
of qualified services facilitate transparency, reuse, and sharing among stakeholders.
7.5. Sensitivity analysis
The analysis has been designed to account for uncertainty and the multidimensional nature of
policy impacts. As further detailed in Annex 4, a scenario-based sensitivity analysis was carried
out to understand how the output of the cost-benefit analysis (CBA) behaves in response to
changes in its inputs and assumptions. It was conducted to verify the uncertainty range
(confidence interval) of the values estimated in the CBA for the most impactful policy measures.
Given the different design, target stakeholders, and the type of impacts generated, the variables
under scrutiny are not uniform or comparable. Therefore, rather than applying a full sensitivity
89
analysis, a bounded variation assessment was performed by introducing minimum and maximum
variation to relevant parameters, thus creating worst and best case scenario for each measure130.
7.6. Comparison per criteria
The following table synthesises the qualitative and quantitative analysis presented in Section 6,
summarising the potential impacts of the policy options across economic, social and
environmental dimensions.
Table 14. Summary of economic, social and environmental impact of the Policy Options, relative to the baseline
Criteria PO1-A PO1-B PO1-C PO2-A PO2-B PO2-C
Economic impact + +++ ++ + ++ +++
Social impact + ++ - + ++ +++
Environmental impact + - ++ o + +
Legend: o neutral impact; + minor positive; ++ positive; +++ significant positive; - minor negative; -- negative
impact; --- significant negative, all with respect to the baseline
With respect to their economic impact131, PO1-B and PO2-C are expected to deliver the strongest
outcomes, mainly driven by long-term benefits due to accelerated infrastructure development,
joint procurement and cloud federation among Member States. PO1-A delivers only modest
benefits, limited to administrative simplification and minor efficiency. By issuing guidelines and
reinforcing the current collaborative framework between data centre developers, cloud service
providers, energy actors, public authorities and other relevant stakeholders, the option would
reduce some uncertainty around data centre planning and deployment. Data centre operators and
authorities would benefit from greater predictability in permitting and common engagement.
However, because this option would not introduce binding legal changes, dedicated support or
formal deployment mechanisms, its impact on investment decisions would remain modest. In
terms of innovation and technological sovereignty, this option could help create a more
predictable environment for data centre development but would not change Europe’s ability to
scale cloud computing capacity. Wider economic effects would also be positive but limited,
considering the reduced friction, better information flows and improvements in investment
conditions. PO1-B is expected to have a more significant economic impact, as it would generate
the strongest infrastructure related economic benefits, driven by increased computing capacity. By
promoting legislative and financial measures in the form of project facilitators for data centre
deployment, areas for fast-track sustainable development, possible national funding mechanisms
for priority projects and monitoring of capacity, this option would create stronger incentives for
investment. Reducing delays linked to data centre deployment would consequently decrease their
time to market and potentially further crowd in private investment. For data centre operators the
main benefits would stem from more predictable permitting procedures, clearer administrative
pathways and potentially lower risks. Increased attractiveness of the Union for investment
decisions would also help address regional capacity gaps. For public authorities, this option would
involve higher administrative responsibilities than PO1-A, as they would need to designate these
zones and contribute to setting up project facilitators. These costs are expected to be offset by
broader economic benefits, including increased investment, job creation and local infrastructure
development. This would also contribute to strengthen the infrastructure based needed to develop
cloud solutions, AI and support digitalisation in the public and private sector. Under this option
130 This reflects the structure of the standard cost model used for most of the policy measures, which relies on a limited number of inputs, thus
limiting the interpretability of traditional one-at-a-time sensitivity testing. 131 The scoring of economic impacts reflects the estimated costs and benefits associated with each policy option, but also their broader effects on industry (including SMEs), public authorities, innovation and technological sovereignty, possible wider economic effects, trade, and the functioning
of the internal market.
90
the Commission’s role would focus on monitoring capacity, supporting coordination across
Member States and ensuring consistency with other EU-level objectives. PO1-C also produces
gains through EU-level fast-tracking and funding, but these remain below PO1-B due to smaller
expected time savings and a more centralised implementation, which is expected to reduce its
effectiveness. The evidence collected suggests that the expected economic benefits under this
option would be weakened by implementation delays due to the additional coordination and
governance arrangements. For operators this would reduce the value of EU-level deployment and
support. For public authorities, the option could offer long-term benefits by creating a more
coherent EU-level mechanism for data centre deployment. However, the need to coordinate at EU
level is expected to create administrative complexity, as national authorities would still need to be
involved in permitting, zoning, energy coordination, environmental assessments and local
processes, including community engagement. The option could still have a relevant economic
impact but the evidence collected suggests that this would be slower to materialise and less
effective in addressing deployment bottlenecks and national decision making procedures. PO2-A
yields limited economic benefits by increasing transparency and visibility of sovereign cloud and
AI computing services, with overall net costs aggregated across stakeholders. For cloud service
providers, especially smaller European ones, this option could still reduce barriers to market entry
and help users better understand available services. However, due to the non-binding nature of the
measures, the impact on actual demand and procurement behaviour would be limited. Similarly,
the impact on technological sovereignty would be positive but modest, contributing the overall
positive but minor economic impact of this option. PO2-B is expected to improve economic
outcomes by reducing cross-border compliance costs and increasing trust in sovereign services,
generating net benefits. Providers would be able to leverage the voluntary frameworks to create
reputational and market benefits. Wider economic benefits are expected and could include
increased demand for sovereign services and overall stronger market competition. However, while
recognising the positive economic impact of the option on different stakeholders, its effectiveness
in achieving these results would ultimately depend on market uptake and public sector adoption.
PO2-C builds on PO2-B and is thus expected to deliver the largest economic impact, despite
higher costs. The sovereignty risk assessment is expected to enable a once-only audit process
across MS with significant savings for providers, whereas the joint procurement and federation
mechanism are expected to enable savings for authorities procuring cloud and AI computing
services and exchanging their idle compute capacity. The impact of this option on innovation and
technological sovereignty is expected to be substantial. By using public demand strategically, the
public sector could support the development and scaling of sovereign cloud and AI computing
services, while still leaving to non-European providers a relevant share of the market, based on
MS needs and individual assessments. Wider economic effects would include increased
competition, higher productivity through broader adoption of advanced cloud and AI computing
services, greater resilience of digital supply chains and stronger opportunity for European SMEs.
While direct economic impacts were more robustly quantifiable, social impacts were not
monetised due to the absence of data on willingness to pay or stated preferences. The impacts
considered include differences in employment, skills development, and public acceptance of the
different options. PO1-B in combination with PM8 and PM9 is expected to produce the best
outcome among the first set of options, especially in terms of additional employment created and
public acceptance of new data centre projects. The option empowers local authorities in siting
decisions, thus fostering decision-making processes closer to citizens. Positive social impacts are
also expected as this option focuses on shaping data centre development towards sustainable,
innovative and strategic projects, e.g. facilities contributing to tangible community benefits such
as waste heat reuse and overall development of local economies. Conversely, PO1-A delivers
91
limited engagement gains, while PO1-C weakens perceived proximity to decision-making. For
cloud and AI services, PO2-B and PO2-C address skills shortages and employability, supporting
broader societal acceptance of technological sovereignty. Finally, environmental impacts were
assessed in environmental terms but not monetised to ensure consistency across impact categories
and avoid bias from uncertain valuation assumptions. In this context, PO1-C would achieve the
strongest performance by combining efficiency-driven innovation, sustainable siting, and
renewable integration, substantially reducing emissions intensity while limiting energy growth. By
contrast, PO1-B’s capacity expansion outweighs efficiency gains, resulting in the highest absolute
increases in electricity demand and CO₂ emissions. PO2-B and PO2-C provide modest
environmental benefits through improved server utilisation and reduced duplication of
infrastructure under the cloud federation.
The multi-criteria analysis (MCDA) was used to complement the CBA by integrating survey-
based evidence on the perceived effectiveness of the measures, as well as their expected
environmental and social impacts. It provided an alternative analytical lens, particularly through
its underlying components, i.e. stakeholder responses of the proposed measures. The aggregate
results were interpreted with caution, as they reflect the original configuration of measures and
options, which was partly superseded by subsequent refinements. Overall, the weighted aggregate
scores across the different assessment dimensions confirmed the consistency of the initial
comparative assessment (see annex 4 section 6). With respect to Problem 1 - limited and
geographically concentrated availability of computing capacity in the EU - economic operators
showed a preference for PO1-B over PO1-C. Similarly, PO1-B scores higher for national public
authorities, driven mainly by better cost outcomes and perceived social and environmental effects.
With respect to Problem 2 - Dependence on cloud and AI computing services supplied by non-
European providers – the results pointed to a clear preference for PO2-C because of its expected
economic and social benefits, including potential effects on transparency and citizen trust. On the
other hand, national public authorities appeared to favour PO2-B due to the voluntary nature of
most of the measures. This should be nevertheless interpreted with caution as national authorities
represented a limited share of survey responses, which may have affected the robustness of the
aggregate scores.
Table 18 compares the different POs against the criteria presented in Section 7, summarising the
relative performance of each option against the baseline scenario in terms of effectiveness,
efficiency, coherence, subsidiarity and proportionality.
Table 15. Comparison of the options per criteria, relative to the baseline132
Criteria PO1-A PO1-B PO1-C PO2-A PO2-B PO2-C
Effectiveness + +++ ++ + ++ +++
Efficiency − +++ +++ − +++ ++++
Coherence ++ +++ ++ ++ ++ +++
Subsidiarity and proportionality +++ +++ ++ +++ ++ +++
Legend: o no relevance; + more effective/efficient/coherent/proportionate than the baseline; +++ much more
effective/efficient/coherent/proportionate than the baseline; - less effective/efficient/coherent/proportionate than the
baseline; --- much less effective/efficient/coherent/proportionate than the baseline
PO1-A’s soft measures contribute to making it a cost-effective and coherent option but limit its
effectiveness, as it contributes to the specific and general objectives only to a limited extent with
respect to the status quo. PO1-B’s administrative simplification and fast-track areas strengthen its
132 “+” (more effective/efficient/coherent/proportionate than the baseline) to “+++” (much more effective/efficient/coherent/proportionate than the baseline); from '-' (less effective/efficient/coherent/proportionate than the baseline) to '---' (much less effective/efficient/coherent/ proportionate than
the baseline).
92
effectiveness by achieving the greatest expected increase in compute capacity and reduction of
bottlenecks for DC deployment, while being proportionate to solve the issue and leaving
implementation within Member States. PO1-C’s EU-level measures improve effectiveness and
efficiency, given the EU-level centralisation of efforts, but constrain subsidiarity through
centralised enforcement. PO2-A improves effectiveness in reaching the specific objectives but
presents overall net costs and falls short of the challenge at hand. PO2-B’s design balances
coherence and efficiency, while achieving only moderate effectiveness given the magnitude of the
problems to be addressed and the voluntary nature of its measures. By encouraging rather than
mandating a gradual shift towards more sovereign solutions in the public sector, PO2-B may
accelerate some porting and migration costs for users and providers, although they are expected to
remain relatively contained and market-driven. Finally, PO2-C generates strong synergies across
all criteria, with the highest costs and savings over ten years, reflecting high ambition at the
expense of increased operational complexity. It is likely to accelerate porting and transition costs
more significantly, particularly for entities deciding to move workloads and applications to the
highest levels of sovereignty assurance. While the option could increase operational complexity
and short to medium-term costs, it would be more proportionate and coherent than other options as
it matches the scale of the interventions to the nature of the challenge. Moreover, the expected
upfront and operational costs should still be weighed against long-term and less easily quantifiable
benefits, including reduced strategic dependencies, increased control over critical digital
infrastructure, improved resilience and greater European technological sovereignty.
8. PREFERRED OPTION
8.1. Outcome of comparison of policy options
Structuring the policy options into blocks addressing individual problems has allowed to evaluate
the most effective, efficient, and proportionate measures from each block of options, while also
analysing the package’s comprehensiveness in addressing both problems and their underlying
drivers. This section brings together the evidence across criteria to assess the relative merits of the
policy options in addressing the identified problems and objectives.
For the first problem, PO1-B performs as the option that most effectively improves the limited
availability of computing capacity in the EU relative to the baseline and alternative options.
Administrative streamlining and fast-track permitting at national level, coupled with strategic
funding and monitoring of capacity, appear as the most impactful options to tackle the bottlenecks
for expanding such capacity in the EU. The option is expected to generate greater benefits than
costs for both private sector and public sector stakeholders. Moreover, public acceptance of data
centres is expected to be highest because local authorities would have a stronger role in decisions
on where and how they are built. This means that decisions would be made closer to the people
affected by them. By contrast, PO1-C is expected to reach a lower level of computing
infrastructure deployment than PO1-B, because PM10 adds a layer of decision-making at EU-
level. Nevertheless, PO1-C reaches the best environmental impact due to the expected positive
contribution of EU-level funding (PM8 and PM9) to R&D and innovation in sustainable
technologies. Given the importance of fostering deployment of data centres with a focus on
sustainability and reducing environmental impacts, for the first problem, the analysis points
towards a package centred on PO1-B, supplemented by PM8 and PM9 as a proportionate
response. This combination is expected to offer a favourable balance between supporting
deployment of capacity and better energy efficiency and environmental performance per unit of
new capacity.
With respect to the second problem, PO2-C emerges as the best performing option across the
comparison criteria. Through clarity in the definition of sovereignty, which is also defined in a
93
proportionate manner, combined with complementary measures addressing procurement and open
source, it will facilitate the uptake of solutions that meet desired sovereignty levels, and provide
opportunities for European providers. It incorporates, in an incremental way, several elements of
the other PO2 policy options. Additional individual PMs are expected to deliver only limited
marginal gains: PM12 and PM13 are soft measures that the Commission could decide to undertake
in the future should the need arise; PM14 lacks perspective on how effective the enforcement of
the Data Act will be (see annex 4). On this basis, PO2-C is retained as the preferred package
addressing the second problem.
As outlined above, the first set of options primarily addresses SO1 and SO2, while the second set
is more directly aligned with SO3 and SO4. Given this differentiated contribution, a combined
application of the best performing options among the two sets is expected to be the most effective
approach to tackle the full range of problems and underlying drivers identified in the problem
definition, which are heterogeneous in nature. Taken together, these options would also support
the achievement of the general objective of ensuringthe functioning of the internal market for
cloud and AI computing services and securing the necessary conditions for the Union’s
competitiveness and strategic autonomy, by providing a coherent response to the interconnected
challenges identified. Therefore, with regards to effectiveness, the preferred package is expected
to address the identified problem drivers in a comprehensive manner and, relative to the baseline,
achieve a “very effective (“+++”) score against all Specific Objectives.
SO1: Increased
computing capacity
SO2: Conditions
for sustainable and
innovative capacity
SO3: Decreased
reliance on non-EU
providers
SO4: Enhanced
resilience of supply of
cloud and AI
computing services
Policy option 1A + ++ n.r. n.r.
Policy option 1B +++ +++ n.r. n.r.
Policy option 1C ++ ++ n.r. n.r.
Policy option 2A + n.r. + +
Policy option 2B + n.r. ++ ++
Policy option 2C ++ n.r. +++ +++
Preferred Package +++ +++ +++ +++
The retained package is hence made of the following policy measures:
PO1-B:
• PM4 - National facilitator
• PM5 - Fast-track areas
• PM6 - National funding support
• PM7 - Deployment targets
• PM8 - EU R&D funding
• PM9 - EU deployment funding for strategic projects
PO2-C:
• PM19 - Mandatory award criteria (which builds over PM16)
• PM20 – Open Source use in the public sector
• PM21 – Mandatory sovereign risk assessments for the use of cloud and AI computing services (which builds
over PM11 and PM15)
• PM22 – Joint EU-level procurement of cloud and AI (which builds over PM17)
• PM23 – SME cloud and AI support scheme (which has synergies with PM18)
• PM24 – Cloud and AI toolbox
Overall, the analysis suggests that PO1-B and PO2-C perform relatively strongly in terms of
economic and social benefits, while PM8 and PM9 play an important role in supporting
environmental sustainability. This highlights the importance of combining such options to
balance growth, sovereignty, and climate objectives in a balanced manner.
94
This package combining PO1-B, PM8 and PM9, is expected to generate positive efficiency
effects, as the expected benefits in the short and medium term seem to outweigh the generated
costs over the assessment period for different stakeholders. The main expected benefits under
PO1-B arise from the measures that simplify procedures and reduce fragmentation across Member
States. Similarly, under PO2-C key benefits are expected for public authorities under the joint
procurement scheme and the federation of capacity. These are altogether expected to reduce
administrative burdens, shorten timelines and improve legal certainty for private operators and
public authorities. This combination of options also entails implementation and transition costs.
There are mainly linked to adapting systems and procedures, familiarising stakeholders with the
new requirements and ensuring compliance. Most of these costs are one-off or transitional in
nature, while several of the expected benefits are recurring over time. This combination of options
is expected to deliver a positive balance of costs and benefits, while achieving the objectives in a
proportionate and cost effective manner.
In terms of coherence, the package is consistent with the policies described under section 1.2. The
analysis does not indicate significant overlaps, as the measures tackle the DC, cloud and AI
markets from different perspectives. Instead, they address the objectives pursued, while securing
the EU’s digital competitiveness and resilience.
This package is also respectful of subsidiarity and proportionality principles. While Member
States in principle can enact at national level many of the measures, both problems addressed are
characterised by common underlying drivers across the EU and would require a degree of
coordination and harmonisation that cannot be ensured through national action only. EU-level
intervention is therefore expected to offer clear added value, as the objectives of this initiative
seem impossible to be achieved sufficiently by Member States alone. At the same time, the
achievement of the expected impacts will be shaped by national implementation choices and
contextual factors, underlining the key role of Member States in materialising the initiative.
The MCDA was also used to examine the options and measures from an additional perspective.
While it assessed the options separately for each problem area, it provided a wat to consider their
performance across effectiveness, environmental and social impact dimensions. This approach
allowed us to capture additional social and environmental effects, for which robust monetary
valuation such as willingness to pay or shadow price estimates, was not available.
8.2. Application of the “One In One Out” (OIOO) Approach
The preferred policy package is expected to lead to both administrative cost savings and
administrative costs for businesses and national public authorities. The details of how these
administrative costs and savings have been quantified, including the underlying assumptions can
be found in Annex 4, section 3. The table below summarizes the main administrative costs and
savings, including details on the related activities, who is affected and their quantification.
Table 16. Application of the OIOO approach - Businesses
Preferred
package Activities
Economic operators
affected
Monetization
(NPV 2027 – 2036)
PO1-B PM5* One-off administrative costs of preparing the
application file to access the fast-track areas
Data centre
operators
EUR 0.8 – 2.8 m
PO1-B PM6 One-off administrative costs of preparing the
application to respond to the calls (est.12
applications every two years)
EUR 0.2 – 1.3 m
PO1-B PM7 Recurrent administrative costs of survey response
time and eventual periodic verification of data on
compute capacity
EUR 0.2 – 0.6 m
PM8 One-off administrative cost of preparing the Data centre EUR 0.7 – 1.6 m
95
Preferred
package Activities
Economic operators
affected
Monetization
(NPV 2027 – 2036)
application under calls for proposals (est. 20
applications every two years)
operators
PM9 One-off administrative cost of preparing the
application under calls for proposals (est. 15
proposals every two years)
Data centre
operators EUR 0.4 – 1.2 m
PO2-C PM19 Recurrent administrative costs of adapting the
offers to the new criteria
Cloud and AI
service providers EUR 13.6 – 96.4 m
PO2-C PM21 Recurrent administrative costs for new audits at
assurance level 2 - 4
Private sector
(independent
auditors)
EUR 4.4 – 4.4 m
PO2-C PM21 Recurrent administrative costs for audit renewals
at assurance level 2 - 4
Private sector
(independent
auditors)
EUR 7 – 7 m
PO2-C PM21 Recurrent administrative costs of intermediate
audits at assurance level 2 - 4
Cloud and AI
service providers EUR 53.9 – 134.8 m
PO2-C PM21 Recurrent administrative burden for private sector
entities under NIS 2 Annex 1 to address non-
technical risks
Private sector
entities operating in
sectors listed under
Annex I of NIS2
EUR 480– 2 620 m
PO2-C PM21 Recurrent administrative cost savings in
intermediate audits from doing it once for all 27
MS
Cloud and AI
service providers EUR 404.3 – 2 021.4 m
PO2-C PM23 One-off administrative cost of preparing the
application under call for applications (est. 16,000
applications per year)
SMEs EUR 27.7 – 83.4 m
*PM4 and PM5 are expected to generate reduced administrative burdens and improved information through more consistent and predictable interactions with permitting bodies. However, these administrative savings have not been monetised due to the lack of robust data and their relative
smaller scale compared to the economic benefits captured through the NPV approach.
Table 17. Application of the OIOO approach - Public Administrations
Preferred
Package Activities
Monetization (EUR,
NPV 2027 – 2036)
PO1-B PM4 Recurrent administrative burden reduction thanks to the project facilitator
which would reduce parallel processing and back-and-forth interactions EUR 83.1 – 138.6 m
PO1-B PM5 One-off administrative costs for drafting the strategies for national data
centre deployment EUR 1.2 – 6.1 m
PO1-B PM5 Recurrent administrative costs of mapping the fast-track areas and updating
the strategies EUR 46.4 – 86.3 m
PO1-B PM5 Recurrent administrative burden reduction thanks to the fast-track areas with
additional reduction in parallel processing and reporting duties EUR 83.1 – 138.6 m
PO2-C PM19 Recurrent administrative cost savings from using standard non-specific
award criteria when drafting the specifications EUR 4.3 – 13.1 m
PO2-C PM19 One-off administrative cost to update the procedures related to the public
procurement of cloud and AI computing services and draft plans on highly
critical use cases involving the purchase of sovereign services
EUR 3.4 – 13.5 m
PO2-C PM19 Recurrent administrative costs to update the plans and track their progress EUR 46.4 – 115.4 m
PO2-C PM20 One-off administrative costs to adapt the procurement templates to promote
open source EUR 6.9 – 13.9 m
PO2-C PM20 Recurrent administrative costs to maintain the Open Source Programme
Office each year EUR 60.5 – 211.7 m
PO2-C PM21 Recurrent cost savings from using the sovereignty scheme in tenders Around EUR 2.5 m
PO2-C PM21 Recurrent administrative costs from validation of the audited services and
renewals EUR 0.3 – 0.8 m
96
9. HOW WILL ACTUAL IMPACTS BE MONITORED AND EVALUATED?
The Commission (DG CONNECT) will be responsible for monitoring the implementation of the
intervention on a regular basis, possibly with the support of other Commission services, EU
agencies, external studies, Member State and market data. A set of possible indicators is provided
in the following table, alongside preliminary questions for the evaluation of the intervention,
which is expected five years after implementation. Annex 11 presents the detailed monitoring
framework devised for this intervention and how this will build upon existing schemes to avoid
duplication and benefits from synergies.
With respect to the expected targets of the intervention, the data provided in this assessment
presents indicative ranges, which should be interpreted as plausible trajectories. In line with the
Better Regulation Toolbox, quantitative targets were based on the problem analysis, baseline
scenario, available evidence from the assessment, including validation through interviews, and the
expected scale of policy impact. They were defined to quantitatively measure the effectiveness of
the intervention in the next years and paired with indicators for a systematic review. They
included a clear timeframe, i.e. by 2030 or 2035, expectations based on the evidence gathered and
the availability of data collection systems to make this monitoring effective.
Table 18. Summary of KPIs by Specific Objective
Operational objectives and
performance thresholds Possible indicators for monitoring/evaluation
Main data sources and
data collection methods
Increase computing capacity
deployed in the EU through
innovative and sustainable
technologies: By 2030, +200%
EU computing capacity vs
2025 and ensuring that such
capacity fully covers its needs
by 2035133; ≥25% of global
share134; energy-efficient
technologies in ≥80% of new
facilities135.
• Installed computing capacity (MW IT load) by
MS
• Aggregate general purpose and AI-optimised
compute, shifting to a measurement by FLOPs as
opposed to MW.
• EU share of global installed computing capacity
• Number of new EU-based DCs operational per
year
• Utilisation rate of EU computing capacity;
measures on PUE, WUE, location-based
emissions and related environmental impact of
data centres
• Deployment of innovative and energy-efficient
technologies (pilots launched and uptake of new
solutions)
• Share of renewable energy in data centres and
waste-heat reuse
• Total annual public and private investment in
EU-based DCs
• Share of new data centre capacity deployed
outside existing hubs and in underserved regions
Survey (and follow-up
interviews) of data centre
operators, national
authorities and
TSOs/DSOs
Desk research
EED reporting
Interviews with experts,
academia, think tanks and
associations
Industry dataset, e.g. Data
Centre Map, EUDCA
EEA emission data
Ensure attractive conditions
for the deployment of • Average permitting time for new data centre
projects
Surveys per (and follow-
up interviews) across
133 Since the precise level of self-sufficiency is impossible to anticipate today, PM7 has been conceived to monitor the evolution of supply and demand of computing capacity across EU Member States. 134 This 25% target of EU-27 global share has been derived from the estimated increase in EU capacity expected under PO1-B. Under the baseline
scenario, the EU-27 share of global data centre capacity is projected to grow from 20% (2025) to 23% (2030), i.e. reaching around 28 GW in 2030. This is based on the Goldman Sachs forecast that total global data centre capacity will reach around 122 GW by 2030. The additional capacity
expected to be enabled by the proposed measures could raise this EU share to around 27% by 2030. To account for uncertainties, a more
conservative threshold of 25% was set to capture the additional percentage points of global capacity. 135 The need to prioritise and have energy-efficient technologies in at least 80% of new installations, measured by PUE, WUE and other relevant
indicators, is considered a necessary threshold to comply with EU goals of climate neutrality by 2050.
97
Operational objectives and
performance thresholds Possible indicators for monitoring/evaluation
Main data sources and
data collection methods
sustainable and innovative
computing capacity: By 2030,
overall permitting time should
be <18 months136, +30% annual
investment vs. 2025137, ≥20 MS
with harmonised
frameworks138.
• Total administrative burden for operators
• Share of projects delayed/cancelled due to
regulatory or infrastructure barriers
• No. of MS with simplified permitting
frameworks
• Cost competitiveness index (€/MW build cost,
€/MWh energy cost vs. US/Asia)
national authorities and
operators
FDI statistics
Eurostat
Decrease the overall reliance
on non-European cloud and
AI computing services: By
2035, ≥ 30% market share of
EU-homegrown cloud and AI
computing service providers in
the EU, i. e. at least doubling it
against the baseline
• Share of total EU cloud and AI computing
services revenue captured by European service
providers
• Number of public sector authorities served by
sovereign providers per MS
• Share of installed EU DC capacity owned by
European providers
• Share of idle capacity shared among MS
Survey (and follow-up
interviews) of data centre
operators and national
authorities
Desk research
Market research
Contribute to the protection
of public order by enhancing
the resilience of supply of
cloud and AI computing
services, in particular in the
public sector: By 2035, 100%
of highly critical use cases in
the public sector operated using
sovereign cloud and AI
computing services139
• Number of cloud and AI services audited under
level 2, 3 or 4
• Compliance rate by contracting authorities (%)
with the sovereignty scheme
• Annual value of EU public procurement of
sovereign cloud and AI computing services
• SME share (%) in awarded public contracts
• Number of public sector solutions released as
open source in the repository, and their
downloads by third parties
Number of assessed and
evaluated sovereign
services in the repository
Survey (and follow-up
interviews) of DC
operators and national
authorities
Desk research
Market research
TED data
Table 19. Preliminary questions for the evaluation of the intervention
Evaluation
criterion Assessment objective and possible evaluation questions
Effectiveness
Goal: assess the extent to which the objectives of the initiative have been achieved and how
benefits have accrued to different stakeholders.
To what extent did the intervention increase EU installed computing capacity and create the
conditions for easier data centre deployment? How did it foster the development and deployment
of innovative and sustainable data centres and a better use of energy sources? To what extent did
it increase clarity around the concept of sovereign cloud and AI computing services? How did it
improve the market share of European cloud and AI computing service operators? What is their
market share for sovereign use cases? To what extent did it increase federated resources across
the public sector and joint procurement for cloud and AI computing services? To what extent did it
increase the use of open source solutions?
Efficiency
Goal: assess the extent to which the initiative has been cost-effective, analysing the relationship
between expected and actual benefits and costs.
Have benefits and cost savings been achieved at proportionate costs for different stakeholders?
136 This has been set to ensure that regulatory procedures for deploying data centres across the EU are reduced in a consistent way. Under the
baseline, average permitting duration was estimated to be of 32 months across the 12 Member States under scrutiny. Reducing this timeline to 18 months is a realistic improvement, even if still above today’s best-performing countries. 137 Reducing administrative barriers has been shown in the literature, and confirmed through interviews, to greatly improve project bankability and
attract additional investments. Hence, a 30% additional investment unlocked compared to the baseline scenario has been set as a reasonable success criterion. This should be measured against the additional capacity deployed. 138 The objective of seeing 20 Member States with harmonised frameworks stems from a willingness to improve visibility and simplify market
access to data centre investors and businesses, so that they can fully exploit the Single Market without additional administrative burdens. 139 Since the objective consists in enhancing the resilience of supply of cloud and AI computing services, in particular in the public sector, a long-
term target of full compliance is expected as a measure of success.
98
Evaluation
criterion Assessment objective and possible evaluation questions
Relevance
Goal: assess the extent to which the objectives of the initiative still reflect current and future
needs.
To what extent the initiative still addresses relevant needs? How is it still aligned with EU
priorities?
Coherence
Goal: assess the initiative’s internal and external coherence, i.e. if the different elements of the
intervention worked together to reach the set goal and if it worked well or overlapped with other
initiatives, both at EU level and national level.
To what extent is the initiative consistent with existing and future energy, digital, competition,
environmental, security rules at EU level and national level?
EU Added
value
Goal: assess the extent to which the initiative brought EU added value compared to what could
have been achieved by Member States alone.
To what extend did EU-level action prevent fragmentation of DC rules? How did it improve cross-
border service delivery and competitiveness of EU providers?
99
ANNEX 0 - ENDNOTES
i The Draghi Report on competitiveness ii Draghi report iii Competitiveness Compass iv JOIN/2023/20 final v Mission letter to EVP Virkkunen
vii The AI Continent Action Plan | Shaping Europe’s digital future viii Europe’s digital decade: 2030 targets | European Commission ix REPORT on European technological sovereignty and digital infrastructure | A10-0107/2025 | European Parliament x Council Conclusions of 5 December 2025 xi Federal Cloud Computing Strategy, February 2011? xii See for example Amazon's Invasion of the CIA Is a Seismic Shift in Cloud Computing? xiii America’s AI Action Plan xiv Accelerating Federal Permitting of Data Center Infrastructure – The White House xv Promoting The Export of the American AI Technology Stack – The White House xvi China plans network to sell surplus computing power in crackdown on data centre glut | Reuters xvii UK Compute Roadmap - GOV.UK xviii UAE Dominates Global Data Centre Rankings as Region Becomes Digital Infrastructure Hotspot -MIT Sloan
Management Review Middle East xix Decision - 2022/2481 - EN - EUR-Lex xx Directive - 2023/1791 - EN - EUR-Lex xxi Regulation – 20200/852 – EN – EUR-Lex xxii Regulation - 2019/424 - EN - EUR-Lex, which is currently under review. xxiii Digital Networks Act, information accessed from the 'Have Your Say' page xxiv Commission collects views in preparation of the European Grids Package - European Commission xxv Savings and investments union - Finance - European Commission xxvi Apply AI Strategy | Shaping Europe’s digital future xxvii Data Act xxviii Regulation - 2022/1925 - EN - EUR-Lex xxix Regulation - EU - 2024/1689 - EN - EUR-Lex xxx Cybersecurity Act (CSA) xxxi Regulation - 2022/2554 - EN - DORA - EUR-Lex xxxii Directive on Security of Network and Information Systems (NIS2) xxxiii Public Procurement Directive xxxiv Cloud Services Market to Exceed 68 Billion in 2010 | Security Magazine xxxv Gelvanovska-Garcia, Natalija; Mačiulė, Vaiva; Rossotto, Carlo Maria. 2024. Advancing Cloud and Data
Infrastructure Markets: Strategic Directions for Low- and Middle-Income Countries. Sustainable Infrastructure
Series. © World Bank. http://hdl.handle.net/10986/41587 License: CC BY 3.0 IGO. xxxvi 53% EU enterprises used paid cloud services in 2025 - News articles - Eurostat xxxvii European Cloud Providers’ Local Market Share Now Holds Steady at 15% | Synergy Research Group xxxviii European IT providers struggle to capitalise on continent-wide growth in cloud demand | Computer Weekly xxxixEuropean industrial technology roadmap for the next generation cloud-edge offering, May 2021. Available at:
https://ec.europa.eu/newsroom/repository/document/2021-
18/European_CloudEdge_Technology_Investment_Roadmap_for_publication_pMdz85DSw6nqPppq8hE9S9RbB8_7
6223.pdf xl CIA Awards Secret Multibillion-Dollar Cloud Contract; CIA awards multibillion C2E cloud contract to AWS,
Microsoft, Google, Oracle, and IBM. xli What made AWS the leader in the cloud industry? xlii Training and Certification for AWS Partners | Digital and Classroom Training | AWS; Top 10 AWS Consulting
Partners; Azure Partners – Find an Azure Expert Partner | Microsoft Azure; Microsoft partnerships drive innovation
100
and growth in Europe in times of uncertainty – Microsoft Pulse; Cloud Solution Provider program overview - Partner
Center | Microsoft Learn; xliii Cloud computing - statistics on the use by enterprises - Statistics Explained - Eurostat xliv Netflix, Prime Video and Disney+ have 85% of Europe’s SVOD market - Mobile Europe xlv European IT providers struggle to capitalise on continent-wide growth in cloud demand | Computer Weekly xlvi Grassano, N., Hernandez Guevara, H., Tuebke, A., Amoroso, S., Dosso, M., Georgakaki, A. and Pasimeni, F., The
2020 EU Industrial R&D Investment Scoreboard, EUR 30519 EN, Publications Office of the European Union,
Luxembourg, 2020, ISBN 978- 92-76-27418-6, doi:10.2760/203793, JRC123317. Available here:
SB2020_final_16Dec2020_online.pdf. xlvii The Future of Hyperscaler Capital Expenditures: A Deep Dive into AI and Cloud Computing – HyperFRAME
Research xlviii See also Arnal, J. (2025). Towards Competitive Cloud Ecosystems: Strategic Responses for Europe’s Digital
Future, IE Center for the Governance of Change. Available here: CGC_Competitive_Cloud_Ecosystems_2025.pdf xlix Crémer, Jacques & Biglaiser, Gary & Mantovani, Andrea, 2024. "The Economics of the Cloud," TSE Working
Papers 24-1520, Toulouse School of Economics (TSE). lThe UK’s Competition and Markets Authority detail this situation in their July 2025 report li Market study into cloud services | ACM lii Gartner Says Worldwide Sovereign Cloud IaaS Spending Will Total $80 Billion in 2026 liii AI and Compute; The cost of compute power: A $7 trillion race | McKinsey; How Can We Meet AI’s Insatiable
Demand for Compute Power? | Bain & Company; AI to drive 165% increase in data center power demand by 2030 |
Goldman Sachs; The 2025 AI Index Report | Stanford HAI liv Statista Technology Market Insights, 2025. Available here: Artificial Intelligence - EU-27 | Statista Market Forecast lv Data Centres | CBRE lvi Global Data Center Trends 2025 | CBRE lvii Big Tech's AI investments set to spike to $364 billion in 2025 as bubble fears ease? lviii Technopolis Group et al. (2025), "Study: Cloud and AI". lix Technopolis Group et al. (2025), "Study: Cloud and AI". lx KPMG | The evolving data centre landscape lxi Data coming from the Cloud and AI Study. Additional data can be found here: Financing Infrastructure for a
Competitive European AI - Groupe d'études géopolitiques lxii Technopolis Group et al. (2025), "Study: Cloud and AI". lxiii Savills | Costs on the rise lxiv Unlocking the European AI revolution | McKinsey lxv Data Centres | CBRE lxvi Data Centres | CBRE lxvii RTE | No new data centres for the capital for the foreseeable future lxviii HOW DATA CENTERS HAVE COME TO MATTER: Governing the Spatial and Environmental Footprint of
the ‘Digital Gateway to Europe’ - Monstadt - 2025 - International Journal of Urban and Regional Research - Wiley
Online Library lxix CREOS annual report 2021 lxx ServerMania | Cloud server prices lxxi Mistral AI warns of lack of data centres and training capacity in Europe | Euronews lxxii Public Cloud - EU-27 | Statista Market Forecast lxxiii Cloud is a Global Market - Apart from China | Synergy Research Group lxxiv European Cloud Providers’ Local Market Share Now Holds Steady at 15% | Synergy Research Group lxxv European Cloud Providers’ Local Market Share Now Holds Steady at 15% | Synergy Research Group lxxvi European Software and Cyber Dependencies lxxvii Competition in the provision of cloud computing services (EN) lxxviii Asterès | Technological dependence on American Software and Cloud services lxxix ChatGPT in the Public Sector - overhyped or overlooked? lxxx AI in Europe: A new opportunity for growth | McKinsey lxxxi Understanding cloud outages: Causes, consequences and mitigation strategies | HCLTech lxxxii What the EU Needs to do to Challenge Big Tech Cloud Dominance | TechPolicy.Press; Amazon web services
return to 'normal operations' after mass outage, tech giant says - BBC News.
101
lxxxiii Nicolas Guillou, French ICC judge sanctioned by the US: 'You are effectively blacklisted by much of the world's
banking system' lxxxiv Kyndryl announces agreement to purchase cloud-services provider Solvinity lxxxv Ibid lxxxvi Dutch parliament calls for end to dependence on US software companies | Reuters lxxxvii ECB Guide on outsourcing cloud lxxxviii Microsoft Can't Keep EU Data Safe From US Authorities lxxxix Section 3 - Capgemini. Cloud Sovereingy: the road ahead xc Page 43 to 47 - Baromètre de la cybersécurité des entreprises du CESIN, xci How Many Data Centers Are in the US? Latest Statistics and Trends - C&C Technology Group xcii Recent Trends in U.S. Services Trade: 2022 Annual Report xciii Global Infrastructure Regions & AZs xciv Discover more about regions and availability zones | OVHcloud Worldwide xcv OECD (2025), “Competition in the provision of cloud computing services”, OECD Roundtables on Competition
Policy Papers, No. 323, OECD Publishing, Paris, https://doi.org/10.1787/595859c5-en, pp. 23-24. xcvi Oaktree-backed firm unveils $1.2 billion Amsterdam 'hyperscale' data centre project | Reuters xcvii https://www.mordorintelligence.com/industry-reports/europe-colocation-market-industry xcviii Data Centres | CBRE xcix Leitmotiv - Toward our Digital Future c Google raises 2025 capex estimate, again, to $91-93bn - plans "significant increase" in data center spend for 2026 -
DCD ci OVHcloud - Results cii Too late to act? Europe’s quest for cloud sovereignty | Clingendael ciii NGP Capital | The Cloud Evolution: From Hyperscaler Dominance to Modular Infrastructure civ See here: Announcing the Regional 2024 AWS Partners of the Year for Europe, Middle East, and Africa? cv See for instance AWS Distribution program or Microsoft’s partner program cvi The economics of the Cloud – Toulouse School of Economics – March 2024 cvii Partnerships Between Cloud Service Providers and AI Developers cviii Competition in the provision of cloud computing services (EN); Partnerships Between Cloud Service Providers
and AI Developers. cix Cloud Computing & Web Hosting | OVHcloud Worldwide; Cloud Computing & Web Hosting | OVHcloud
Worldwide. cx Eindrapport In het kader van het quickscan-onderzoek naar technische, organisatorische en juridische gaps tussen
Europese/Nederlandse cloudproviders en Amerikaanse hyperscalers voor het ministerie van Economische Zaken |
Tweede Kamer der Staten-Generaal cxi Eindrapport In het kader van het quickscan-onderzoek naar technische, organisatorische en juridische gaps tussen
Europese/Nederlandse cloudproviders en Amerikaanse hyperscalers voor het ministerie van Economische Zaken |
Tweede Kamer der Staten-Generaal cxii European Cloud Providers: What Are the Options Today? - InfoQ cxiii Best Strategic Cloud Platform Services Reviews 2026 | Gartner Peer Insights cxiv CMS. (n.d.). Expert Guide on Real Estate Data Centre Consenting. CMS Law. cxv (2024). European data center overview 2024 cxvi European Real Estate Market Outlook 2026 cxvii The Real Estate Power Play Behind Europe’s Data Center Growth cxviii See McKinsey: Unlocking the European AI revolution | McKinsey; IEA: Energy demand from AI – Energy and
AI – Analysis - IEA; The Shift Project: Environmental-impacts-of-digital-technology-5-year-trends-and-5G-
governance_March2021.pdf; Ember: Grids for data centres: ambitious grid planning can win Europe's AI race cxix IDC Report Reveals AI-Driven Growth in Datacenter Energy Consumption, Predicts Surge in Datacenter Facility
Spending Amid Rising Electricity Costs. See also: CERRE_Report_DCs_FinalPDF.pdf pp 13-15. cxx For example in Dublin: Data Centres in Ireland, and in Amsterdam: Challenges in the Dutch Data Center Market. cxxi Grids for data centres: ambitious grid planning can win Europe's AI race; Ember, 2025; Reuters, 2025; BCG,
2025; Power connection requests for Italy data centres rise to 42 GW at end-March | Reuters; See also: Single Market
- Compendium of obstacles - 21 May 2025 cxxii Why Retrofit Could Dominate Data Centre Builds This Decade | Data Centre Magazine cxxiii Data centre developers explore linking up to UK gas pipelines
102
cxxiv Europe primed for data centre ABS financing as investment soars | News | About Us | Linklaters cxxv Why data center finance is diversifying cxxvi AI Drives Cloud Player Capex Amid Cautious Overall Spend cxxvii Knight Frank: Global Data Centres Report cxxviii China invests $6.1 billion in computing data center project, official says | Reuters cxxix North America Data Center Report Midyear 2025 cxxx Circular water solutions key to sustainable data centres | World Economic Forum; Mytton, D. Data centre water
consumption. npj Clean Water 4, 11 (2021). Data centre water consumption. cxxxi See for example Revealed: Big tech’s new datacentres will take water from the world’s driest areas. Globally,
however, total data centre water use is much smaller than that of other sectors. See: icef.go.jp/wp-
content/themes/icef_new/pdf/roadmap/icef2025_roadmap.pdf. cxxxii See for example: ECIPE | The EU’s Trillion Dollar Gap in ICT and Cloud Computing Capacities: The Case for a
New Approach to Cloud Policy cxxxiii Eurostack | Deploying the Eurostack: What’s needed now. cxxxiv Sovereign Cloud Technologies – EU-Lisa technology brief – June 2025 cxxxv Concession from the Department for digital transformation to the Polo Strategico Nazionale cxxxvi Germany launches government cloud – Press release cxxxvii CISPE’s Buying Cloud Services in the Public Sector provides a comprehensive overview of public procurement
difficulties cxxxviii ECIPE, Occasional paper No. 04/2025: ’Boosting Efficiency and Quality in EU Public Services: The Need for a
European Multi Cloud-First Strategy’: ECI_OccasionalPaper_04-2025_LY04.pdf cxxxix V-ICT-OR delivers innovative Flemish government services with Azure Container Apps | Microsoft Customer
Stories cxl Digitale Abhängigkeit von Microsoft: Risiken und Auswirkungen auf die EU-Wirtschaft cxli "We're done" - major government organization slams Microsoft Teams as it drops Windows for good; Danish
government agency to ditch Microsoft software in push for digital independence; Denmark’s Digital Declaration of
Independence: A Growing European Revolt Against Big Tech Dependency. cxlii What is the Log4j vulnerability? cxliii “EDLER, J. "Technology sovereignty for the EU: Needs, concepts, pitfalls and ways forward". Available at
European Commission cxliv AWS European Sovereign Cloud cxlv EU Sovereign Cloud | Oracle Europe cxlvi Capgemini and Orange are pleased to announce the launch of commercial activities of Bleu, their future “cloud de
confiance” platform - Newsroom Orange Group cxlvii Clarence | Le premier cloud souverain déconnecté en Europe cxlviii Sovereign AI | Oracle Europe cxlix The next chapter for UK sovereign AI | OpenAI cl See for example: Die souveräne Cloud für den öffentlichen Sektor - STACKIT or Sovereign Cloud for Tech Innovation |
Scaleway. cli Sovereign Cloud in the EU: Providers, Challenges, and Opportunities clii Introducing the Overview of the AWS European Sovereign Cloud whitepaper | AWS Security Blog cliii 6G Infrastructures for Edge AI: An Analytical Perspective cliv When AI Takes Time to Think: Implications of Test-Time Compute clv This is a challenge already observed in the United States, where electricity prices have risen in regions with high
data-centre concentration. It highlights the need for the EU to ensure a more balanced distribution of data-centre
deployment across Member States, in order to prevent similar electricity price spikes in specific areas. Additional
information available here: How AI Data Centers Are Sending Your Power Bill Soaring; $64 billion of data center
projects have been blocked or delayed amid local opposition — Data Center Watch clvi Global Data Center Trends 2025? clvii Time to place our bets: Europe’s AI opportunity? clviii Europe Public Cloud Market Size & Outlook, 2023-2030. clix Artificial Intelligence Index Report 2025 clx Technopolis Group et al. (2025), "Study: Cloud and AI". clxi IEA (2025). Energy and AI, pp. 93-96. Available at: https://www.iea.org/reports/energy-and-ai clxii Technopolis Group et al. (2025), "Study: Cloud and AI".
103
clxiii European Commission clxiv ACTON, M., BOOTH, J. and PACI, D., 2025 Best Practice Guidelines for the EU Code of Conduct on Data
Centre Energy Efficiency, Publications Office of the European Union, Luxembourg, 2025,
https://data.europa.eu/doi/10.2760/9449356, JRC141521. clxv Technopolis Group et al. (2025), "Study: Cloud and AI". clxvi Blind, K. et al. (2021). The impact of Open Source Software and Hardware on technological independence,
competitiveness and innovation in the EU economy. European Commission. Available at: Study about the impact of
open source software and hardware on technological independence, competitiveness and innovation in the EU
economy clxvii See Special report 12/2025 - The EU’s strategy for microchips ; Competition in artificial intelligence
infrastructure (EN) and Report_Emerging-Resilience-in-the-Semiconductor-Supply-Chain.pdf clxviii Technopolis Group et al. (2025), "Study: Cloud and AI". clxix As mentioned in the Copenhagen Economics Study (Available here), the data centre sector can contribute to
increasing FDI by fostering clusters along parts of the digital infrastructure value chain which can, in turn, attract
other businesses, e.g. suppliers or technology firms benefitting from proximity to the facilities. Other factors
contributing to the selection of data centre locations are also presented here: 2025-Data-Center-Site-Selection-
Dynamic-Brief.pdf clxx The EU attracted the most FDI projects among world’s regions according to figures from fDi Markets. Data centre
investors announced several mega projects, i.e. with over $1bn in committed CapEx across Europe and FDI worth
more than $69bn in 2024. Figures available here: fDi’s European Cities and Regions of the Future 2025 clxxi OECD (2025), “Competition in the provision of cloud computing services”, OECD Roundtables on Competition
Policy Papers, No. 323, OECD Publishing, Paris, https://doi.org/10.1787/595859c5-en. clxxii British International Investment insight, How-does-access-to-a-local-data-centre-affect-business-productivity.pdf clxxiii AWS services recover after daylong outage hits major sites clxxiv WEF_Global_Cybersecurity_Outlook_2025.pdf and CrowdStrike outage: We finally know what caused it - and
how much it cost | CNN Business clxxv Accelerating Europe’s AI adoption: The role of sovereign AI clxxvi Right Scaling for Right Pricing: A Case Study on Total Cost of Ownership Measurement for Cloud Migration |
Springer Nature Link clxxvii Artificial intelligence: Europe needs to start dreaming again clxxviii See Digitalisation and Employment, International Labour Organization, 2022 clxxix Findings of an unpublished Research briefing by FGS Research made available to the Commission in November
2025 (survey of 3,022 adults across 6 markets - UK, US, Japan, France, Germany, Italy – with a sub-set of questions
specific to France, Germany and Italy only) indicates that (1) 71% of surveyed populations think that “Technological
sovereignty helps create and protect local jobs” as opposed to 22% who insisted that Technological sovereignty harms
and threatens local jobs. clxxx See Infrastructure or Intrusion? Europe’s Conflicted Data Center Expansion clxxxi See Eurostat (2024, 2025), E-government activities of individuals via websites, Individuals - frequency of
internet use and Meetings via the internet by size class of enterprise clxxxii OpenAI reports that Chatgpt search alone has 120 milion average monthly users in the EU. clxxxiii Ibid (1) on average, 65% of citizens in France, Germany and Italy insist that “It is worth investing more to keep
technologies like satellites, cloud services, and AI systems under national control, even if that means costs are
higher”, as opposed to the average of 28% thinking that “It is more important to keep costs low, even if that means
relying on foreign companies for things like satellites, cloud services, and AI systems”. (2) On average, 54% of
citizens think “Governments should lead and invest heavily in technology development to secure sovereignty”,
whereas 36% that “The private sector should drive technology innovation with minimal government intervention”. (3)
But only 39% think “It is worth spending more on developing our own technologies, even if that means less funding
for other areas like healthcare”, while 55% argue that “Public money should prioritise essential services like
healthcare and education, even if that means relying on foreign technology”. clxxxiv This is in line with past expectations and figures. EEA’s indicator shows that the GHG intensity of power
generation in the EU has been falling for decades and was about 40% lower in 2024 than ten years before. See here:
Greenhouse gas emission intensity of electricity generation in Europe | Indicators | European Environment Agency
(EEA) clxxxv The IEA 4E EDNA policy work underlines that reducing infrastructure use (PUE) can be achieved by smaller
data centres through grants or technical support, separate from Minimum Energy Performance Standards or
104
obligations which may be challenging to apply to smaller data centres. Available here: Policy development on energy
efficiency of data centres draft final report v1.05. Smaller operators typically lag behind hyperscalers as they lack the
capital and human resource capacity to invest in efficiency innovations, leading to structurally higher PUEs, and often
depending on available technologies. clxxxvi https://www.networkworld.com/article/972407/5-ways-to-boost-server-efficiency.html clxxxvii See also: Al, data, and computing: shaping infrastructures for a decarbonised world - The Shift Project IL clxxxviii IEA (2025). Overcoming energy constraints is key to delivering on Europe's data centre goals. Available at:
https://www.iea.org/commentaries/overcoming-energy-constraints-is-key-to-delivering-on-europe-s-data-centre-goals clxxxix IEA (2025). Energy and AI. Available at: https://www.iea.org/reports/energy-and-ai cxc European Environment Agency (2024). Water abstraction by economic sector in the 27 EU Member States, 2000-
2022. Available at: https://www.eea.europa.eu/en/analysis/indicators/water-abstraction-by-source-and/water-
abstraction-by-economic cxci Circular thinking for data centres - Arup cxcii Alissa H., Nick T., Raniwala A. et al., Using life cycle assessment to drive innovation for sustainable cool clouds,
Nature, Vol. 641, 2025, pp. 331–338, available at: Using life cycle assessment to drive innovation for sustainable cool
clouds | Nature cxciii Schneider Electric. (2023). Quantifying Data Center Scope 3 GHG Emissions to Prioritize Reduction Efforts—
White Paper 99. https://www.se.com/ww/en/download/document/SPD_WP99_EN/ cxciv Swedish audit reports (2022). See: Central government initiatives to stimulate investments in data centres (RiR
2022:18)
EN EN
EUROPEAN COMMISSION
Brussels, 3.6.2026
SWD(2026) 502 final
PART 2/2
COMMISSION STAFF WORKING DOCUMENT
IMPACT ASSESSMENT REPORT
Accompanying the
Proposal for a
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strengthening Europe's cloud and AI
ecosystem (Cloud and AI Development Act)
{COM(2026) 502 final} - {SEC(2026) 502 final} - {SWD(2026) 503 final}
Contents
0. GLOSSARY ................................................................................................... V
ANNEX 1: PROCEDURAL INFORMATION ...................................................... 7
1. LEAD DG, DECIDE PLANNING & CWP REFERENCES .......................... 7
2. ORGANISATION AND TIMING .................................................................. 7
3. CONSULTATION OF THE RSB ................................................................... 7
4. EVIDENCE, SOURCES AND QUALITY ................................................... 17
ANNEX 2: STAKEHOLDER CONSULTATION (SYNOPSIS REPORT) ........ 18
1. INTRODUCTION ......................................................................................... 18
2. OVERVIEW OF CONSULTATION ACTIVITIES ..................................... 18
2.1 Public Consultation on the Cloud and AI Development Act .............. 18
2.2 Study to support the impact assessment .............................................. 18
2.3 Overview of workshops, seminars and roundtables: ........................... 18
2.4 Awareness raising events in Member States and international outreach
............................................................................................................. 19
2.5 Bilateral meetings with industry, academic institutions, think tanks .. 19
3. PARTICIPANTS AND METHODOLOGY ................................................. 19
3.1 Participants .......................................................................................... 19
3.2 Methodology ........................................................................................ 20
4. ANALYSIS OF THE STAKEHOLDER CONSULTATION ...................... 20
4.1 EU businesses and public authorities’ dependency on non-EU cloud and
AI providers ......................................................................................... 20
4.2 Bottlenecks to expand (sustainable) data centre capacity within the EU
............................................................................................................. 21
4.3 Barriers to public sector transformation and public procurement ....... 23
4.4 Lack of common concept and operational criteria for sovereign cloud
and AI computing services .................................................................. 25
ANNEX 3: WHO IS AFFECTED AND HOW? ................................................... 27
1. PRACTICAL IMPLICATIONS OF THE INITIATIVE .............................. 27
2. SUMMARY OF COSTS AND BENEFITS ................................................. 30
3. RELEVANT SUSTAINABLE DEVELOPMENT GOALS ......................... 34
ANNEX 4: ANALYTICAL METHODS .............................................................. 36
1. METHODOLOGY ........................................................................................ 36
2. GENERAL ASSUMPTIONS FOR THE MODELLING OF THE POLICY
MEASURES .................................................................................................. 37
2.1. Valuation framework and time horizon ............................................... 37
2.2. Unit cost parameters/ standard cost model .......................................... 37
2.3. Estimates related to data centres .......................................................... 39
2.4. Estimates related to Public Procurement ............................................. 52
2.5. Indirect benefits from cloud adoption ................................................. 56
3. IMPACTS OF POLICY MEASURES IN TERMS OF COSTS AND
BENEFITS .................................................................................................... 57
3.1. PM1: Expanding the Alliance for Industrial Data, Edge and Cloud with
a working group on data centres and extending membership to relevant
players .................................................................................................. 57
3.2. PM2: Creating a forum for exchanges between relevant public and
private stakeholders involved in the buildout of data centres ............. 59
3.3. PM3: Adopting guidelines on building sustainable data centres in the
EU ........................................................................................................ 61
3.4. PM4: Project facilitators for the roll-out of data centres ..................... 64
3.5. PM5: Mechanism for Member States to identify areas to fast-track data
centre deployment ............................................................................... 68
3.6. PM6: National funding support for data centres ................................. 72
3.7. PM7: Set deployment targets and monitor progress ............................ 74
3.8. PM8: EU Funding for R&D and innovation ecosystems for cloud and
AI ......................................................................................................... 76
3.9. PM9: EU deployment funding for strategic projects ........................... 79
3.10. PM10: EU-level identification of areas for fast-track data centre
deployment .......................................................................................... 80
3.11. PM11: Creating EU-level harmonized criteria for sovereign cloud and
AI computing services ......................................................................... 84
3.12. PM12: Creating EU guidelines for sovereign Cloud AI computing
services for public procurement .......................................................... 88
3.13. PM13: Annual conference on digital sovereignty ............................... 90
3.14. PM14: Interoperability flanking measures .......................................... 92
3.15. PM15: Voluntary sovereign risk assessments for the use of cloud and
AI computing services in the public sector ......................................... 94
3.16. PM16: Non-mandatory specific award criteria for the procurement of
cloud and AI computing services ...................................................... 101
3.17. PM17: Public Sector Cloud Federation ............................................. 103
3.18. PM18: Vendor-neutral EU cloud/AI skill certificates ....................... 106
3.19. PM19: Mandatory specific award criteria for the procurement of cloud
and AI computing services ................................................................ 109
3.20. PM20: Boosting open source in public administrations ................... 111
3.21. PM21: Mandatory sovereign risk assessments for the use of cloud and
AI computing services in the public sector ....................................... 116
3.22. PM22: EU-level Procurement of cloud and AI computing services . 125
3.23. PM23: Financial support for SMEs to adopt cloud and AI ............... 127
3.24. PM24: Tools to enrich EU cloud and AI computing services offering
129
4. STRENGTHS AND LIMITATIONS OF THE ANALYSIS ...................... 131
5. SENSITIVITY ANALYSIS ........................................................................ 132
6. MULTI-CRITERIA DECISION ANALYSIS ............................................ 138
6.1. Models used for the multi-criteria analysis ....................................... 138
6.2. Development and comparisons of policy options (MCDA) .............. 143
7. ENVIRONMENTAL IMPACT ANALYSIS ............................................. 155
7.1. Installed IT capacity .......................................................................... 156
7.2. Utilisation and efficiency assumptions .............................................. 156
7.3. Conversion to annual electricity demand .......................................... 156
7.4. Emissions calculations ....................................................................... 157
7.5. Limitations ......................................................................................... 157
8. SO3 – HOW REALISTIC IS IT TO SET AN OBJECTIVE FOR EU
PROVIDERS TO REACH A MARKET SHARE OF 30% BY 2035? ...... 158
9. TYPES OF PROCEDURES, PERMITS AND DATA CENTRE
DEPLOYMENT TIMELINES PER MEMBER STATE ............................ 160
ANNEX 5: COMPETITIVENESS CHECK ....................................................... 172
1. OVERVIEW OF IMPACTS ON COMPETITIVENESS ........................... 172
SYNTHETIC ASSESSMENT ............................................................................ 172
ANNEX 6: SME CHECK ................................................................................... 176
ANNEX 7: EXTERNAL COHERENCE WITH RELEVANT EU LEGISLATION
AND POLICY INITIATIVES .................................................................... 182
ANNEX 8: KEY TECHNICAL CONCEPTS ..................................................... 196
ANNEX 9. BASELINE SCENARIO (POLICY OPTION 0) ............................. 199
1. Limited and geographically concentrated availability of computing
capacity .............................................................................................. 199
2. Dependence on cloud and AI computing services supplied by non-
European providers ............................................................................ 202
ANNEX 10. CADA INTERVENTION LOGIC ................................................. 207
ANNEX 11. OPERATIONAL MONITORING AND EVALUATION SYSTEM
..................................................................................................................... 208
1. Output monitoring: deliverables and direct effects ........................... 208
2. Outcome monitoring: mid-term change ............................................ 209
3. Impact monitoring: long-term structural effects ................................ 211
4. Evaluation arrangements and timing ................................................. 212
ANNEX 12. COSTS OF MIGRATING AND PORTING APPLICATIONS .... 214
12.1 Considerations about migrating and porting ..................................... 214
12.2 Cost modelling for public administrations to port services (cloud-to-
sovereign cloud) ................................................................................ 218
12.3 Phases of a porting operation ............................................................ 220
12.4 Aspects considered for the analysis of costs ..................................... 224
12.5 Limitations of the analysis ................................................................. 235
ANNEX 13. COMPARATIVE ANALYSIS OF SELECTED CLOUD SERVICES
..................................................................................................................... 236
0. GLOSSARY
Term or acronym Meaning or definition
AI Artificial Intelligence
AZ Availability Zone
CEN European Committee for Standardization
CENELEC European Committee for Electrotechnical Standardization
CfE Call for Evidence
Colocation data
centre
A data centre in which one or more customers install and manage
their own network or networks, servers and storage equipment and
service
Colocation data
centre operator
An organisation who manages and leases/sells space, security,
network access, power and cooling capacity from a colocation
data centre to one or more customers who install and manage their
own network or networks, servers and storage equipment and
services
CSP Cloud service provider
Data Centre (DC)
A data centre (DC) is defined as a structure or a group of
structures used to house, connect and operate computer systems/
servers and associated equipment for data storage, processing
and/or distribution, as well as related activities.
DSO
Distribution System Operators operate and manage local and
regional distribution networks, delivering electricity to end users
from the transmission system, such as Data Centres up to a given
size. See also TSO.
EC European Commission
Enterprise data
centre operator
Enterprise data centre operator is a physical or legal person who
manages the entire enterprise data centre, including the building
and the use of the information technology services delivered
EED Energy Efficiency Directive
ESO(s) European Standardisation Organisation(s)
EU European Union
Term or acronym Meaning or definition
FLAPD Refers to the data centre markets located in Frankfurt, London,
Amsterdam, Paris and Dublin.
FTE Full Time Employee
GDPR General Data Protection Regulation
Hyperscaler
Hyperscalers are very large cloud service providers. They are
characterized by their ability to provide cloud computing at very
large scale. Hyperscalers include Amazon (Amazon Web
Services), Microsoft (Azure), and Google (Google Cloud
Platform).
IA Impact Assessment
ICT Information and Communication Technology
IPCEI Important Project of Common European Interest
MS Member State(s)
OSS Open-Source Software is software that is released under a license
that allows users to view, modify, and distribute the source code.
PC Public Consultation
PUE
Power Usage Effectiveness, or PUE, is a metric used to measure
the energy efficiency of data centres. It is defined as the ratio of
total facility energy to the energy used by IT equipment. A PUE
of 1.2 implies that for each unit of energy spent in powering IT
equipment, 0.2 units are spent for non-IT equipment such as
cooling. A lower PUE indicates better energy efficiency, as it
means less energy is being used for non-IT purposes.
Simpl Smart middleware platform for data spaces
SME(s) Small- and Medium-sized Enterprise(s)
TFEU Treaty on the Functioning of the European Union
TS Technical Specification
TSO
Transmission System Operators, or TSOs, manage the high-
voltage transmission networks that transport electricity over long
distances, ensuring stability and reliability across regions. Above
a given size, Data Centres connect to the grid directly through
TSOs. See also DSO.
7
ANNEX 1: PROCEDURAL INFORMATION
1. LEAD DG, DECIDE PLANNING & CWP REFERENCES
The lead DG is Directorate-General for Communications Networks, Content and Technology
(DG CNECT), UNIT E2: Cloud and Software.
The DECIDE reference number of this initiative is PLAN/2025/815.
The Commission work programme for 2025 includes a legislative action for a cloud and AI
Development Act, under the header “3.1 A new plan for Europe’s sustainable prosperity and
competitiveness”.
2. ORGANISATION AND TIMING
The impact assessment started in 2025, with the public consultation and call for evidence
published on 9 April 20251.
The impact assessment on a possible cloud and AI Development Act was coordinated by an
Interservice Coordination Group (ISCG). The Commissions Services participating in the ISCG
were: Secretariat-General, Legal Service, DG Communications Networks, Content and
Technology, DG for Internal Market, Industry, Entrepreneurship and SMEs, DG for Energy,
DG for Trade, DG for Digital Services, DG for Environment, DG for Climate Action, and DG
for Competition.
The ISCG met on 24 July 2025, on 12 September 2025, 28 November 2025, and 25 January
2026. It was consulted throughout the different steps of the impact assessment process, notably
on all stakeholder consultation materials and deliverables from the external contractor and on
the draft impact assessment before submission to the Regulatory Scrutiny Board (RSB).
3. CONSULTATION OF THE RSB
This version was submitted to the RSB on 30 April 2026. The RSB issued a positive opinion
with comments on 8 May 2026. The comments received from the Board have been addressed
in the revised version of the impact assessment as detailed below.
Table 1. Modifications of the impact assessment report in response to RSB comments
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
(B) Key issues
The report should sufficiently
describe the policy measure
establishing a set of non-price
award criteria and the one
encouraging the use, reuse,
development and sharing of open-
source assets by the public sector.
PM16 and PM19 have been reformulated to
support EU added value. The non-award criteria
defined under these PMs should be of ancillary
and non-decisive value in view of the overall
tender and are to be included, voluntarily, by
contracting authorities in the procurement of
cloud and AI computing services that are not off
the shelf, standardised or commercially
available services. These criteria will earn
Main report
sections 5.1.1
and 5.1.2 and
Annex 4.
1 AI Continent – new cloud and AI development act.
8
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
additional points to tenderers. Furthermore, the
criteria have been updated. Given that the costs
and benefits entailed majorly administrative
costs (e.g. updates of templates) these values
remain unchanged.
The description of PM20 in section 5.1.2 has
been updated to make more explicit what it is
intended with ‘use, reuse, and share’ in this
context, arguing on the motivation to follow
such an approach.
The internal coherence between
the policy measures addressing
sovereign cloud and AI services
and those addressing critical
dependencies and supporting the
use of Open Source should be
sufficiently explained and
assessed.
Under section 7.2. the coherence between
sovereign cloud and AI services, critical
dependencies and open source interplay with
each other and are mutually reinforcing each
other.
Main report
section 7.2
The costs related to migration and
porting and the related
uncertainties should be
sufficiently reflected in the
assessment of the total costs in the
main report.
A portion of the costs of porting applications
have been included as part of the cost – benefit
analysis. The text argues that the central value
of EUR 86.3 bn should be considered as a
maximalist approach, where all out of the 30%
of the cloudified solutions requiring sovereign
level 2, 3, and 4 of the 6 400 essential entities
under NIS2 would be ported to a sovereign
cloud service as of result of the intervention. A
more plausible scenario has been considered,
where all applications at level 4 (1% of the total
applications) and half of the applications at
level 3 (5% compared to the estimated 9%) will
be ported over 3 years as a result of the
intervention.
Main report
6.1.2, 7.2 and
Annex 12,
section 12.4
(C) What to improve
The policy measures aimed at
reducing critical dependencies
(PM16/PM19) should be better
described, in particular why the
criterion “outside of the country of
dependencies” is used rather than
“within the EU”, on the basis of
which criteria the dependency
threshold is set at 50%, how the
non-price award criteria are to be
used by public buyers, how they
will be used and assessed. The
report should also better discuss
the representativeness of the
assumptions in PM15 regarding
what percentages of use cases
would respectively require
sovereignty levels 1, 2, 3 and 4.
PM16 and PM19 have been reformulated to
support EU added value. The non-award criteria
defined under these PMs should be of ancillary
and non-decisive value in view of the overall
tender and are to be included, voluntarily, by
contracting authorities in the procurement of
cloud and AI computing services that are not off
the shelf, standardised or commercially
available services.
Under PM15, both in the main report and Annex
4, a clarifying sentence has been included to
argue on the representativeness of the
assumptions.
Main report
5.1.1 and 5.1.2
and Annex 4.
9
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
Policy measure PM20 encourages
the use of Open Source in the
public sector and requires the
contracting authorities to assess
the equivalence/superiority of
Open Source over proprietary
solutions in the tendering
procedure. PM22 encourages EU
level joint public procurement.
Building on the annex, the report
should better analyse the impact of
these measures on public
authorities, including the need for
specialised expertise and interplay
with envisaged Open-Source
Programme Offices, as well as
how a uniform approach across the
EU will be ensured. The impact of
these measures on the companies
offering the relevant IT services
should also be discussed
The description of PM20 in Annex 4 has been
expanded. It is discussed that the initial
administrative costs incurred due to the
comparative assessment of solutions
(proprietary vs. open source) in staff training
would result in capability benefits in the long
term. Furthermore, public authorities may rely
on OSPOs as centres of excellence in open
source, at technical, legal, procedural and
strategic level, who can support public
authorities with standardised evaluation
frameworks and methodologies. Uniformity
may be achieved by means of recommendations
and experience sharing in the OSPO networks.
The new text also includes how these new
conditions would impact businesses, both
providers of proprietary solutions and open
source solutions, and the overall market.
Annex 4,
description of
PM20; main
report section
6.1.2 and 6.1.4
Regarding external coherence: as
the policy measure PM11,
defining the sovereignty levels
and the associated requirements,
includes criteria based on the
European Cybersecurity
Certification Scheme for Cloud
Services (EUCS), which has not
yet been adopted, the report
should describe in more detail the
interplay of the two initiatives as
well as implementation
uncertainties for EUCS
PM11 includes the inter-relation between the
sovereignty levels and EUCS as one of the
criteria to fulfil.
This footnote argues that as part of the ‘One
Europe, one market’ roadmap agreed by the
Parliament, the Council and the Commission,
the co-legislator have agreed to finalise
negotiation for this initiative ed by Q4 2027.
Adding one year for the measures to take effect,
this implies an enter into force in early 2029.
EUCS technical work is done and has been
adopted by CEN-CENELEC Technical
Specifications. The candidate scheme has
therefore reached an advanced stage of
development, which now needs to be
transformed into an Implementing Act to be
adopted under the Cybersecurity Act, a process
much shorter than CADA’s interinstitutional
negotiations.
Main report
5.2.2 and
Annex 4, as a
footnote in
PM11.
In view of the impact on the total
costs of the initiative, the costs
related to migration and porting,
described in box 2 and in the
annex, should be reflected in a
summary table of costs and
benefits.
The table and text in section 7.2 have been
modified to reflect the costs of porting. Under
6.1.2 the estimated costs are detailed based on
the explanations provided under Annex 12,
section 12.4, which has also been updated.
Main report
6.1.2, 7.2 and
Annex 12,
section 12.4
10
Table 1. Modifications of the impact assessment report in response to RSB comments
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
(B) Key issues
(1) The key proposed policy
measures are not sufficiently
specified to allow for the
assessment of those measures and
whether they can address the
identified problems of EU
competitiveness and strategic
autonomy. Their proportionality is
not adequately demonstrated.
In response to the Board’s comments regarding
the measures designed to respond to Problem 2
and its corresponding drivers, the updated
impact assessment proposes a more refined and
proportionate approach to how sovereignty
would be defined in CADA and how it would
be practically implemented by public
administrations.
PM11 now defines sovereignty along four
levels (instead of one level, which required
exemptions for particular cases), of which
higher levels come with higher sovereignty
demands. This allows for greater granularity in
answering customers’ needs and reflects with
more nuance the reality of available and future
service offerings. The layered approach
reinforces the proportionality of the measure. It
aligns better with the available offer from
European operators. It increases trust and
enables broader cloudification of on-premises
solutions of the most critical use cases, while
creating new market opportunities for sovereign
European providers. PM11 now also includes a
dedicated synthesis table providing a
consolidated view of the gradually demanding
sovereignty criteria.
The assessment of conformity against the
respective sovereignty levels is also more
balanced and cost-effective for all parties
involved, notably in terms of administrative
burden on public authorities. As introduced by
PM15, compliance against sovereignty level 1
would be based on self-assessments conducted
by the service provider. Verifying compliance
of service against sovereignty level 2-3-4 would
be done through third party’s auditors and
verified by national competent authorities
designated by Member States (a mechanism
similar to that used in the Digital Services Act).
We make it explicit that services from non-
European providers can qualify for Level 1 and
Level 2, but not for Level 3 and Level 4. The
measure remains anchored in addressing the
protection of public order in the public sector
(SO4). But since the new Level 3 and Level 4
are reserved to EU providers (on sovereignty
grounds), it nevertheless justifies that this
measure also addresses the overall dependence
Main report:
section 5.2.2.
(and all
subsequent
analytical
sections)
11
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
on non-European cloud and AI computing
services (SO3).
The sovereignty risk assessment newly
introduced by PM15 allows Member States to
identify which public sector use cases within
require the use of which sovereignty level. This
approach reinforces the subsidiarity dimension
of the measure. It empowers Member Stats to
integrate sovereignty criteria in their established
practices and to better choose or change
providers where the sovereignty risk assessment
has determined so. It limits transition costs, by
allowing Member States to consider their own
budgets so that the transition to sovereign
solutions naturally follows their existing
investment schedules. Finally, it avoids big one-
off expenditures by public authorities and
diffuses any risk of a supply crunch since the
uptake of sovereign services will be more
gradual.
In this context, to ensure methodological clarity
and appropriate degree of delineation between
the different measures analysed, the previous
PM19, part of PO2-B, which consisted in
sovereignty audits to be conducted in the public
sector has been removed altogether as it was too
close of the sovereignty risk assessments now
conducted under PM15 and PM21. A new
PM19, now part of PO2-C, has been spun off
from PM21 and consists in the award criteria
made mandatory.
(2) The analysis of effectiveness
and efficiency does not adequately
reflect all the costs (i.e. transition
costs). The benefits appear to be
over-estimated.
The updated approach to sovereignty is
reflected in the way costs and cost savings are
calculated. Based on the feedback of the Board,
the impact assessment now includes in the costs
not only those corresponding to the
establishment of the measures, but also
transition costs directly in the calculations, and
not at a macro-economic level.
In response to the Board’s targeted comments
and based on new evidence, the impact
assessment introduces an improved sensitivity
analysis of the effective price mark-up of
sovereign solutions, as described in section
2.3.4. This results into a more robust range-
based calculation of impacts across the
applicable PMs detailed in Annex 4, section 3.
We now also analyse the costs for a customer to
port individual IT systems from one provider to
Main report:
section 2.3.4.,
6.1.2., 6.1.5
Annexes:
Annex 4,
section 3.
Annex 12
12
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
another. This is summarised in a new box under
section 6.1.2 and detailed in the new Annex 12
provides which also covers the case of
migration (cloudification) of legacy
applications. It includes the
migration/switching journey broken down into
distinct phases and accounts for three gradual
application sizes for which the corresponding
costs are calculated.
In response to the Board’s comments regarding
the previous approach to quantification of
benefits, section 6.1.5. now provides a more
narrowly defined and qualitative analysis of
macro-economic effects. The first part of the
new qualitative analysis focuses on benefits for
public administrations related to improved
sovereignty stemming from the use of sovereign
services, lesser dependency on 3rd-country
providers. The second part provides a
qualitative analysis of the commercial spillover
in terms of improved market opportunities for
providers serving use cases associated with
Level 3 and Level 4, related for example to
improved visibility and perception in terms of
service quality from the perspective of private
buyers.
(3) The analysis of coherence must
be reinforced as there are apparent
overlaps with existing and
upcoming legislation.
This comment is addressed under the more
detailed point 4 of the What to improve section.
(C) What to improve
(1) Given the market situation and
the stated specific objectives, the
interplay between the criteria for
authorising sovereign cloud
services and existing legislation in
third countries should be clarified
and the resulting impacts
analysed. The report must provide
a robust estimate of the number of
entities affected by mandatory
requirements regarding sovereign
solutions, as well as which
sovereign cloud services EU
service providers currently do not
offer. The report should explain
how “countries with
dependencies” will be identified
as well as how and by whom the
criterion of
equivalence/superiority will be
established for the prioritisation of
open source. Without these
The updated PM11 provides under section
5.2.2. a granular descriptionof the four levels of
sovereignty, including a breakdown of the
gradually demanding requirements. PM11 now
also includes clarifications as to which types of
providers could qualify for which sovereignty
level, all in view of the increasing level of
assurance provided between Levels 1 and 4
against the risks defined under Problem 2.
New analysis provided under PM15 in Annex 3,
section 3.15. connects the new stratified
approach to sovereignty with current market
realities with regard to providers in scope of
CADA. The section details the number and type
of providers who could reach the respective
levels of sovereignty. We assess that 59 non-EU
headquartered providers meet Level 1
requirements and would be able to qualify to
Level 1 or 2 should they decide so. 226 EU
headquartered providers could qualify as Level
1 or 2 and would be able to qualify as Level 3
Main report:
section 5.2.2.
(and all
subsequent
analytical
sections),
section 6.1.2.,
section 2.3.2.
Annexes:
Annex 3,
section 3.1.5.,
section 3.2.1.
Annex 12,
Annex 13
13
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
clarifications the impact and
proportionality of the intervention
in terms of costs, including
transition costs, and benefits
cannot be credibly assessed.
should they decide so. Some of these providers
are large companies with proven track record
already in serving governments. We also clarify
that at this moment, few European providers can
achieve Level 4 but more are expected to reach
this level soon.
When it comes to the requested strengthened
analysis of the number of entities impacted by
the obligations under PM21, section 6.1.2.
includes real life evidence-based calculations of
costs of porting applications between two cloud
environments with a view for three distinct
types of public authorities (small, medium,
high) and the implied IT system complexity.
Annex 3, section 3.21. On this basis,three
illustrative cases are presented to show how
costs can vary across different types of public
authorities. These scenarios provide a
benchmark for individual authorities, while the
aggregated cost across all 6400 public entities
possibly in scope is presented under the new
Annex 12.
The new Annex 13 provides a comparative
analysis between offerings of the three leading
hyperscalers and three selected European
providers. Complementing the previous
description of the difference in terms of the
number of offerings under PD1, section 2.3.2.,
the Annex provides a much more granular and
nuanced assessment of this. A set of dedicated
tables shows that the selected European
providers already offer services broadly
equivalent to those of AWS, Microsoft Azure,
and Google Cloud Platform across the core
infrastructure categories that are most widely
used: compute, storage, network, and managed
container orchestration. The comparison has
been extended to office automation suites, one
of the most widely adopted SaaS categories in
the public sector, where credible European and
open-source alternatives are also available.
Where previous interviews have shown that
most public administrations use an average of
twenty services, instead of the whole range
offered by hyperscalers, the Annex
demonstrates that EU providers already offer
the most used services and with a similar level
of functionality.
With respect to the identification of countries of
dependence, PM16, under section 5.2.2. now
details the methodology to establish the
14
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
necessary degree of dependence as well as the
necessary evidence collection and notification
to Member States contracting authorities.
Finally, regarding the approach to establishing
equivalence or superiority of an open source
solution to define its priority over a proprietary
one, PM20, under section 5.2.2. now provides
further operational details related to this
assessment.
(2) The analysis of benefits must
be based on robust and verifiable
assumptions and methodologies.
Although the report acknowledges
the arbitrary choice of certain
assumptions and parameters used
for assessing the impact of key
measures under the preferred
option, those (for example
increased market shares, risk
parameters allocation between the
various options, mark-up
percentage for sovereignty
increased costs) appear to drive
the results of the analysis.
In response to this comment, the modelling has
been reviewed for all PMs where changes were
implemented. Where appropriate, to ensure the
most convincing analysis attainable, we have
shifted from a quantitative to a qualitative
assessment, including with respect to increased
market share of European providers (as
explained in response to the Key issue #2). The
analysis of a price mark-up for sovereign
solutions and the sensitivity analysis thereof has
also been redone and is now based on newly
available empirical evidence, as detailed in
section 2.3.4.
Main report:
Section 2.3.4.,
6.1.5.
(3) The report must provide a more
comprehensive analysis of
technical, functional and
operational consequences of key
measures (i.e. PM 20, 21). As
certain measures (in particular PM
21, embedding PM 11, 15 and 16)
aim at building an EU sovereign
cloud, the impacts of the
limitations of supply from third
countries must be assessed,
including the time and costs
estimated for EU service-
providers to offer equivalent
substitutes. The report should
assess the pass-through of these
cost
As part of the assessment of economic impact
on industry, including SMEs and private sector
essential entities, section 6.1.1. now includes a
dedicated Focus box detailing in qualitative
terms (with several illustrative pieces of
quantitative evidence) the costs for service
providers to develop sovereign services or raise
the conformity level of existing services. This
section should be read in conjunction with the
new Annex 13 providing a comparative analysis
between offerings of the three leading
hyperscalers and three selected European
providers.
Main report:
section 6.1.1.
Annexes:
Annex 13
(4) The report should provide a
detailed analysis of coherence,
in particular of the
authorisation scheme for cloud
and AI computing services with
the security risk assessment in
the Cybersecurity Act and of
public procurement measures
with the upcoming revision of
the public procurement rules.
This should be accompanied by
Section 7.3. now includesa detailed description
of delineation and complementarity between
CADA and the Cybersecurity Act 2 (CSA 2) on
one hand and the upcoming Public Procurement
Act on the other.
With regard to the CSA 2, this section details
the complementarity between the cybersecurity
measures established by the Act and the
sovereignty criteria put forward by CADA. The
risk assessments introduced by the CSA 2 and
the possible prohibitions and other mitigating
Main report:
Section 7.3
Annex 7
15
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
an analysis of the potential
additional legal complexity and
administrative burden.
measure, including with respect to high-risk
vendors, provides a cybersecurity baseline
which CADA builds over by introducing a
targeted approach to mitigating sovereignty
risks specific to the provision and use of cloud
and AI computing services. Similarly, the
gradual sovereignty levels established by
PM11, PM15, PM21 incorporate increasingly
demanding compliance against the three
assurance levels of the future European
cybersecurity certification scheme (EUCS). The
section also explains the absence of any overlap
between the requirements of the CADA
sovereignty levels and the applicable
requirements and those assessed under the
EUCS, which exclusively covers technical
cybersecurity aspects. Finally, PM21 details
further how private sector essential entities
listed under Annex I of NIS 2 will have to
integrate into the cybersecurity risk assessment
they already conduct, the assessment of the risks
stemming from their use of cloud and AI
computing services. Accordingly, section 6.1.1.
clarifies the costs of such entities integrating
sovereignty considerations in their existing risk
assessment activities.
With respect to applicable legal framework of
the Public Procurement Directives, currently
under review, all measures of CADA related to
public procurement are compatible with the
rules in place and those anticipated to be
proposed by the Commission under the new
framework. Notably the measures related to
restricted access to public procurement under
PM15 and PM21 incorporate the justifications
necessary to comply with the requirements of
the exemptions foreseen in the Government
Procurement Agreement of the World Trade
Organisation. Additionally, based on targeted
consultation with DG GROW, the section now
details the interplay between horizontal
measures related to admissible means of
incorporating “EU preference” requirements in
public procurement which the Public
Procurement Act should provide and the sector-
specific approach leveraging award criteria
under PM16 and PM19 of CADA.
16
Table 2. Modifications of the previous version of the impact assessment report in response to RSB first
negative opinion
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
(B) Key issues
(1) The key proposed policy
measures are not sufficiently
specified to allow for the
assessment of those measures and
whether they can address the
identified
problems of EU competitiveness
and strategic autonomy. Their
proportionality
is not adequately demonstrated.
A clearer description of the market historical
context and evolution has been added before
introducing the problems of the intervention.
Moreover, problem drivers 1 - 4 have been
updated to better reflect relevant failures and the
regulatory complexity at play.
Main report:
section 2.1,
section 2.2,
section 2.3
(2) The aspects of EU strategic
autonomy are insufficiently
reflected in the report.
The general objective, and the specific
objectives 3 and 4 have been updated to include
a clearer strategic autonomy dimension.
Main report:
section 4
(3) The proposed policy measures,
including on EU ‘sovereign’
solutions, are not clearly defined
and the causal links in the
intervention logic are not
sufficiently substantiated.
The description of the policy measures has been
improved both in section 5.2 and in annex 4,
notably PM11, PM15 and PM21. The causal
links have been included also in section 5.2.
Furthermore, the table in section 5.2.3 has been
updated to show clearer the link between the
problem drivers, the problems, the policy
options, policy measures and specific
objectives.
Finally, a new table has been added in section
5.2.4 to better demonstrate which aspects of
each problem driver the proposed policy
measure address.
Main report:
section 5.2
(notably 5.2.1,
5.2.2, 5.2.3 and
5.24)
(4) The analysis of effectiveness
and efficiency does not
sufficiently reflect the adjustment
costs, technical feasibility and
unintended consequences linked
to the proposed policy measures.
Adjustment and administrative costs have been
cross-checked and updated in the main report
under section 6 and Annex 4. While quantitative
estimates of all possible adjustment costs and
unintended consequences were at times limited
by data availability, the effectiveness and
efficiency analysis have been revised and the
discussion to address transition costs, technical
feasibility and possible side effects of the
proposed measures, expanded. This has been
added under section 7.1 and 7.2.
The sensitivity analysis in section 7.5 has also
been updated to clearly present the uncertainties
linked to the proposed policy measures.
Main report:
section 7.1,
7.2, 7.5
Annex 4
(C) What to improve
1) The report should better analyse
the root causes of the insufficient
cloud capacities in the EU and the
declining market share of
European services providers.
Section 2.1 (Problem context) has been
reformulated to include an analysis of the main
drivers for the current situation of the declining
market share of EU providers.
Main report:
Section 2.1,
Section 2.2,
Section 2.3
17
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
Based on an analysis of the main
drivers of competitiveness, such as
price, scope of services,
innovation and quality, including
cyber security, it should analyse
potential market failures. It should
also analyse potential regulatory
failures based on available
evidence of investment decisions.
It should analyse more thoroughly
the current practices and
regulatory frameworks in Member
States and how they impact the
Internal Market. It should describe
the risks of reliance on non-EU
providers and the potential trade-
offs with aspects of strategic
autonomy.
Problem drivers 1-4 have been updated to
showcase better the different failures, such as
market and regulatory related ones. Problem
driver 1 has been updated seeking to respond to
the aspects related to the main drivers of
competitiveness. Problem driver 2 has also been
updated and reformulated to better substantiate
the fragmentation of the different regulatory
frameworks across MS. Problem driver 3 has
been updated to discuss the risks of reliance to
non-EU providers, notably for the public sector.
2) The report should be clear
whether – and if yes how - the
objectives of this initiative relate
to the EU’s strategic autonomy.
What level of autonomy is to be
achieved and in which domains?
The content of policy measures
should be clearly defined,
including the criteria for
“sovereign solutions” and “EU-
made”. The report should better
justify how the target of market
share of EU providers is set.
The general objective as well as specific
objectives 3 and 4 have been updated to include
a clearer strategic autonomy dimension.
The explanations of the main elements of the
policy measures have been updated in section
5.2. Further explanations and details have been
included in annex 4 under each PM11, PM15,
PM16 and PM21 to clarify this.
Section 6.1.5 has been updated to include a
justification of how the increase of market share
from 15% to 30% has been used and what has
been the rationale for such an assumption.
A new section has been included in Annex 4
(section 8.3) aiming to better explain the
feasibility of the move towards a market share
of 30% by EU providers per customer segment.
Main report:
Section 4,
Section 5.2,
section 6.14
Annex 4
section 7.2,
section3 and
section 8.3
3) The report should assess the
potential impacts related to the
criteria for “sovereign solutions”.
It should analyse in sufficient
detail the levels of sovereignty to
be reached and be transparent
about technical prerequisites
needed. The report should assess
all the adjustment costs and
quantify them to the extent
possible and be clear on who will
bear these costs. The report should
provide a more comprehensive
analysis of unintended
consequences, including on cyber
security, in particular related to
The report has been modified in section 5.2 and
annex 4, where a clearer definition of the
sovereignty criteria is provided.
Annex 4 and section 6.1 have been amended to
clarify which adjustment cots have been
considered for the sovereignty aspects.
Additional explanations and justifications have
been included to motivate the rationale of why
certain adjustment costs have not been
considered.
The sovereignty criteria that have been
discussed in the report will result in a unique
level of sovereignty. However, the report has
been edited to include the notion of
“sovereignty-readiness” in section 5.2 and
Main report:
section 5.2,
section 6.1,
section 2.2.3.
Annex 4
section 3,
description of
PM11, PM15
and PM21
18
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
sovereignty and open-source
requirements.
annex 4, to clarify that not all cloud and AI
computing service providers will start from the
same situation and what this may entail.
The unintended consequences of cybersecurity
and open source have been described in PD3,
which has been adjusted accordingly.
4) The analysis of benefits should
be based on robust assumptions, in
particular for PO2B and PO2C
where stated benefits are very
high. For example, all
implications should be factored in
the analysis, such as investments
needed to maintain cyber security,
and the scale needed to make them
viable in economic terms. In terms
of the environmental footprint, the
report should explain the
mechanisms leading small(er) EU
cloud providers to significantly
reduced Power Usage
Effectiveness values and how
European-funded R&D initiatives
are expected to deliver superior
results when compared to the
efforts of global hyperscalers.
For PO2-B and PO2-C in annex 4 a clarification
has been added both under PM15 (applicable to
PO2-B) and PM21 (applicable to PO2-C) on the
costs that have been included, the ones that have
been excluded and the rationale behind them.
The report has also been updated to include
additional considerations on the environmental
impact of data centre capacity, with clearer
explanations of how EU-funded R&D
initiatives and other Policy Measures are
expected to accelerate the adoption of energy-
efficient technologies within data centres and
thus leading to a long-term reduction in PUE
levels, compared to the other options and
baseline. The timing dimension of this
reduction was also further considered as
programmes require time to develop and market
new solutions that would optimise energy-
efficiency. The report does not suggest that EU-
funded R&D is expected to outperform or
deliver superior results when compared to
hyperscaler-led investments in innovation.
Rather, the discussion focuses on how public
R&D support and related policy instruments
could help facilitate the diffusion and adoption
of current and new technologies across the
wider data centre ecosystem, and in particular
among non-hyperscale operators, which have
been found to exhibit higher average PUE levels
than hyperscale facilities.
Main report:
Section 6.3
Annex 4
5) The report should allow for
verifiability of the economic
modelling. It should quantify and
monetise only benefits which can
be clearly linked to the policy
measures and whose
quantification can be substantiated
by evidence. Currently the benefit
cost ratios presented in the report
are not plausible. After revision of
the analysis of costs and benefits,
the sensitivity analysis should be
reviewed in order to allow to
assess uncertainty related to the
most impactful variables behind
the projections.
The cost benefit analysis has been revised. The
benefits modelled in annex 4 include only those
that can be quantified and monetised; however,
as an additional exercise, the indirect benefits,
not quantified, had already been included in the
previous version of annex 4 under most policy
measures.
Most of the assumptions remain in line with
what had been previously presented. These
were validated in one-to-one interviews, CATI
surveys and the final validation workshop of the
study led by Technopolis. In the event of
uncertainty, assumptions were discussed in the
final validation workshop. Except for a couple
Main report:
section 7.5
Annex 4
19
Opinion of the RSB by point How comments have been addressed Sections in IA
report/Annex
of them (authorisation effort and PUE) all other
assumptions were confirmed by attendees.
The explanations for the sensitivity analysis
have been improved in section 7.5
6) The coherence,
complementarity and potential
synergies with other initiatives -
such as the Cybersecurity Act,
Data Act or the public
procurement revision - should be
analysed in more detail.
Additional clarifications on external coherence
with the Cybersecurity Act, the Data Act and the
public procurement revision have been added to
section 7.3, on top of the existing analysis in
Annex 7.
Section 7.3 of
the main report
4. EVIDENCE, SOURCES AND QUALITY
The impact assessment is based on several sources, including:
1. Stakeholder consultation activities (see Annex 2)
2. External support study carried out by an independent consultant (Technopolis Consulting
Group Belgium coordinator - SPECIFIC CONTRACT No 4500075392 implementing
framework contract No FW-00141707 CNECT/2022/OP/0036 - Framework Contract for
the provision of Studies and related services on digital policy issues Study: cloud and AI
– STUDY 2024-046)
20
ANNEX 2: STAKEHOLDER CONSULTATION (SYNOPSIS REPORT)
1. INTRODUCTION
This annex provides a summary of the consultation activities undertaken to gather evidence
and ensure that the public interest is well reflected in the design of potential policy measures
for cloud and AI computing services development within the EU. It presents the range of
stakeholders consulted, describes the main consultation activities and provides an analysis of
their views and the main issues they raised. The objective of the consultation activities was to
collect the views and experience of stakeholders on the key problems and associated drivers
related to cloud and AI infrastructures in Europe, including current and future capacity needs,
sustainability, investment conditions, and the availability of secure EU-based providers and
services. In addition, it aimed to gather information and their opinions on the definition of
relevant policy objectives linked to those problem areas and the identification, definition and
screening of policy measures that could eventually be incorporated into policy options for this
impact assessment as well as gather information and opinions on their likely impacts.
2. OVERVIEW OF CONSULTATION ACTIVITIES
2.1 Public Consultation on the Cloud and AI Development Act
The Public Consultation (‘PC’) consisted of an online questionnaire available in English,
French and German; and a Call for Evidence (CfE) for stakeholders to submit detailed position
papers outlining their views and recommendations on the objectives and proposed actions
envisaged by the cloud and AI Development Act, while also giving direct evidence to inform
the design of policy options. The questionnaire and the Call for Evidence ran from 9 April 2025
to 3 July 2025. In addition to the OPC, the following consultation activities were conducted:
• Targeted Workshops with relevant stakeholders;
• Ad-hoc contributions and targeted consultations with Member States, small-and-
medium sized enterprises (SMEs) and other relevant stakeholders;
• Targeted consultations with industry representatives and public authorities through
surveys and workshops as part of a study commission by the Commission to support
the impact assessment.
2.2 Study to support the impact assessment
The Commission contracted a study to a consortium led by Technopolis Group to provide
empirical evidence on the currently available computing capacity across Europe and to assess
present and projected future cloud computing needs within the EU. The study addresses
knowledge gaps in cloud computing development and provides policy recommendations to
support the Impact Assessment. The study incorporates multiple stakeholder engagement
activities, including targeted interviews (over 60), surveys (with over 250 replies) and
workshops (over 100 participants). The results of those activities are also included in this annex.
2.3 Overview of workshops, seminars and roundtables:
• Workshop with 70 industry actors on industry needs for cloud computing capacity for
AI demand (4 October 2024, Brussels, Belgium);
• Seminar with industry, academia and public authorities on the economic dynamics of
the AI stack (4 April 2025, online);
• Workshop with industry on the rationale and policy options that underpin the Act (14
May 2025, Berlin, Germany);
21
• Workshop with the Information Technology Industry Council (15 May 2025, Brussels,
Belgium);
• Roundtable on investment in cloud compute with financial investors (20 May 2025
Brussels, Belgium);
• Presentation on future EU cloud and AI policy with industry (3 June 2025, Bilbao,
Spain);
• Workshop on the preliminary findings of the accompanying study of the cloud and AI
Development Act (6 June 2025, Brussels, Belgium);
• Roundtable on policy measures to facilitate data centre integration to the EU grid (18
June 2025, Brussels, Belgium);
• Roundtable with European Cloud Service Providers’ CEOs on developing sovereign
cloud offerings in the EU (23 June 2025, Brussels, Belgium);
• Workshop during the 6th General Assembly and Alliance Forum of the European
Alliance for Industrial Data, Edge and Cloud (8 July 2025, Brussels, Belgium).
• Roundtable with American Chamber of Commerce (22 October 2025, Brussels,
Belgium);
• Workshop on the final findings of the accompanying study of the cloud and AI
Development Act (12 November 2025, Brussels, Belgium);
2.4 Awareness raising events in Member States and international outreach
The Commission conducted stakeholder activities across several Member States, participating
in bilateral meetings with a diverse range of stakeholders in Poland, Germany, the Netherlands
and Spain to discuss the problem drivers and core elements of the proposed legislation. The
Commission also maintained dialogue with Member States’ relevant authorities through the
European Alliance for Industrial Data, Edge and Cloud (‘Cloud Alliance’) and bilateral
meetings in Brussels. Furthermore, the Commission engaged in bilateral discussions with third
countries, including Japan, Switzerland and the United Kingdom, to present and discuss the
considered policy options while gathering insight on best practices.
2.5 Bilateral meetings with industry, academic institutions, think tanks
The Commission conducted over 100 bilateral meetings with a diverse array of stakeholders,
including industry representatives, academic institutions, think tanks and civil society
organisations to gather feedback on the design of policy options for the cloud and AI
Development Act. These meetings facilitated an open dialogue with both European and
international partners regarding their perspectives, concerns, and recommendations across the
different policy pillars of the cloud and AI Development Act.
3. PARTICIPANTS AND METHODOLOGY
3.1 Participants
A total of 436 responses to the PC were received: 243 for the consultation survey and 193 for
the Call for Evidence. The largest share came from EU citizens, accounting for 33% of all
responses. This was followed by companies (28%) and associations and trade unions (18%).
Additional contributions came from academic and research institutions (9%), civil society and
consumer organisations (7%), and public authorities (4%). Only 2% of all respondents were
non-EU citizens.
In terms of geographical origin, 90% of respondents (393) were based within the EU. The
remaining 10% (43 responses) came from non-EU countries, including the US (25 responses),
United Kingdom (10), Switzerland (2). The Member States with the highest participation rate
22
were Spain (78 responses; 18%), Germany (57; 13%), and both France and Belgium (56; 10%
each). No responses were received from Bulgaria, Croatia, Latvia, or Malta.
The 243 respondents to the questionnaire comprised 66 companies and private businesses
(27%) including 17 SMEs, 96 EU citizens (39%), and 13 public authorities (0.4%). Out of the
66 companies and private businesses, 57% were AI developers or deployers, 56% were
cloud/edge/AI users, 48% were cloud/edge/telecommunications providers, 45% were data
centre operators, and 7% were financial institutions.2
3.2 Methodology
The responses to the OPC were subjected to quantitative and qualitative analyses. In the
quantitative analysis, the respondents were ordered into the main categories considered relevant
for the analysis. The answers were mapped and analysed to identify patterns as regards “who
thinks what?”. When considering the views of respondents on questions where they were asked
to rank on a scale of 1 to 5 (1 = not very important, 5 = very important), only responses ranked
4 or 5 were considered when determining respondents’ positions on each issue. The qualitative
analysis of open responses and position papers was carried out in two stages. In a first step, the
content of the submissions was synthesised to capture the main experiences and feedback of
respondents. In a second step, these summaries were clustered according to the type of issue
raised, so that contributions addressing comparable concerns were considered together
irrespective of their specific wording. This approach allowed for a structured overview of the
main perspectives of the respondents.
4. ANALYSIS OF THE STAKEHOLDER CONSULTATION
The respondent input gathered through the OPC highlighted four core policy challenges, which
directly informed the structure of potential policy options under consideration for the cloud and
AI Development Act. The main challenges and problems identified through the OPC are:
1. The strategic dependency of EU businesses and public authorities on non-EU providers
of AI and cloud services, raising questions about sovereignty, resilience, and
competitiveness;
2. Structural and regulatory bottlenecks hindering the expansion of data centre capacity
across the EU, particularly in the face of growing AI-related demand;
3. Barriers to the digital transformation of the public sector, including limited capacity,
fragmented compliance, and complex procurement procedures; and
4. The absence of a common EU-level definition and operational criteria for 'sovereign'
cloud and AI computing services, leading to legal uncertainty and inconsistent
implementation across Member States.
4.1 EU businesses and public authorities’ dependency on non-EU cloud and AI
providers
The hyperscalers are recognised for their ability to deliver comprehensive, ready-to-use
business services that European providers have yet to match. For example, in terms of AI
computing services, out of the respondents that provided an answer in the questionnaire, 48%
indicated that only hyperscalers could offer those services and only 3% said that European
providers were capable of offering those services. Despite this reliance, respondents
emphasized the critical need for trusted EU-based cloud service providers, particularly for
2 The sum of the percentages is higher than 100 as one company may be involved in more than one sector.
23
processing sensitive data. When selecting cloud service providers, users prioritise cybersecurity
(97%), data location (87%), connectivity (81%), and innovation (81%) as their primary criteria.
Among respondents who provided an answer, when selecting a cloud provider, 82% were
concerned that provider was headquartered in a third country that poses specific cybersecurity
threats to the EU. In addition, 58% were concerned about vendor lock-in and 42% considered
as a challenge the limited access to computational capacity in the EU. Finally, 89% considered
that the location of the data was one of the main factors impacting their choice of
Cloud/Edge/AI provider.
In the CfE, public authorities showed widespread concern about Europe’s dependency on non-
EU cloud and AI providers; however, there was some divergence on the solutions. Some
authorities described the situation as a “market failure” due to the lack of viable European
alternatives in key service areas and called for supplier diversification. Other public authorities
advocated risk-based approaches to avoid protectionist outcomes. Some EU citizens questioned
the structural reliance on private actors and proprietary systems while advocating for an EU-
first approach and arguing that transformation cannot succeed while relying on foreign vendors
whose strategic incentives may not align with public interest objectives.
4.1.1 Policy measures
In the CfE, respondents were generally in favour of simplifying permitting requirements and
one-stop-shop to facilitate the deployment of data centres in the EU (see Section 4.2.1 below).
European companies and business associations expressed strong interest in setting up
sovereignty requirements and introducing “Buy EU” criteria in public tenders, whereas their
global counterparts cautioned against digital protectionism and argued digital trust could be
achieved without exclusion based on geographical origin (see Section 4.4.1).
4.2 Bottlenecks to expand (sustainable) data centre capacity within the EU
Data centre operators (‘operators’) demonstrated strong investment intentions in AI
infrastructure, with 80% of them planning to expand capacity within the EU. Among operators,
77% replied future infrastructure would be dedicated to general purpose computation, 70%
indicated it would be used for AI model inference, 60% on AI model training, and 33% on edge
AI.
Despite this enthusiasm, operators face substantial barriers that may hamper their expansion
plans. Complex and lengthy permitting procedures emerged as the primary challenge (87%),
followed closely by limited access to energy resources (83%). The prevalence and severity of
these obstacles is illustrated by the fact that 57% of operators need on average between 3 and
5 years for the data centre to be operational, and 13% of them need between 5 and 7 years.
The presence of those bottlenecks is supported by the responses to the CfE where connection
to the electrical grid was also identified as a key bottleneck by several respondents. One
business association indicated it could take up to 7 or 8 years for transmission access and two
operators mentioned grid connection could take up to a decade. In addition, one business
association considered those problems were exacerbated by growing local opposition to the
construction of data centres, a phenomenon often termed “not in my backyard”.
The factors considered by operators regarding building new data centres or expanding existing
infrastructures are energy availability (91%), as well as long approval times (83%), permitting
24
procedures (82%) and the complexity of the regulatory framework (75%).3 Those factors reflect
the importance given by operators to grid connections and the entire permitting process.
Access to finance also appears to be a common hurdle with 50% of operators and 53% of
Cloud/Telco/edge providers considering high interest rates as a limitation when expending their
capacity in the EU.4 In addition, several SMEs expressed mounting concerns over high
operational costs when it comes to infrastructure expansion in the EU.
In the CfE, public authorities agreed on the need for increased data centre capacity while
stressing different strategic priorities. Industry associations identified fragmented permitting
regimes across Europe, limited energy access, and regional disparities as persistent constraints
on capacity scaling. It is important to note that in the CfE, some EU citizens challenged the
expansion narrative entirely, criticizing the approach for the ignoring social and environmental
costs of AI and advocating for binding limits on resource consumption rather than efficiency
improvements alone.
4.1.2 Policy measures
Overall, across all respondent groups, 45% considered that having a one-stop-shop service to
deal with permits at different administrative levels should be a priority action at EU level to
boost the availability of sufficient and adequate cloud capacity for AI workloads; and 46%
considered it essential to reduce the amount of time needed to obtain the necessary permits and
environmental authorisations. Finally, 55% were in favour of creating expedited approval
mechanisms and clear conditions for critical or strategic projects.
Table 2. EU policy measures on simplification of infrastructure permitting procedures
Policy Measure Companies
(SMEs)
Data
Centre
operators
Public
Authorities
EU
Citizens
Business
associations/
trade unions
CSOs5
One stop shop service
for managing permits
59% (41%) 83% 46% 43% 56% 15%
Reduce time to obtain
permits and
environmental
authorisations
71% (53%) 93% 23% 38% 60% 21%
Expedited approval
mechanisms for
critical/strategic
projects
79% (71%) 97% 54% 49% 57% 27%
In the CfE, several industry associations and companies recommended the creation of EU-level
zoning guidelines, one-stop-shop permitting hubs, and fast-track authorisation for sustainable
projects. This was echoed by investment respondents who stressed that streamlining permitting
procedures would significantly enhance the sector's attractiveness for capital allocation.
In terms of sustainability, 69% of respondents supported funding for R&D of energy-efficient
technologies. This was echoed in the CfE by a Member State’s government which suggested
R&D&I for data centre technologies enhancing sustainability, as well as several business
3 The respondents who did not provide an answer to that question were not included to calculate the percentage of
respondents. 4 The respondents who did not provide an answer to that question were not included to calculate the percentage of
respondents. 5 The category Civil Society Organisations (CSOs) include NGOs, consumer organizations, environmental organizations,
and academic/research institutions.
25
associations and companies. In addition, 62% were in favour of tax incentives for using
sustainable technologies. In terms of EU actions that should be prioritised to boost the
availability of sufficient and adequate cloud capacity for AI workloads, 72% supported the
creation of clear environmental compliance requirement et EU level, and 66% were in favour
of unified guidelines at EU level for energy efficiency for computing infrastructure.
Table 3. EU policy measures to address sustainability aspects of data centres
Policy measure Companies6 DC
operators
Public
authorities
EU
citizens
Business
associations /
trade unions
CSOs7
Clear environmental
compliance
requirements
64% 73% 46% 86% 60% 70%
Addressing energy
availability
73% 80% 54% 77% 50% 61%
Addressing land
availability
47% 47% 23% 65% 47% 52%
Funding for research
and development of
energy-efficient
technologies
61% 57% 54% 83% 60% 58%
Tax incentives for using
sustainable technologies
70% 73% 54% 67% 60% 39%
In the CfE, multiple business associations and companies suggested the introduction of targeted
tax incentives to build sustainable and energy efficient cloud infrastructure. Some considered
this would be more useful than duplicating the current environmental regulations. One
company suggested to create tax incentives for data centre operators that implement sustainable
technologies, including waste heat reuse or immersion cooling. In addition, one business
association was in favour of public-private partnerships to build sustainable, innovative cloud
infrastructure. Some public authorities recommended strategic infrastructure funding to be tied
to interoperability and climate goals.
Overall, this shows a strong support by different respondent groups to address existing
bottlenecks in the deployment of data centres and to ensure those are sustainable and energy
efficient.
4.3 Barriers to public sector transformation and public procurement
Cloud adoption in the public sector is already widespread with 85% of respondents (i.e. public
authorities) storing data in cloud environments. This data often includes sensitive information
(46%), commercially sensitive material (38%), or content related to public security and health
(23%). The sensitive nature of this data creates heightened concerns about exposure to third-
country authorities and the cybersecurity risks associated with cloud service providers
headquartered in certain jurisdictions.
Such a problem is not theoretical as 69% of public authorities acknowledged that their procured
providers were subject to non-EU jurisdictions including laws with extraterritorial effect. In
addition, 77% of public authorities cited security risks as one of the primary obstacles
preventing broader public sector cloud adoption, alongside regulatory barriers, including
6 In each policy measure, companies include data centre operators, cloud/edge/telecommunications providers, cloud/edge/AI
users, AI developers and AI deployers, and financial institutions. 7 The category Civil Society Organisations (CSOs) include NGOs, consumer organizations, environmental organizations, and
academic/research institutions.
26
procurement requirements (68%), and insufficient technical capacity to assess and procure
appropriate cloud services (62%). In addition, 62% of public authorities mentioned fear of
vendor lock as a challenge in the adoption of cloud.
These concerns are echoed by EU citizens since 76% considered public services in the EU
should not be allowed to store citizen data with non-EU cloud providers, only 9% considered
it depended on the type of service or data.
In the CfE, regional and local authorities identified administrative barriers such as fragmented
national audits, inconsistent certification requirements, and burdensome permitting processes
as core obstacles. Overall, multiple respondents emphasized that underlying market structure
issues such as vendor lock-in and limited portability represent real barriers.
4.1.3 Policy measures
To address these barriers, public authorities advocated for coordinated EU-level action, with
84% supporting the establishment of cybersecurity guidelines, 77% supporting the adoption of
standards, open specifications, and mechanisms to ensure interoperability. Finally, 77%
advocated for a mechanism to allow federation of cloud services across public administration
within and across Member States. In terms of policy actions to address the current problems
faced by public administrations regarding procurement, 85% of public authorities support the
creation of guidelines with standard criteria to procure cloud services and 69% support
guidelines with standard award criteria. Regarding cloud and AI computing services more
generally, 61% would support the creation of a marketplace for cloud and AI computing
services.
In addition, overall, 74% of respondents support an open source software ecosystem. The
majority of public authorities, academic institutions and EU citizens would favour an obligation
to release the code developed with public money onto open-source repositories (61%), as well
as a common open-source licensing schema across the EU (64%).
Table 4. EU policy measures to support procurement of cloud and AI computing services
Policy Measure Companies8 Public
Authorities
EU
Citizens
Business
associations / trade
unions
CSOs9
Guidelines with standard
criteria to procure cloud
services
50% 85% 61% 20% 55%
Standardised tender
vocabulary and requirements
38% 54% 36% 20% 30%
Guidelines with standard
award criteria
32% 69% 26% 17% 39%
Criterion for solutions with
added value and innovation
27% 23% 30% 33% 18%
Improvement of skills and
capabilities
32% 69% 41% 40% 21%
Marketplace of cloud and AI
computing services
26% 62% 31% 20% 27%
Supporting an open source
software ecosystem
62% 31% 92% 73% 60%
8 In each policy measure, companies include data centre operators, cloud/edge/telecommunications providers, cloud/edge/AI
users, AI developers and AI deployers, and financial institutions. 9 The category Civil Society Organisations (CSOs) include NGOs, consumer organizations, environmental organizations,
and academic/research institutions.
27
In the CfE, some national and local governments as well as several public authorities
emphasized empowering public buyers through model clauses and guidance to help public
administrations to procure cloud services. However, some warned against mandatory
obligations that could conflict with EU procurement principles or overly limit their flexibility.
Some companies framed public procurement as a critical but underused leverage point, though
they diverged sharply on solutions. European companies argued that current frameworks fail
to empower buyers to choose sovereign, sustainable, or interoperable solutions, calling for
uptake targets for trusted EU cloud services and criteria based on sovereignty and
environmental performance. Some argued for a diversification of cloud supply, including
strengthening EU-based alternatives and governance but without restricting the use of non-EU
providers.
In contrast, global providers support the EU's ambitions to bolster innovation and security but
warn against overly restrictive sovereignty rules that could hinder competitiveness and limit
access to cutting-edge technologies if they exclude providers based on corporate origin.
Sovereignty requirement for critical use cases must be carefully defined to avoid fragmentation.
They advocate for a pragmatic and targeted approach to defining ‘sovereignty’ to avoid
unintended consequences such as straining the EU's cloud ecosystem or fragmenting the Single
Market. To strengthen resilience and avoid vendor lock-in, several companies and business
organisations recommend that public authorities adopt multi-cloud strategies for non-critical
use cases.
In general, all respondent groups, especially public authorities and both EU and non-EU
companies, showed strong support for open source to foster sovereignty, interoperability and
competition.
4.4 Lack of common concept and operational criteria for sovereign cloud and AI
computing services
In the CfE, respondents consistently identified the absence of a common EU definition and
certification framework for sovereign cloud and AI computing services as a fundamental barrier
to building trust, driving adoption, and enabling cross-border scalability. The lack of
harmonized cybersecurity standards particularly limits providers’ ability to scale services
across Member States, creating fragmentation that undermines the Single Market potential.
This regulatory gap has intensified demands for trusted EU-based cloud providers capable of
delivering sensitive data services with robust guarantees of legal and operational sovereignty,
interoperability, and protection from extraterritorial interference. These concerns align with
public sentiment, as EU citizens expressed significant wariness about foreign control of public
data where 85% reported low to very low trust in providers based outside of the EU, and 76%
considered public services in the EU should not be allowed to store citizen data with non-EU
cloud providers.
In the CfE, there was near-universal agreement that the term ‘digital sovereignty’ lacks a clear
and actionable definition, particularly in the context of cloud and AI computing services. Public
authorities acknowledged that the concept of ‘sovereign cloud’ remains ill-defined, creating
uncertainty across the policy landscape. Several EU citizens emphasized that sovereignty
should ensure services critical to public continuity cannot be disrupted unilaterally by non-EU
actors.
4.1.4 Policy measures
Among public authorities, 77% would support a criterion to ensure sovereignty, autonomy,
resilience and availability. In the CfE, public authorities specifically called for a clear, uniform
28
definition of cloud sovereignty that encompasses data ownership, exclusive EU jurisdiction,
and comprehensive contractual safeguards. Many advocated for deploying EU-level secure
cloud offerings comparable to existing frameworks like SecNumCloud or EUCS High+. Some
respondents repeatedly emphasized the urgency of finalising the EUCS certification scheme
and integrating it into procurement frameworks, while leveraging public procurement rules to
reinforce sovereignty requirements and mitigate national security risks. Several public
authorities proposed extending NIS2 scope to include suppliers serving essential public
services.
Some public authorities and national governments advocated for functional, risk-based
approach to sovereignty as full technological autonomy for all cases is not realistic. Sovereignty
should be anchored in clear operational criteria focused on retaining control over data and
operations, switching providers, and avoiding undue foreign legal influence rather than
mandating EU-only solutions. Regulatory bodies emphasized that sovereignty must not become
a pretext for fragmentation, instead advocating for open architectures, multi-cloud approach,
and guaranteed access to key inputs for challenger firms.
Among industry associations, there was a split between those calling for functional definitions
based on EU legal jurisdiction and vendor independence through federated European solutions,
and those warning that overly restrictive criteria could impede cross-border services and harm
innovation. Several European companies pushed for enforceable criteria involving EU
ownership, legal immunity from third-country laws, and operational autonomy, with some
proposing sovereignty frameworks differentiated by use case criticality. In contrast, non-EU
providers urged against rigid definitions, recommending non-discriminatory and risk-based
criteria grounded in technical safeguards rather than ownership or nationality.
Table 5. Level of support for EU policy measures to protect against unlawful access to sensitive data by
third-country legislation with extraterritorial reach and risk associated with supply chain dependencies
Policy Measure Companies10 Public
Authorities
EU
Citizens
Business
associations /
trade unions
CSOs11
Criterion ensuring sovereignty,
autonomy, resilience and
availability
45% 77% 63% 23% 52%
Criteria to narrowly identify
highly critical use cases for
public procurement
30% 62% 48% 20% 52%
Pursue international cooperation
with third countries that address
such risks
45% 69% 40% 70% 27%
Criteria to differentiate between
third countries depending on
whether they pose specific threats
to the Union
29% 54% 44% 27% 24%
10 In each policy measure, companies include data centre operators, cloud/edge/telecommunications providers, cloud/edge/AI
users, AI developers and AI deployers, and financial institutions. 11 The category Civil Society Organisations (CSOs) include NGOs, consumer organizations, environmental organizations,
and academic/research institutions.
29
ANNEX 3: WHO IS AFFECTED AND HOW?
1. PRACTICAL IMPLICATIONS OF THE INITIATIVE
This initiative aims at providing a framework to ensure the functioning of the internal market
for cloud and AI computing services in the Union and secure the necessary conditions for the
Union’s competitiveness and strategic autonomy. The preferred policy options and measures
are expected to increase computing capacity deployed in the EU through innovative and
sustainable technologies, ensure attractive conditions for the deployment of such computing
capacity, decrease the overall reliance on non-European cloud and AI computing services and
contribute to the protection of public order by enhancing resilience of supply of cloud and AI
computing services in particular for the public sector.
The key stakeholder groups concerned include data centre operators, cloud and AI computing
service providers, private sector entities operating in sectors listed under Annex I of NIS2,
national and local public authorities, and the European Commission. However, all businesses
that rely on digital services across a wide range of sectors, as well as citizens will also be
indirectly impacted, albeit in varying degrees. The following table presents an overview of the
main impacts and stakeholders affected for the preferred options.
Table 6. Main impacts and stakeholders affected
Affected stakeholder Main impacts
Costs Benefits (Direct and indirect)
Businesses Data centre
operators ⎯ Direct compliance costs:
⎯ Administrative costs (e.g.
preparing applications and
ensuring compliance with
funding rules; paperwork
for fast track permitting)
⎯ Adjustment costs (e.g.
adjustments to meet
funding eligibility criteria
and access the areas)
⎯ Improved predictability of
regulatory direction (e.g.
coordination efficiencies and
reduced information asymmetries)
⎯ Reduced fragmentation of national
approaches and administrative
simplification, e.g. fewer bilateral
interactions, clearer requirements,
and coordinated processing
⎯ Reduced permitting delays (e.g.,
faster time to market and higher
NPV, labour time saved)
⎯ Direct financial support for
innovation and deployment
⎯ Energy savings
⎯ Possibility for improved access to
investment opportunities (CAPEX
de-risking)
⎯ Easier cross-border operations
30
Affected stakeholder Main impacts
Costs Benefits (Direct and indirect)
Cloud and AI
computing
service
Providers
⎯ Direct compliance costs:
⎯ Administrative costs (e.g.
providing information for
repository; participation to
audits; preparation of
applications for funding
scheme)
⎯ Adjustment costs (e.g.
alignment with sovereignty
criteria; participation to
public procurement
procedures; eventual
alignment with standards)
⎯ Improved predictability of
regulatory direction (e.g.
coordination efficiencies and
reduced information asymmetries)
⎯ Reduced fragmentation of national
approaches
⎯ Reputational gains from alignment
with EU standards
⎯ Easier cross-border operations
⎯ Faster procurement
⎯ Direct financial support for
innovation
⎯ Improved access to investment
opportunities
⎯ Reduced marketing costs
⎯ Increased revenues
⎯ Easier uptake of innovative
solutions for cloud and AI
computing services
⎯ Productivity gain and business
modernisation
⎯ Improved business continuity
Private sector
entities
operating in
sectors listed
under Annex I
of NIS2
− Direct compliance costs:
⎯ Administrative costs (i.e.
expanded cybersecurity risk
assessment to address non-
technical risks)
Public
authorities
National and
local public
authorities
⎯ Direct compliance costs:
⎯ Administrative costs (e.g. data
collection of fast-track areas;
sovereignty risk assessments;
drafting of national strategies;
updates of procedures/
templates related to public
procurement; reporting
activities to maintain the open
source programme office and
on the implementation of
national strategies)
⎯ Adjustment costs (e.g.
redesign of permitting
processes and reprioritising
planning resources for fast-
track areas; alignment of
national measures with EU-
level rules; integrating risk
assessment outcomes within
procurement; transition costs
linked to new cloud and AI
computing services providers;
participation in the public
sector cloud federation;
alignment of procurement
criteria; set up and
management of the
infrastructure for the open-
source repository)
⎯ Enhanced dialogue with industry
⎯ Better monitoring of sector
developments
⎯ Alleviated constraints and delays
resulting in reduced pressure from
industry over bottlenecks
⎯ Reduced time dedicated to
individual data centre projects
⎯ Saved effort from common
mechanism and sharing of idle
capacity
⎯ Faster and more reliable
procurement processes
⎯ Reduced screening efforts and
efforts needed to verify eligibility
⎯ Streamlined verification of
compliance
⎯ More sovereign cloud and AI
computing services
⎯ Reduction of proprietary license
expenditures and lower licensing
costs
⎯ Reduced duplication of work and
efforts
⎯ Reduction of total cost of
ownership for IT systems
⎯ Shorter procurement time for
innovative solutions; reduced time
needed to find providers
31
Affected stakeholder Main impacts
Costs Benefits (Direct and indirect)
European
Commission ⎯ Administrative costs (e.g.
design and monitor EU-level
funding instruments; setting up
and overseeing the repository of
audited sovereign services;
financial and coordination
support)
⎯ Adjustment costs (e.g., setting
targets; monitoring progress;
new supervisory capacity to
enforce EU-level targets; design
of training programme; setting
up the certification framework;
developing and deploying the
federation platform; setting up
the voucher scheme for SMEs;
setting up the cloud and AI
toolbox)
⎯ Enforcement costs (e.g. running
EU-level infrastructure, defining
targets and ensuring
compliance)
⎯ Harmonisation of practices and
approaches across the EU
⎯ Increased ability to steer the
market towards EU priorities and
accelerated innovation
⎯ Reduced duplication of efforts at
national level
⎯ Faster and more reliable
procurement processes;
⎯ Reduced screening efforts and
efforts needed to verify eligibility;
⎯ Streamlined verification of
compliance;
⎯ Reduction of proprietary license
expenditures and lower licensing
costs;
⎯ Reduced duplication of work and
efforts;
⎯ Reduction of total cost of
ownership for IT systems;
⎯ Shorter procurement time for
innovative solutions; reduced time
needed to find providers
Consumers
and
citizens
Consumers and
citizens ⎯ Potential budgetary trade-offs
from public funding
⎯ Possible cost pass-through from
compliance obligations to end
users
⎯ Higher prices12
⎯ Improved transparency and
stakeholder engagement
⎯ Faster and more sustainable roll-
out of digital infrastructure
⎯ Improved digital service quality
and lower costs
⎯ Improved transparency on security
properties
⎯ Enhanced protection of
fundamental rights, especially
privacy and data protection
⎯ Enhanced transparency and trust
⎯ Higher employability
⎯ Greater trust in public digital
services
12 Concerning data centre expansion, potential increased costs via grid tariffs are an important element. ACER’s 2024
Electricity Infrastructure Monitoring Report shows that future grid costs are highly sensitive to rising electricity consumption
and increased grid-investment expenses, with a 10% increase in either factor possibly resulting in a 1 to 4.5 EUR/MWh rise in
grid costs by 2050. As data centres are among the fastest-growing and electricity-intensive sectors, they influence both drivers
of higher grid costs with consequent possible increases in electricity prices. The Finnish National Roadmap for Data Centres
similarly highlights that the sector’s expansion could strain the power grid and contribute to higher electricity prices unless
data centre development and grid planning are coordinated proactively. See also: Overcoming energy constraints is key to
delivering on Europe's data centre goals – Analysis - IEA; ACER Monitoring Report on cross-zonal electricity trade and
congestion management in the EU (2025) and National Roadmap for Data Centres : Rapporteur's Report.
Moreover, in regions where water resources are limited, inefficiently managed industrial demand can exacerbate competition
with households, increase the likelihood of water restrictions, and call for expensive network upgrades – costs that may
ultimately affect consumer water bills depending on the tariff structure and regulatory cost distribution. Thus, proactive
governance of data centre water usage is also key to protect citizens’ access to water and prevent unintended increases in
residential water costs.
32
Affected stakeholder Main impacts
Costs Benefits (Direct and indirect)
⎯ Continuity and resilience of
critical services, like health,
defence, and finance
This initiative is expected to generate costs for data centre operators and cloud and AI
computing service providers, but also significant savings and direct economic benefits. While
administrative and adjustment costs arise from, inter alia, participation in funding programmes
and fast-track permitting procedures, or from the audit procedures and risk assessments, these
are offset by the efficiencies and benefits that emerge from administrative simplification,
reduced fragmentation of national approaches, and greater incentives for investments,
improving the attractiveness of the Single Market.
National public authorities are expected to incur administrative and adjustment costs for
assessing applications for support measures and permitting, planning resources for fast-track
areas, monitoring compliance, and participating in the cloud federation. However, these costs
are balanced by significant savings resulting from reduced inefficiencies and duplications
caused by the lack of harmonisation. National public authorities will also directly benefit from
open source solutions as well as indirectly from additional cloud and AI capacity at lower costs.
Ultimately, this initiative aims to support the digital transformation and resilience of public
authorities. Moreover, while some Member States might need a reinforcement of capabilities,
the EU would bear the cost of oversight, with the aim to create efficiency gains in the
cooperation across Member States. The administrative and adjustment costs borne by the
European Commission, related to the implementation of the measures and their supervision are
expected to be outweighed by significant long-term benefits. Society at large is expected to
benefit from the wider availability of digital services, underpinned by robust digital
infrastructures. As the same time, the deployment of sustainable infrastructures will accelerate
the green transition, ensuring that digitalisation goes hand in hand with environmental
responsibility.
The assessment of the impacts draws on multiple data sources, including the targeted
stakeholder consultation (interviews and surveys), the public consultation, and desk research
in the context of the impact assessment support study. To the extent possible, the impacts are
quantified based on available assessments or modelling by Technopolis et al. (2025)13. Where
dedicated modelling was not possible due to the lack of tools or data, a qualitative assessment
was performed, based on existing studies and input from stakeholders. Annex 4 provides further
details on the methodological approach, detailed tables and estimates.
2. SUMMARY OF COSTS AND BENEFITS
13 Technopolis et al. (2025), “Study: Cloud and AI”
I. Overview of Expected Benefits (total for all provisions) – Preferred Option
Description Indicative range
(NPV 10-years) Comments
Direct benefits
Economic benefits from
administrative simplification and
fast-track areas for providers
(PM4, PM5) EUR 8 – 27 bn
These are the large, expected benefits accruing to data centre
operators which are foreseen to benefit from reduced times to build
new facilities. It has been calculated as the net present value of
discounted economic benefits over 10 years. The economic benefits
that are expected from these measures extend beyond the
measurable impact (see below on indirect benefits).
33
14 Harvard economist Jason Furman reported in September 2025 that investment in information processing equipment &
software was 4% of GDP in the US in 2024, but it was responsible for 92% of GDP growth in the first half 2025; US GDP
excluding these categories grew at a 0.1% annual rate in the first half of 2025. More information available here: Without data
centers, GDP growth was 0.1% in the first half of 2025, Harvard economist says | Fortune.
Cost savings from administrative
simplification and fast-track areas
for public authorities (PM4, PM5)
EUR 166 – 277 m
These include administrative cost savings accruing to public
administrations as they reduce parallel processing of projects and
back-and-forth interactions with operators.
Cost savings from the increased
use of open source in public
administrations (PM20) EUR 2 – 12 bn
By reusing and adapting existing software rather than
commissioning bespoke solutions, administrations can reduce
duplication of effort and lower the total cost of ownership of ICT
projects.
Costs savings for authorities from
using standard non-specific award
criteria when drafting tender
specifications (PM19)
EUR 4 – 13 m
Cost savings for public authorities stem from use of standard
criteria when procuring cloud and AI computing services that can
be reused across all public tenders.
Costs savings for authorities from
using audited services in tenders
(PM21) ⁓ EUR 2.5 m
Cost savings for public authorities stem from the use of the
repository of sovereign cloud and AI computing services and audit
reports, which are expected to save administrations time in their
verification of the documentation for the evaluation of the offers.
Cost savings for cloud and AI
service providers from being able
to use sovereignty audits across all
MS, i.e. once and for all (PM21)
EUR 2 – 5 bn
These direct savings for providers stem from their ability to get the
audited in one Member State and participate in procurement of
services for critical use cases across all EU27.
Cost savings from cloud
federation and joint procurement
for public administrations (PM22) EUR 19 – 49 bn
Recurrent savings stem from the cloud federation, i.e. from
reducing idle capacity, getting cheaper computing resources and
reducing the coordination effort for sharing computing capacity.
The participation in the joint procurement framework and the
aggregation of demand in common EU-level procurement
processes is also expected to significantly reduce the required FTEs
for procuring cloud and AI computing services and service prices
from higher scale and purchase power.
Direct cost savings in AI-based
transformation project design for
SMEs (PM23)
EUR 160 – 500 m
The targeted scheme to provide financial support to SMEs for
adopting cloud and AI computing services foresees a yearly budget
of EUR 40 050 000 over the 10-year period, mostly dedicated to
supporting the design and planning phase of cloud and AI-based
transformation projects for SMEs. The grants are fixed amounts that
SMEs can spend in consultancy services to design digital
transformation projects based on cloud and AI technologies.
Indirect benefits
Indirect benefits coming from data
centre deployment targets
This was assessed
qualitatively or
based on existing
literature.
Decrease of the capacity gap and contribution to the overall GDP
and workforce driven by the data centre sector and related fields. For operators, clearer EU-wide targets and consistent monitoring
could help reduce investment uncertainty, reducing effort in
oversupplied or poorly sited locations. National authorities would
also benefit from improved infrastructure foresight, with better
planning for grid reinforcements and targeting of support schemes.
Indirect benefits from additional
data centre capacity on GDP
growth and employment levels
This was addressed
under Section 6.1.5
and 6.2.
Economists have attributed the growth in GDP growth from
investment in data centres and information processing technology14.
In additional to these gains, there are likely to be indirect benefits
for society at large, e.g. in terms of improved digital infrastructure,
skills development and increased attractiveness of the EU for
complementary industries. These spillover effects, while possibly
significant, are not directly quantified due to the lack of reliable
data.
34
(1) Estimates are gross values relative to the baseline for the preferred option as a whole (i.e. the impact of
individual actions/obligations of the preferred option are aggregated together); (2) In the comments column the
key stakeholder group as main recipient of the benefit is presented;(3) Spillover effects and broader social benefits
of enhanced digital infrastructure for citizens and consumers were hard to quantify given their long-term and
diffuse nature and were thus assessed only qualitatively.
II. Overview of expected costs, indicative ranges as discounted NPV over 10 years – Preferred option
Citizens/Consumers Businesses
(including SMEs)
Administrations (national public
and European Commission)
One-off Recurrent One-off Recurrent One-off Recurrent
Legislative and
financial intervention
enforced nationally:
Administrative
simplification and
fast-track areas for the
roll-out of data
centres; support
measures and
deployment targets
also based on national
data centre strategies
(PO1-B)
Direct
adjustment
costs
EUR 6 – 26 m EUR 28 – 43 m EUR 2 – 14 m
Direct
administrative
costs
EUR 1 – 4 m EUR 0.2 – 0.6 m EUR 1 – 6 m EUR 46 – 86 m
Direct
enforcement
costs
EUR 30 – 89 m
EU R&D funding and
deployment funding
for strategic projects
(PM8, PM9)
Direct
adjustment
costs
EUR 1 – 2 m EUR 0.6 – 1.2 m
Direct
administrative
costs
EUR 1 – 3 m
Indirect benefits coming from
funding for R&D and strategic
projects
This was assessed
qualitatively or
based on existing
literature.
It is important to underline that the expansion of data centre
capacity entails considerable environmental impacts, including
increased energy consumption, water use, land uptake and
associated carbon emissions. These impacts may generate new
social and environmental costs, with possible consequences for
overall welfare, local ecosystem and community acceptance. For
these reasons, measures that promote sustainable solutions are
fundamental to ensure that new facilities are designed and operated
with sustainability principles at their heart. This fundamental to
maximise net social benefits and safeguard the public good over the
long term.
Indirect benefits from the
increased use of open source
solutions in public administrations
This was assessed
qualitatively or
based on existing
literature.
When public administrations increase their procurement and use of
open source software, they create additional demand for open and
interoperable solutions across the wider economy.
Indirect benefits from the
sovereignty risk assessment
scheme
This was assessed
qualitatively or
based on existing
literature.
The sovereignty assessment framework provides national
authorities with a more robust mechanism to procure and uptake
cloud and AI computing services, protecting critical use cases
where public order is at stake. It allows for the diversification of
dependencies, and services from non-European providers would
still be able to qualify for 90% of the public sector’s needs. At the
same time, it contributes to reducing exposure to non-EU
dependencies and single provider concentration risks, strengthening
the resilience and autonomy of cloud and AI services.
Indirect benefits from joint EU-
level procurement
This was assessed
qualitatively or
based on existing
literature.
The increase in public procurement of cloud and AI sales volume
that is addressable by European providers may help increase their
scale and competitiveness and their share in the European and
global cloud and AI markets.
35
II. Overview of expected costs, indicative ranges as discounted NPV over 10 years – Preferred option
Citizens/Consumers Businesses
(including SMEs)
Administrations (national public
and European Commission)
One-off Recurrent One-off Recurrent One-off Recurrent
EU-coordinated
procurement and
support framework for
sovereign cloud and
AI computing
services: OSS use in
the public sector;
mandatory use of
audited cloud and AI
computing services in
highly critical public
sector use cases and
award criteria; private
entities cybersecurity
risk-assessment
integration to consider
risks from using non
audited providers;
national strategy
definition and
monitoring; joint EU-
level procurement;
public sector cloud
federation;
SME support scheme;
Cloud and AI toolbox
(PO2-C)
Direct
adjustment
costs
EUR 173 m –
352 m
EUR 140 m –
420 m (*) EUR 1.3 – 2.7 bn
Direct
administrative
costs
EUR 27 m – 83
m
EUR 560 m – 2.9
bn
EUR 10 m – 27
m
EUR 142 m – 293
m
Direct
enforcement
costs
Direct
regulatory fees
and charges
⁓ EUR 3 m ⁓ EUR 3.4 m
(*) The mandatory use of sovereign cloud and AI computing services would trigger an acceleration of porting of some
applications to sovereign cloud services of assurance levels 2 to 4, requiring an anticipation of expenses of EUR 3 to 15 bn, as
reported in section 12.4 of Annex 4. Since they represent planned expenditure that would be incurred in the future as part of
the regular cloud contract renewal cycles and independent of the preferred option, they have been scoped outside of the
summary cost table so as not to conflate structural renewal costs with the incremental financial impact of the measure itself.
(1) Estimates (gross values) to be provided with respect to the baseline; (2) costs are provided for each
identifiable action/obligation of the preferred option otherwise for all retained options when no preferred
option is specified; (3) If relevant and available, information on costs is presented according to the
standard typology of costs (adjustment costs, administrative costs, regulatory charges, enforcement
costs, indirect costs;); (4) although relevant environmental and social costs have been considered in this
assessment, they were not quantified here due to the lack of sufficient and reliable data.
III. Application of the ‘one in, one out’ approach – Preferred option
One-off
(annualised total net present
value over the relevant period)
Recurrent
(annualised total net present value
over the relevant period)
Total
Businesses
New administrative
burdens (INs) (EUR 30 m – 90 m) (EUR 559 m – 2.8 bn) (EUR 589 m – 2.8 bn)
Removed administrative
burdens (OUTs) EUR 404.3 – 2 bn EUR 404.3 – 2 bn
36
III. Application of the ‘one in, one out’ approach – Preferred option
One-off
(annualised total net present
value over the relevant period)
Recurrent
(annualised total net present value
over the relevant period)
Total
Net administrative
burdens* (EUR 30 m – 90 m) (EUR 154.7 m – 0.8 bn)
(EUR 184.7 m – 809
m
Adjustment costs**(EUR 179 m – 378 m)
National public administrations
New administrative
burdens (INs)
(EUR 12 m – 34 m) (EUR 154 m – 414 m) (EUR 166m – 448 m)
Removed administrative
burdens (OUTs)
EUR 90 m – 154 m EUR 90 m – 154 m
Net administrative
burdens* (EUR 12 m – 34 m) EUR 64 –2-60 m EUR 68 m – 194m
Adjustment costs**EUR 209 m –507 m EUR 1 bn – 3 bn
Total administrative
burdens*** EUR 42 m – 124 m EUR 218.7 m – EUR 860 m
EUR 252.7 m –1.00
bn
(*) Net administrative burdens = INs – OUTs;
(**) Adjustment costs falling under the scope of the OIOO approach are the same as reported in Table 2 above.
(***) Total administrative burdens = Net administrative burdens for businesses + net administrative burdens for public
administrations
3. RELEVANT SUSTAINABLE DEVELOPMENT GOALS
IV. Overview of relevant Sustainable Development Goals – Preferred Option
Relevant SDG Expected progress towards the Goal Comments
e.g. SDG no. 4 – quality
education
Increase in training opportunities for enhancing
digital skills, contributing to quality education and
lifelong learning.
e.g. SDG no. 7 - affordable
and clean energy, 12 -
responsible consumption and
production, 13 - climate
Increased energy efficiency of data centres will
save energy and CO₂e emissions per GW of
additional capacity added with respect to the
baseline over the next 10 years, contributing
positively to SDG no. 7 (affordable and clean
energy) and SDG no. 13 (climate action) including
through the promotion of renewable energy
sources to power data centres. Greener data
centres would contribute to reduce their carbon
footprint and support the clean energy transition.
Promotion of energy-efficient hardware and
optimisation of cooling systems, also thanks to the
use of AI-integrated solutions, is also expected to
reduce waste and resource consumption. The
proposal embeds the concept of “sustainable-by-
design” data centres, including high
environmental and energy-efficiency obligations,
e.g., through alignment with the upcoming Union
scheme for rating the sustainability of data centres.
Affordable energy for all must be
sustained through the renewal and
upgrade of the energy grids (which falls
outside of the scope of the current
initiative), but which is a relevant
prerequisite to move closer to this SDG
target.
37
IV. Overview of relevant Sustainable Development Goals – Preferred Option
Relevant SDG Expected progress towards the Goal Comments
SDG no. 8 – Decent work
and economic growth
Expected increase in employment in relation to
data centre rollout and GDP growth. Expected
increase of SMEs and startups’ use of cloud to
support inclusive economic growth.
SDG no. 9 – Industry,
innovation and infrastructure
Increase in the number of data centres and next
generation infrastructure to meet the needs of
business and citizens in the EU, strengthen
innovation and build resilient digital
infrastructure.
Cloud computing and AI also enable
digital transformation across industries,
boosting productivity and new services.
The initiative aims to increase the EU’s
share in the global cloud market from
15% to 30%.
38
ANNEX 4: ANALYTICAL METHODS
1. METHODOLOGY
This impact analysis has followed a stepwise approach.
First, a modelling analysis of each proposed policy measure has been carried out, using a
structured set of parameters, assumptions and data inputs15. The data included has been
gathered from the following sources:
• Eurostat, notably for the labour costs.
• The supporting study “Study: cloud and AI” led by Technopolis Group16, including
over 60 targeted interviews, surveys with over 250 replies and two workshops with over
100 participants each.
• Tenders Europe Daily (TED), for the calculation of the procurement contract values by
public administrations in the field of cloud and AI computing services.
• Desk research, with sources cited in the corresponding measures.
• Best practices documented in the industry or similar approaches followed in other world
regions or in EU countries.
Section 2 of this Annex explains the general assumptions used for the modelling of the different
policy measures. These assumptions apply across several policy measures and are therefore
reused throughout the analysis.
Section 3 includes a detailed explanation of the modelling approach for each policy measure.
For each measure, the analysis covers:
• The scope of the policy measure and how it would be implemented in practice.
• The different types of costs, including the one-off and recurrent administrative and
adjustment costs per stakeholder
• The assumptions, values and sources used to estimate these costs.
• The expected cost savings and benefits, including the assumptions and the reasoning
behind them.
• Where relevant, indirect savings, that cannot be fully monetised, but that nevertheless
play an important role for assessing the overall effect of the policy measure.
A sensitivity analysis was carried out to identify the parameters with the greatest variability
and the strongest potential impact on the cost-benefit results. This analysis helps understand
the robustness of results under the different assumptions. A final summary table presents the
overall results for each policy measures, under three scenarios (central-min-max), as well as
costs and benefits for the key stakeholder groups impacted.
The modelling has resulted in a cost-benefit analysis for each policy measure, which was
aggregated and presented in sections 6 and 7 of the core text of the impact assessment and in
the Annex 3. In most cases, the analysis followed the Standard Cost model, whereas in some
instances the NPV or other quantification approaches have been used to estimate expected
benefits.
15 This consisted of a cost-benefit model built around defined assumptions and evidence sources, not on a macroeconomic
model. 16 Technopolis et al. (2025). “Study: Cloud and AI”
39
Section 4 discusses the strengths and limitations of the approach and of the data used for the
analysis. To investigate the robustness and stability of the results, a min-max sensitivity
analysis was performed (See section 5 of this Annex for more details and assumptions checked
in interviews and CATI surveys).
Section 6 presents the multi-criteria decision analysis (MCDA), which was carried out using
three different models, to further assess the strengths and weaknesses of each option.
Wider economic impacts, indirect effects or non-market social and environmental externalities
can be material but are excluded from the monetisation within the Cost-Benefit Analysis due
to limitations in available data and the lack of robust methodologies for monetising these
effects.
Section 7 of this Annex includes the assumptions behind the environmental impact analysis
performed, while section 8 includes additional literature used to discuss the wider economic
effects associated with the policy options.
Section 9 shows the analysis performed by Technopolis Group as part of the study “Study:
cloud and AI” on the different procedures, permits and data centre deployment timelines across
several Member States, aiming to demonstrate the fragmented approach currently existing in
the EU.
2. GENERAL ASSUMPTIONS FOR THE MODELLING OF THE POLICY MEASURES
2.1. Valuation framework and time horizon
Price base.All figures are expressed in real 2025 euros; no general price inflation is applied.
All unit costs are expressed in real 2025 euros and are based on Eurostat hourly labour cost. In
2024, the EU-27 average hourly labour cost for the whole economy was EUR 33.5; services
averaged EUR 33.3; information and communication at EUR 46.3; and the mainly non-business
economy (excl. public administration) averaged EUR 34.2. These figures are taken from
Eurostat’s latest annual release and Statistics Explained pages17.
Discounting. Present values are computed at a 3% real discount rate. The discounting base date
is the beginning of 2027; cash-flow timing follows an end-of-year convention. The cumulation
period is 2027-2036 inclusive. The discount factor for a flow in year t is:
DF = 1
(1 + 0.03)−2027
Horizon. Impacts are accumulated over 2027–2036 inclusive (10 years).
2.2. Unit cost parameters/ standard cost model
Working day length. 8 hours/day.
Working days in the year. 220 working days/year
Day-rate proxies (EU averages with sector uplifts):
Eurostat’s EU-27 hourly labour cost converted to day-rates:
17 EU hourly labour costs ranged from €11 to €55 in 2024 - News articles - Eurostat
40
Table 7. Labour costs and rates used (Source: Eurostat and Statistics Explained Hourly labour costs page (Table for
breakdown by activity)
Baseline aggregate Hourly labour cost
(2024, EU-27)
Day-rate (8h)
Services EUR 33.3 / h EUR 266.40 / day
Mainly non-business (excl. public admin) EUR 34.2 / h EUR 273.60 / day
Information and communication EUR 46.3/h EUR 370.40 / day
Eurostat’s 2024 reports, for the EU-27, an average hourly labour cost in Information and
communication (NACE J) at EUR 46.3/hour18. This level reflects the high ICT skill mix and
sits above the overall services economy. Anchoring directly to NACE J provides a relevant
comparator for data-intensive activities. To capture systematic differences in skill composition
and overhead, modest multipliers are applied to the J baseline, justified by reports on ICT
labour market tightness and hard-to-fill vacancies19. Each uplift is a modelling parameter,
justified by EU evidence on sector pay premia and administrative overheads:
• Cloud service providers (CSPs): ICT activities (NACE J62 – Computer programming,
consultancy, and related activities; J63 – Information service activities) carry higher
earnings and labour costs than services overall due to specialist roles (software
engineers, site reliability engineers, cybersecurity). Eurostat SES shows ICT employees
are concentrated in higher-skill/tertiary categories; Eurostat reports tight markets for
ICT specialists and persistent hard-to-fill vacancies, supporting a positive premium vs.
average services. A 30% uplift is set on this baseline, which yields to EUR 481.52/day
(46.3×8×1.3)
• AI computing service providers: AI computing service providers offer computing
infrastructure for the running (inference) of AI systems. Due to the complexity and
novelty of running these clusters of AI GPUs, more specialised roles are needed
compared to those running cloud infrastructures. Data show employees in these groups
cluster in the highest wage deciles within ICT and Eurostat reports structural labour
shortages in ICT that are particularly acute for advanced digital skills. This systematic
premium reflects both scarcity of talent and high value-add of AI services compared
with broader ICT20. A 45% uplift is set on the baseline, which yields to EUR 537.08/day
(46.3×8×1.45)
• Data centre operators: DC operations mix technical operators, electricians, facility
engineers and IT technicians. The mix is more operations-heavy and less developer-
heavy than CSP software teams, so the premium over services is lower than for CSPs
but still positive relative to the broad aggregate due to 24/7 operations and specialised
compliance (e.g., uptime, safety). A 8% uplift positions DC operations below CSPs but
above average ICT services, consistent with observed role mixes in NACE J63.11. This
+8% leads to EUR 400.03/day (46.3×8×1.08).
• National public authorities: Eurostat does not publish NACE O separately in this
dataset, but reports EUR 34.2/h for the “non-business economy excl. O” aggregate, with
18 [lc_lci_lev] Labour cost levels by NACE Rev. 2 activity 19 ICT specialists in employment - Statistics Explained - Eurostat 20 AI Jobs Barometer | PwC
41
a figure for the mainly non-business economy (excl. public administration) as EUR
34.2/h. We apply a small +5% uplift to proxy: (i) policy/project staff engaged in inter-
institutional coordination (above average non-business roles), and (ii) inclusion of
specialist grades not perfectly captured by the “excl. public administration” aggregate.
This remains conservative relative to some MS where public-sector employer costs are
higher than the EU average. This +5% on non-business (excl. public admin) leads to
EUR 287.28/day (34.2×8×1.05).
• European Commission: Coordination at EU level entails multilingual drafting, legal
vetting, translation, dissemination, and inter-DG and MS engagement. Relative to
Member-State averages for non-business activities, Commission project teams typically
face higher overhead (procurement, quality assurance, inter-service consultation). The
+15% factor is a policy-evaluation convention to reflect those overheads while staying
close to EU-wide non-business costs. Using the baselines and multipliers above +15%
on non-business (excl. public admin) leads to EUR 314.64/day (34.2×8×1.15).
• Private sector essential entities operating in sectors listed in Annex I of NIS2:
Annex I NIS2 lists the following sectors as highly critical sectors: Banking and
Financial market infrastructures (NACE Rev2 - K), Health (NACE Rev2 - Q), Drinking
Water and Waste water (NACE Rev2 - E). Energy (NACE Rev2 - D), Transport (NACE
Rev2 – H), Digital infrastructure and ICT Service Management (NACE REv2 - J),
Public Administration and defence (NACE Rev2 – O) and Space (NACE Rev2 – C).
The values that Eurostat yields for this query for 2024, the most current data is 46.8
€/hour21
Using the baselines and multipliers above
• CSPs: 46.3 EUR /h×8×1.30 = EUR 481.52 / day.
• Data-centre operators: 46.3 EUR /h×8×1.08 = EUR 400.03 / day.
• AI service providers: 46.3 EUR /h×8×1.45 = EUR 537.08 / day
• National authorities: 34.2 EUR /h×8×1.05 = EUR 287.28 / day.
• European Commission: 34.2 EUR /h×8×1.15 = EUR 314.64 / day.
External contractor rates: are set at 50% over external AI developer, EUR 537.08/day x 1.5=
EUR 805,62/day.
Travel. EUR 700 per in-person trip per participant (transport and incidentals for one day and
one night).
Webinars. 2 hours per participant, converted to 0.25 day for labour costing.
In-person meeting/event time. Each in-person engagement entails 2.5 participant-days (1.0
meeting days + 1.5-day for preparation and follow-up).
2.3. Estimates related to data centres
2.3.1 Data centre supply growth
There is no single, EU-wide dataset produced by regulators or energy system planning exercises
that consistently locates historical data centre capacity at site level in terms of commissioned
21 Query used available at https://ec.europa.eu/eurostat/databrowser/view/lc_lci_lev__custom_20086615/default/table
42
IT load. The estimation of new data centres deployed under the different policy options builds
on the scenarios of data centre growth (measured in MW) identified in Technopolis et al.
(2025), “Study: Cloud and AI”. The analysis relied primarily on the Data Centre Map dataset,
which is built entirely on historical data centre sites that have already been developed and
commissioned, i.e. the 2025 dataset reflects the cumulative outcome of past capacity
development rather than a forward-looking projection. The dataset includes the year of launch
of the sites for only around one-third of the recorded sites, with uneven coverage across
countries and operators. As a result, it was impossible to construct a consistent historical
baseline of commissioned capacity across the EU without strong assumption or a systematic
bias.
Nevertheless, to present forward-looking projections based on past22 and future trends, the
study estimated possible forecasts for data centre capacity across the EU27 based on the
characteristics of three strategic scenarios (low growth, central growth and high growth). These
projections illustrate the potential evolution of installed capacity over the next decade,
reflecting variations in natural resources, demand drivers, policy intervention, technological
innovation, and market responsiveness.
Table 8. Overview of low growth, central and high growth scenarios (Source: Technopolis et al. (2025))
Scenario CAGR
(2025-2036) Key assumptions and trends
Low growth 9% • Across 2025–2036, EMEA data centre capacity grows at an
average ~9% CAGR, below the ~13.8% historic pace seen in
2018–2024. Growth remains concentrated in primary markets
e.g. Germany, France, Ireland, and the Netherland while many
secondary and emerging markets struggle to secure large,
anchor-tenant investments or hyperscale commitments. The
build-out is most intense in 2025–2030 (approximately 14%
CAGR), driven by committed pipelines, cloud region rollouts,
and AI-ready retrofits. From 2030–2036, momentum slows to
roughly 4.2% CAGR as power, permitting, and capital
discipline increasingly constrain new capacity additions.
• Power availability is the binding constraint, with insufficient
electrical capacity to support large-scale AI compute in
countries such as Netherlands, Ireland, Germany with limited
investment in grid modernisation, increasingly observed
between 2030-36. These constraints limit the development of
new high-density sites and delay infrastructure upgrades.
• In the absence of major policy shifts, a fragmented regulatory
landscape persists across Europe. Disjointed national
strategies, coupled with tightened environmental restrictions
on energy usage and prolonged permitting processes, create
significant barriers to new data centre development. These
22 To contextualise historical growth dynamics we performed desk research and looked at past industry-level analyses, which
document that the FLAP-D markets experienced a 17% CAGR in MW supply between 2018 and 2023, while secondary
European markets saw a 23% CAGR in the same period. Other industry market reports tracked data centre power capacity and
trajectories, showing an increase from 9 GW in the early 2020s to above 11 GW in 2024. These trends confirm the growth
dynamics of the European market identified in the study, but they often differ substantially in geographic scope (e.g.
considering the EMEA or UK market). No existing study combines the uses a methodology in a comparable manner to that of
our study, making any comparisons misleading. See: Roland Berger and spotlight-eu-data-centre---november-2022-.pdf;
spotlight-eu-data-centre---may-2024---final.pdf
43
Scenario CAGR
(2025-2036) Key assumptions and trends
hurdles constrain growth, keeping capacity expansion below
historically observed levels.
Central
growth
(used as
baseline)
12% • Capacity growth aligns with the CATI survey average
(~13% CAGR) and broader industry momentum through
2025–2030, then eases over 2030–2036. Expansion remains
concentrated in FLAP-D hubs, with steady investment in
secondary markets (Nordics, Spain, Italy) and slower uptake
in developing markets (around 6.49% CAGR). In the
baseline, aggregate growth is ~17.7% over 2025–2030
(consistent with the survey’s ~13% CAGR lens) as
committed pipelines, AI-readiness upgrades, and new cloud
regions come through. From 2031–2036, growth moderates
to ~7.1%, reflecting tighter site selection driven by power
and land availability, longer permitting lead times, and
higher financing costs.
• Electrical grid limitations remain an issue in certain primary
markets such as Ireland and the Netherlands. However,
regulators and operators in several markets have begun
investing in grid modernisation and demand management.
• Policy evolution remains gradual and inconsistent, creating
a fragmented regulatory landscape across Europe. Efforts to
modernise infrastructure and streamline permitting are
underway, but progress is uneven, often slowed by
regulatory complexity and localised decision-making.
Governments aim to balance digital infrastructure
expansion with environmental priorities by introducing
power efficiency targets, and sustainability requirements.
High
growth
15% • The growth rate accelerates to 15% from 2025-2036 with
front loaded 21.7% CAGR in 2025-2030 before moderating
to 10.4% from 2030-36. This is improving on the baseline
scenario by following the trajectory of advanced, leading-
edge markets such as Northern Virginia in the US, where
data centre capacity grew at a 25% CAGR between 2014 and
2021, with capacity doubling from 2018 to 202123. This
scenario sees aggressive geographic expansion beyond
primary and secondary markets, with new regional hubs
emerging in Southern Europe, the Baltics, and Central &
Eastern Europe. Growth in developing regions significantly
exceeds historical trends with a CAGR of 13% and outpaces
current market projections.
• Major public and private investment in grid modernisation
leads to substantial increases in available electrical capacity,
effectively removing power availability as a constraint on
data centre expansion. This enables large scale AI
deployments, including high density GPU workloads to
23 NVTC (2022). The impact of data centres on the state and local economics of Virginia 2022. Available at:
https://www.spotsylvania.va.us/DocumentCenter/View/29411/2022-Virginia-Data-Center-Report--32022?
44
proceed rapidly not only in primary markets but in secondary
and developing markets.
• Proactive, coordinated policy action is witnessed across
Europe. National and EU-level governments prioritise digital
infrastructure development, implementing supportive
strategies such as AI innovation hubs, tax incentives, and
expedited permitting processes. Environmental goals are
addressed in parallel, but through mechanisms that
encourage rather than restrict growth, such as mandates for
renewable energy procurement and circular economy
initiatives.
As of 2025, all three scenarios converge at an estimated 12.4 GW of installed data centre
capacity. This uniform starting point persists into 2026, with all paths reaching 15.1 GW, based
on relatively firm visibility of near-term buildout plans and permitted sites. However, the
trajectories begin to diverge thereafter as each scenario follows the key assumptions outlined
above from 2027 onwards. This projection broadly aligns with the literature, even though some
sources present slightly less ambitious estimates, such as a forecasted capacity of 13.1 GW by
202724.
By 2030, under the central scenario, capacity reaches 28 GW. In the high growth scenario,
driven by heightened demand and accelerated infrastructure deployment, capacity increases
more sharply to 33 GW. Meanwhile, the low growth scenario sees more modest expansion,
with capacity reaching just 24 GW by the end of the decade. The CAGR from 2025-36 for the
low growth scenario is 9%, for the central growth scenario is 12% and for the fast growth
scenario is 15%.
Extending the forecast horizon to 2036 introduces a higher degree of uncertainty, largely due
to evolving market dynamics, potential technological breakthroughs, and a lack of visibility
into planned or in-progress developments beyond the 2026–2030 window. That said, the
divergence between the scenarios becomes more pronounced. In the low growth scenario,
installed capacity reaches 31 GW, while the central trajectory leads to 42 GW. The high growth
scenario sees capacity expand dramatically to over 60 GW, representing a 55% increase over
the central forecast, as shown in the Figure below.
24 Savills (2024). European data centre power capacity projected to rise to approximately 13,100 MW by 2027. Available at:
https://www.savills.co.uk/insight-and-opinion/savills-news/362723-0/european-data-centre-power-capacity-projected-to-rise-
to-approximately-13-100-mw-by-2027
45
Figure 1. Data centre capacity across the EU27 in GW from 2025-36 across three growth scenarios
Source: Technopolis et al. (2025) by means of a CATI survey (n=100 companies, 280 Data Centres), DataCentreMap
These findings complement existing literature, which forecasts a rapid growth in European data
centre capacities in the coming years driven by the breakthrough of AI25. While a
diversification of data centre markets is similarly expected, the literature places more emphasis
on constraints affecting the growth of emerging markets, including limited power grid capacity
and less developed economic or business ecosystems26.Technologies typically follow an S-
curve growth rate and so is the expectation about data centre growth, implying that at some
point in the future, the CAGR will start declining. The study has modelled a partial inflection
of growth rates after 2030 but finds that other inflexion points are not expected in the time
horizon contemplated here.
To estimate near-term growth, the Data Centre Map dataset was leveraged, since it includes
site-level information of facilities "Under Construction" and "Planned." All data centres
identified as "Under Construction" were assumed to be operational by 2026, reflecting standard
build timelines and aligning with industry expectations. Facilities classified as "Planned" were
incorporated into our longer-term forecasts, with their capacity distributed evenly across the
2026–2030 period. Information from government sources, company announcements, and
market reports has been consulted to cross-check against these estimates. These sites were then
assigned to NUTS-1 regions through geospatial mapping and manual checks where coordinates
are missing. Projections beyond 2027 are also adjusted for regional and market maturity
categories (primary, secondary, developing) across the three scenarios.
To assess the realism of projected capacity, a natural resource constraint at the NUTS 1 level
was incorporated. This was based on the national electricity generation capacity from 2020–
2024 available from the European Commission’s Europe and Industry dataset and downscaled
to the NUTS 1 level. This data was used to project grid capacity to 2036, applying historical
25 JLL (2024). EMEA Data Centre Report Q4 2024. Available at: https://www.jll.com/en-uk/insights/emea-data-centre-report.
McKinsey analysis estimating global demand for data centre capacity to grow by a compound average growth rate of 19% in
the pessimistic scenario and 27% in the more optimistic one, with efficiency improvements, flexible workload management
and supportive regulation. Available at: : https://www.mckinsey.com/industries/technology-media-and-
telecommunications/our-insights/ai-power-expanding-data-center-capacity-to-meet-growing-demand 26 ICIS. Data centres: Hungry for power. Available at: https://www.icis.com/explore/resources/data-centres-hungry-for-power
46
CAGR trends. A validation threshold was then introduced to ensure that forecasted data centre
capacity does not exceed a set share of regional power generation. This percentage was 20%
for the central scenario, to reflect a situation where data centres become a relevant electricity
user in line with emerging patterns in highly digitised regions; a more conservative 10% for the
low scenario to reflect tighter grid constraints; while no constraints were considered for the
high scenario under the assumption that major public and private investment in grid
modernisation would enable substantial increases in available electrical capacity.
2.3.1.1 Scale, timing and distribution of new capacity
This section provides additional information on the scale, timing, geographic distribution and
functional orientation (AI-readiness) of new data centre capacity within the EU-27 under the
central growth (baseline) scenario.
Scale: the split between hyperscalers and colocation facilities is illustrated here below.
AI-readiness: based on the results of the CATI survey, the study estimated that in 2025
approximately 12% of total EU-27 data centre capacity is AI-ready, defined as infrastructure
capable of supporting high-density AI workloads (e.g. GPU-accelerated compute, advanced
cooling, and power delivery). Survey responses indicate that hyperscale operators report a
substantially higher share of AI-ready capacity than colocation providers, a finding that mirrors
the scale of hyperscaler investment in advanced GPU architectures. Building on this baseline,
three forward-looking scenarios for AI-ready capacity as a share of total data centre
infrastructure were developed.
o Slow growth scenario: By 2036, approximately 35% of total data centre capacity is
AI-ready, corresponding to an average growth rate of ~32% over the ten-year
period.
o Baseline growth scenario: By 2036, approximately 64% of data centre infrastructure
is AI-ready.
o Fast growth scenario: By 2036, approximately 85% of data centre infrastructure is
AI-ready, directly aligned with the upper-bound expectations reported by CATI
survey participants
These results were also validated through the several interviews with relevant stakeholders and
two workshops organized in the context of the study.
In practice, in relation to computing capacity, the market and technical reality does not lend
itself to a distinction between different types of supply with the exception of the training of
large AI models. As discussed in Annex 8, the training of AI models requires large,
concentrated amounts of computing power. For training, low latency is not as important as for
the later stages in an AI model’s lifecycle. This sets AI training facilities somehow apart. To a
lesser extent, large facilities dedicated to a single client (typically hyperscalers) do not need to
be in proximity to an economic centre, to the difference of colocation data centres which tend
to be located next to their clients, so closer to economically active areas. The policy intervention
is mostly agnostic on these considerations as the most pressing needs of data centres do not
differ significantly based on their end use. One of its objectives is to create the right conditions
for a faster build-out of data centres and is agnostic to their ultimate use, allowing the market
Type 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 Colocation / CSP / telco (primary markets) 4210 5120 6068 7090 8238 9523 10577 11490 12317 13055 13703 14369 Colocation / CSP / telco (secondary markets) 2401 2921 3462 4045 4699 5433 6033 6555 7026 7447 7817 8197 Colocation / CSP / telco (developing markets) 793 965 1143 1336 1552 1794 1993 2165 2321 2460 2582 2707 Hyperscaler 5036 6125 7258 8481 9854 11391 12651 13744 14734 15615 16391 17187
12440 15131 17932 20953 24343 28141 31254 33954 36398 38577 40492 42460
47
to find the right balance. That is why the capacity gap presented in Annex 9 (baseline scenario)
looks at total data centre capacity and does not differentiate data centre according to their use
(in the colocation context, the data centre operators would not even know for which
downstream purpose their clients use of the server rooms offered by the colocation provider).
2.3.1 Data centre demand growth
The estimation of data centre demand builds on three scenarios (measured in MW) identified
in Technopolis et al. (2025), “Study: Cloud and AI”. The analysis presents corresponding
projections for data centre demand across the EU27 region based off the characteristics of three
strategic scenarios (low growth, central, and high growth). These projections illustrate the
potential evolution of enterprise demand over the next decade.
Table 9. Overview of low growth, baseline and high growth scenarios for data centre demand
Scenario CAGR
(2025-2036) Key assumptions and trends
Low growth 10% • Enterprise IT budgets grow conservatively, constrained by
macroeconomic headwinds, regulatory uncertainty, and
cautious capital deployment across the EU27.
• IT spending as a share of enterprise turnover holds steady at
around 3.1%, reflecting a cautious approach to digital
transformation. Within this, spending on data centre grows
more slowly than in previous years, with a CAGR of
approximately 8%, a notable deceleration from the 14%
observed between 2021 and 2025.
• Workload growth remains relatively modest, with most
enterprises focusing on efficiency gains rather than
significant expansion of compute-intensive workloads. AI
adoption progresses slowly, resulting in only incremental
increases in processing and storage demand.
• Digital policy has limited influence on demand in this
scenario. Private enterprises see limited incentives or
pressures to scale AI or data-driven services. The lack of
cohesive policy momentum results in a focus on maintaining
existing infrastructure rather than investing in new capacity
to support advanced workloads.
Central
growth
(baseline)
13% • Enterprise IT budgets across the EU27 continue to expand at
a steady and predictable pace, reflecting a stable
macroeconomic environment and a measured approach to
digital transformation.
• IT spending as a share of enterprise turnover increases
gradually, reaching 3.7% by 2032. From 2032 onwards, this
ratio stabilises, indicating a maturing digital spend profile
with spend only increasing to 3.8% by 2036. Within this
broader trend, investment in data centre infrastructure grows
in line with the historical CAGR of 14% through to 2030,
reflecting sustained demand for compute and storage
infrastructure. This growth supports both the retrofitting of
legacy environments and the deployment of new, AI-capable
facilities, particularly in markets where cloud adoption and
enterprise workloads continue to scale. From 2032 onwards,
the growth rate of data centre spending decelerates, but
48
Scenario CAGR
(2025-2036) Key assumptions and trends
outlays continue to increase, sustained by lifecycle refreshes,
AI-capable retrofits, compliance and resilience upgrades,
and steady colocation demand
• Workloads evolve steadily, with increased deployment of AI,
machine learning, and data-driven services contributing to
higher computational requirements.
• Digital policies boost private sector demand. As
governments roll out e-government services, data-sharing
initiatives, and sector-specific AI guidelines, private
enterprises respond by adopting initiatives which steadily
drive increases in compute and storage requirements.
High
growth
17% • In the fast growth scenario, enterprise IT budgets across the
EU27 rise sharply as digital infrastructure becomes a central
pillar of business strategy. This acceleration is driven by the
widespread deployment of AI, real-time analytics, edge
computing, and data-intensive applications. Organisations
across all sectors and geographies invest aggressively in
next-generation infrastructure, resulting in unprecedented
demand for data centre capacity.
• IT spending as a share of enterprise turnover increases to
4.1% by 2033. After reaching this threshold, the share
stabilises through 2036, with digital operations have become
deeply embedded in business models. Within this expanded
IT budget, the share allocated specifically to data centre
systems grows substantially reaching 23% of total IT
spending by 2035, underpinned by a CAGR of
approximately 11% through to 2036. This growth supports
large-scale, AI-optimised infrastructure rollouts across both
traditional hubs and new emerging markets.
• Workload growth is exponential, driven by generative AI,
large language model training, real-time analytics, and
autonomous operations with enterprises requiring
significantly more computational power.
• Proactive and coordinated digital policies act as a powerful
enabler of private sector demand. Large-scale public
investments in AI, national cloud services, and public sector
digital transformation signal long-term commitment to
digital growth, de-risking private sector investment.
Enterprises accelerate their own AI deployment, edge
computing, and data management strategies to align with
evolving regulatory frameworks, increased data flows, and
market expectations.
As of 2025, all three scenarios converge at an estimated 15.3 GW of installed data centre
demand. However, the trajectories begin to diverge thereafter. By 2030, under the baseline
scenario, demand reaches 37 GW. In the fast growth scenario, driven by increased spending,
enterprise demand increases only slightly more sharply to nearly 40 GW. Meanwhile, the slow
growth scenario sees more modest demand, reaching just 31 GW by the end of the decade. The
49
CAGR from 2025-36 for the low growth scenario is 10%, for the baseline scenario is 13% and
for the high growth scenario is 17%.
Extending the demand forecast horizon to 2036 introduces greater uncertainty, reflecting the
unpredictable nature of long-term market drivers, potential shifts in enterprise digital strategy,
and the impact of emerging technologies such as generative AI and edge computing.
The divergence between the scenarios becomes much more pronounced. In the low growth
scenario, demand reaches 43 GW, while the baseline trajectory leads to 62 GW. The fast growth
scenario sees capacity expand dramatically to over 83 GW, representing a 43% increase over
the central forecast and a 93% increase over our slow growth forecast.
Figure 2. Data centre demand across the EU27 in GW from 2025-36 across three growth scenarios
Source: Technopolis et al. (2025) by means of DataCentreMap, STL analysis
The literature aligns most closely with the status quo scenario, estimating European data centre
demand at around 35 GW by 2030. In terms of growth, the study’s findings are supported by
global projections in the literature, which forecast a compound annual growth rate of 19% to
22% between 2023 and 2030. AI is also consistently identified as a key driver of this demand,
with some sources expecting global data centre capacity to triple as a result .
Based on these projections, a supply-demand gap is expected to emerge by 2030 under all three
scenarios. In the low growth scenario, the shortfall reaches approximately 7 GW, reflecting
modest demand outpacing limited infrastructure expansion. Under the central trajectory
(baseline), the gap widens to around 9 GW, as steady demand growth outstrips the pace of new
capacity being built. In the high growth scenario, the mismatch becomes even more
pronounced, with demand exceeding supply by over 13 GW, driven by accelerated adoption of
AI workloads and large-scale digital infrastructure investment. By 2036 the gap becomes even
more pronounced across all three of the scenarios. In the low growth scenario, the shortfall
reaches approximately 12 GW, in the baseline reaches 19 GW and in the high growth it reaches
23 GW. In the high growth scenario, the gap, though still large at 23 GW, is only slightly higher
than in the baseline. This is due to the assumption that rapid increases in demand are matched,
at least in part, by accelerated capacity deployment enabled by more favourable investment
conditions, improved regulatory agility, and large-scale public-private initiatives. However,
even under this high-investment environment, the pace at which demand scales, particularly
50
for AI and high-performance compute, still exceeds the ability of the market to deliver
sufficient, AI-ready capacity.
Figure 3. Gap between data centre supply and demand in GW in 2036
Source: Technopolis et al. (2025) by means of CATI survey (n=100 companies, 280 Data Centres)
2.3.2 Data centres benefitting from each Policy Option
The analysis of policy options uses the baseline and high growth results as anchor points to
estimate the impact of different policy options, with intermediate values interpolated to reflect
varying scenarios and intervention intensities. Growth trajectories under the different policy
options are thus modelled according to three alternative compound annual growth rate (CAGR)
trends, i.e. 15%, 14% and 13%, which reflect possible deviations from the central forecast
expansion path (12%). To translate capacity expansion (MW) into the number of new data
centres, the following assumptions are made:
• The average data centre capacity has been derived from past deployment trends
• A forward-looking adjustment is applied, assuming an annual increase of 1% in
average data centre size, to capture the tendency toward larger facilities over time
• For each year between 2026 and 2036, the additional MW capacity projected under
the different policy growth scenarios is calculated
• The number of new data centres is then obtained by dividing the incremental annual
capacity (MW) by the estimated average data centre capacity for that year.
• It is then assumed that only a percentage of new data centres each year will benefit
from the policy intervention.
This approach ensures that the model accounts not only for aggregate capacity growth, but also
for the evolution in average facility size, which affects how many new sites are required to
deliver the expected expansion.
The tables below present the estimated number of new data centres (DC) benefitting under each
policy option (PO1-A, PO1-B, PO1-C). These figures are derived by multiplying the projected
number of new DCs coming online each year by the assumed percentage that would qualify
for, or take advantage of, the measures under each option. The modelling distinguishes,
wherever relevant, between the private market (colocation and hyperscalers) and the public
sector. Enterprise data centres are not considered in the analysis.
51
To assess the impact of each option on data centre investment, the outcomes under each policy
scenario have been compared to the baseline scenario, in which no new policy is introduced.
Since the objective of each policy option is to alter investment conditions and increase data
centre deployment, each scenario has been defined to reflect this potential induced build-out.
To reflect uncertainty in the voluntary uptake of the mechanism, the analysis varies the share
of new projects benefitting from the different policy options between 30% (low), 50% (central,
reported below) and 70% (high). This range captures realistic differences in data centre
facilities adoption dynamics, without assuming either full uptake or very limited benefit from
each option.
Table 10. Capacity growth and new data centres under the baseline, considering no intervention
(Source: Technopolis et al. (2025))
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 CAGR
Capacity growth
(MW) 12440 15131 17932 20953 24343 28141 31254 33954 36398 38577 40492 42460 12%
Avg. capacity of
one DC (MW) 13 13.1 13.3 13.4 13.5 13.7 13.8 13.9 14.1 14.2 14.4 14.5 1%
No. of new DC
built 205 211 226 251 278 226 194 174 153 133 136
Table 11. Capacity growth and new data centres benefitting from PO1-A
(Source: Technopolis et al. (2025))
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 CAGR
Capacity growth
(MW) 12440 15131 18189 21573 25409 29754 33518 36893 39965 42722 45168 47712 13%
Average capacity
of one DC (MW) 13.0 13.1 13.3 13.4 13.5 13.7 13.8 13.9 14.1 14.2 14.4 14.5 1%
No. of new DC
built 205 231 253 284 318 273 242 218 194 170 175
% of DC
benefitting from
the option
0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
No. DC benefitting
from the option
0 12 25 43 64 68 73 76 78 77 88
Table 12. Capacity growth and new data centres benefitting from PO1-B
(Source: Technopolis et al. (2025))
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 CAGR
Capacity growth
(MW) 12440 15131 18705 22839 27632 33186 38440 43409 48003 52189 55974 59991 15%
Average capacity
of one DC (MW) 13.0 13.1 13.3 13.4 13.5 13.7 13.8 13.9 14.1 14.2 14.4 14.5 1%
No. of new DC
built
205 269 309 354 407 381 357 326 294 264 277
% of DC
benefitting from
the option
0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
No. DC benefitting
from the option 0 13 31 53 81 95 107 114 118 119 138
Table 13. Capacity growth and new data centres benefitting from PO1-C
(Source: Technopolis et al. (2025))
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 CAGR
Capacity growth
(MW) 12440 15131 18447 22202 26505 31435 35912 40041 43827 47249 50315 53538 14%
Average capacity
of one DC (MW) 13.0 13.1 13.3 13.4 13.5 13.7 13.8 13.9 14.1 14.2 14.4 14.5 1%
52
2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035 2036 CAGR
No. of new DC
built 205 250 280 318 361 324 296 269 241 214 222
% of DC
benefitting from
the option
0% 0% 5% 10% 15% 20% 25% 30% 35% 40% 45% 50%
No. DC benefitting
from the option 0 13 28 48 72 81 89 94 96 96 111
2.3.3 Impact modelling
In order to evaluate the impact of each policy option (PO1-A, PO1-B, PO1-C), a Discounted
Cash Flow (DCF) model has been carried out, for a single data centre, starting at 13 MW
capacity in 2025, based on the estimations calculated by Technopolis et al. (2025).
A dual approach has been used, including a company-level DCF (calculating the Weighted
Average Cost of Capital (WACC) through a build-up methodology) and an external
benchmarking at the MS-level (using WACC averages found in the literature for each MS27),
to ensure that the DCF analysis is both methodologically rigorous and relevant for
policymakers.
Firstly, a company level DCF has been carried out, extracting companies investing and
operating data centres from Data Centre Map28. The following data has been extracted from
companies when publicly available: WACC (if available), Cost of equity, Cost of debt, D/E
ratio and beta. For private companies i.e., companies whose financial data are not publicly
available, the Build-up model has been used to estimate the WACC factors.
The same analysis has been performed at the MS level, using WACC averages found in the
literature. The following parameters have been gathered up or estimated for a 13 MW data
centre: CAPEX, OPEX (using the energy consumption as a proxy), total revenues (fixing the
expected EBITDA margin and power provisioning charged to customers has been kept in an
average 165 €/kW/month29 30) and corporate tax rate31.
Energy price volatility and country-specific dynamics have been partly reflected in the model’s
calibration. Electricity prices used to calculate operational expenses for data centres in the
discounted cash flow model have been based on non-household prices from Eurostat data and
adjusted to reflect the two sourcing models identified in the CATI survey, i.e. grid supply and
power purchase agreements (PPA), using a 40/60 mix. This includes declining prices driven by
increasing renewable penetration, stable PPA prices for new contracts and stable network
charges and non-recoverable taxes and levies. Volatility has been captured through these time-
varying trajectories.
The DCF for a 13 MW project has been then performed as a European average and for each
country under study, with the following assumptions:
• IT CAPEX is disbursed in two phases of 7 and 6MW capacity.
• Ramp up period: the capacity utilisation is adjusted in the first 3 years of the operations: 3
years (50% year 1, 75% year 2, 100%-year 3+).
27https://www.berec.europa.eu/system/files/2024-
07/BoR%20(24)%20102%20BEREC_WACC%20parameters%20Report_2024_1.pdf.pdf 28 https://www.datacentermap.com/ 29 https://www.cbre.com/insights/reports/global-data-center-trends-2025 30 https://www.cbre.com/insights/books/european-real-estate-market-outlook-2025/data-centres 31 https://taxsummaries.pwc.com/quick-charts/corporate-income-tax-cit-rates
53
• Depreciation of the construction is 15 years distributed equally while IT CAPEX presents
different depreciation periods. Maintenance CAPEX has been calculated as a 3% of the
total construction CAPEX yearly.
• The DCF has been presented for the first 10 years of project, using the Gordon Growth
model to calculate the Terminal value (an average 2% perpetual growth rate has been
considered for all countries).
As mentioned above, the WACC used has been taken from the BEREC report that uses the
methodology defined by the European Commission Notice of 2019 for consistent WACC
calculations across EU national regulatory authorities and supports regulatory pricing decisions
in the electronic communications infrastructure sector including data centres32.
The numbers obtained under the previous section have been used in the DCF model to estimate
net benefits for data centre operators. The next present value of an average 13 MW project was
calculated under the baseline scenario and under each option, varying key elements such as
permitting timelines and PUE levels. The NPV difference between each policy option and the
baseline was then multiplied by the number of facilities expected to benefit under each option,
as derived from the possible capacity paths.
It is important to underline that the 13 MW representative data centre corresponds to the
average installed capacity of data centre facilities identified in the underlying database used to
construct the EU baseline. It thus reflects the observed mean of capacity distribution cross
hyperscale and colocation facilities. This size is not held constant over time. As above, the
average facility size is projected to increase based on observed historical trends, thus capturing
the shift towards larger, capital-intensive and possibly also AI-capable facilities. It is also
relevant to note that announcements of new data centres with 200 MW+ capacity are often
planned and built as campus or cluster of multiple facilities implemented as smaller units, e.g.
of 10 to 20 MW, that altogether compose the larger capacity, rather than a single installation.
The heterogeneity in data centres, e.g. hyperscale facilities versus colocation operators, has
been incorporated into operating and financial parameters used to derive costs. Such differences
were captured through PUE assumptions and utilisation rates, with hyperscalers operating at
higher utilisation and lower PUEs than colocation centres on average. These parameters had a
direct impact on electricity consumption and thus operating costs used in the Discounted Cash
flow model. Similarly, as mentioned above, the WACC was calculated considering different
business realities to reflect the different costs of capital. While any averaging compresses the
dispersion, this simplification was needed in our analysis to estimate aggregate costs and
benefits based on observed market dynamics and structures.
2.3.4 Energy efficiency
Starting with an average Power Usage Effectiveness, PUE, of 1.29 each policy option is
expected to achieve different levels of improvement over time as shown in the figure below,
with PO1-C achieving the largest impact because of the contribution of public funding to R&D
and innovation in sustainable technologies.
Colocation PUE was calculated using information from Data Center Map (coming out an
average of 1.39) whereas hyperscaler PUE was calculated using the latest reported hyperscaler
figures (coming out at an average of 1.13). The result PUE of 1.29 for 2025 is the capacity-
32 BEREC. (2025). BEREC report on WACC parameter calculations according to the European Commission’s WACC Notice
of 6th November 2019 (WACC parameters Report 2025). Body of European Regulators for Electronic Communications.
https://www.berec.europa.eu/en/all-documents/berec/reports/berec-report-on-wacc-parameter-calculations-according-to-the-
european-commissions-wacc-notice-of-6th-november-2019-wacc-parameters-report-2025
54
weighted average of PUE levels identified for colocation and hyperscale facilities in the EU.
This approach was used to capture the differences in efficiency across businesses of different
sizes. Moreover, it has been validated with the industry during interviews and workshops, for
which it was also subject to refinements and further changes as some stakeholders provided
additional data on their PUE levels and market presence.
PUE improvements across the policy options are expected based on different mechanisms and
time horizons. Under PO1-A, such improvements are mainly driven by the diffusion of best
practices produced by the guidelines and as part of the established forum for exchanges. These
are assumed to generate positive spillovers on the use of increasingly energy-efficient
technologies beyond the baseline, where individual companies (usually the largest ones) have
most of the incentives to invest in such technologies, without necessarily sharing the outcome
of their investment. Under PO1-B, PUE reductions are driven by the stronger incentives for
data centres to increase energy-efficiency due to the sustainability requirements linked to
accessing the fast-track areas. These conditions are expected to increase incentives for
operators to adopt higher-efficiency and sustainable solutions relatively early, as this could
become a prerequisite for benefitting from accelerated permitting. Under PO1-C, in addition,
data centre operators would benefit from targeted R&D programmes to optimise energy
efficiency and, in some cases, deployment funds for particularly sustainable technologies and
strategic uses. Since these programmes require time to develop, test and de-risk new solutions,
especially for smaller operators, the most substantial PUE improvements are expected to
materialise in the later years of the period. Quantitatively, this is reflected in a somewhat flatter
PUE reduction in the early years and a comparatively steeper profile towards 2032–2035, when
a larger share of new capacity can incorporate the mature, higher-efficiency designs.
Figure 4. Assumptions in terms of PUE improvement across the different policy options for colocation
providers and public sector 2025-35 (Source: Technopolis et al. (2025))
2.4. Estimates related to Public Procurement
2.4.1 Number of public authorities procuring cloud and AI computing services
With regards to the number of public authorities that procure cloud and AI computing services
and could participate at the EU-level procurement process, the assumption made is that there
is one public authority in each NUTS1 and NUTS2 areas, and in most of the NUTS3 areas33,
33 NUTS1, NUTS2 and NUTS3 areas in Europe are close to 1,500 (https://ec.europa.eu/eurostat/web/nuts). An adjustment has
been made to slightly reduce the number of NUTS3 areas with a public authority procuring cloud and AI computing services
in the bigger member states where the administrative distribution at that level is more granular.
55
resulting in a figure of 1 206 public authorities in Europe. This figure has been confirmed when
filtering the number of unique buyers of cloud and AI computing services in the procurement
notices registered in the years from 2021 to 2024 at the TED portal34 (the filtering has been
done applying a search of “cloud” and “AI” strings over the TED notices with Common
Procurement Vocabulary (CPV) 72 and 48). Not all public authorities launch procurement
processes every year; in 2024, for instance, there were just 463 Public Administrations (PA)
opening cloud and AI computing service tenders.
2.4.2 Public procurement contracts of cloud and AI computing services
The following table shows the contract award notices in EU27 by public administrations, as
extracted from TED using the TED API35. For the ICT award notices, ‘contract award notices’36
with CPV code 72 and 48 have been considered, whereas for cloud and AI, combinations of
the keywords “Cloud” or “AI” have been used as a subset of the previous. 2024 is taken as the
baseline for the calculations. Then, the count of each of the results have been extracted, whereas
for the share, the percentage has been calculated.
Table 14. Count of Public procurement award notices in TED for 2024. (Source: Technopolis et al. (2025),
elaborated from TED data)
Count of Award Notices 2024
Cloud & AI awards 1 043
ICT awards 33 299
Share of Cloud & AI in ICT awards 3.13%
Total awards 30 7249
Share of Cloud & AI in total awards 0.34%
For the calculation of the CAGR the period spanning from 2021 – 2024 is considered. 2020
has been disregarded due to the COVID-19 pandemic effects. The values resulting from a query
from TED are shown next:
Table 15. Evolution of public procurement award notices 2021 – 2024. (Source: Technopolis et al. (2025),
elaborated from TED data)
Count of Award Notices 2021 2022 2023 2024 Avg 2021-2024
Cloud & AI Awards 815 915 1,017 1 043 948
ICT Awards 29 235 31 566 34 673 33 299 32 193
Total Awards 272 004 292 162 316 899 307 249 297 079
Share Cloud & AI vs Total 0.30% 0.31% 0.32% 0.34% 0.32%
The data above results in a CAGR for the period 2021 – 2024 of 8.6%. Conservatively, it is
assumed that the CAGR between now and 2030 will grow at the current pace whereas it will
grow half as fast afterwards.
Table 16. CAGR for public procurement award notices for ICT and cloud and AI. (Source: European
Commission based on data provided by Technopolis et al. (2025))
2021-24 2025-30 2031-36
34 TED is the official source for public procurement in the EU. However, it is to be noted that a percentage of the data sets do
not present complete information, with some fields missing. 35 https://docs.ted.europa.eu/api/latest/index.html 36 TED also stores the contact notices which include closed procedures with no awardees. This number is however rather small
and could fall under the margin of error.
56
CAGR Cloud & AI Awards 8.6% 8.6% 4.3%
CAGR ICT awards 4.4% 4.4% 2.2%
CAGR Total awards 4.1% 4.1% 2.1%
With regards to the highly critical use cases, the assumption is that 10% of all contract award
notices represent these narrowly scoped use cases37. Applying this to 2024 figures, the number
of procurement procedures of cloud and AI computing services in that year is 52. The CAGR
for this specific case is considered to be the same as for other types of use cases. The data
provided in TED is not disaggregated enough to provide information on the criticality of the
use cases and in some cases, the sectors are not present, making very difficult a further
extrapolation.
The above assumptions yield the following projections:
Table 17. Projected number of cloud & AI awards (Source: European Commission based on data
provided by Technopolis et al. (2025))
Projections Cloud & AI Awards Cloud & AI awards Highly
critical services
2025 1 132 113
2026 1 229 123
2027 1 335 133
2028 1 449 145
2029 1 573 157
2030 1 708 171
2031 1 781 178
2032 1 858 186
2033 1 937 194
2034 2 020 202
2035 2 107 211
2036 2 197 220
An important differentiation along the document shall be the delineation between procurement
tenders and bids. A tender is the procedure that a contracting authority launches while a bid is
an offer that a cloud and AI computing service provider shall submit in response to the call for
tender. The assumption is that as an average there are 3.5 bids for each tender launched and
awarded.
Where audit is voluntary, the assumption is that 30% of the tenders will require providers to
offer an audited service, unlike when audit is mandatory, where this requirement is 100%.
The projections resulting from these assumptions are depicted next:
Table 18. Projections for the number of bids of cloud and AI computing services that need to be audited
(Source: European Commission based on data provided by Technopolis et al. (2025))
37 Member States have reported in various settings that 10% of their services fall under the category of highly critical services.
57
Projections
Cloud & AI bids for highly critical
services (HCS) audit mechanism
mandatory
Cloud & AI bids for HCS audit
mechanism voluntary
2025 396 119
2026 430 129
2027 467 140
2028 507 152
2029 551 165
2030 598 179
2031 623 187
2032 650 195
2033 678 203
2034 707 212
2035 737 221
2036 769 231
2.4.3 Public procurement contract values of cloud and AI computing services
The baseline data taken into consideration for the contract values of cloud and AI computing
services procured has been extracted from the TED database using the TED API38. To calculate
the ICT contract values, the sum of the contract values of the Contract Award notices under
CPV72. For cloud and AI contract values, the keywords “cloud” or “AI” have been used to
discriminate.
Table 19. Public procurement contract values 2021 – 2024. (Source: Technopolis et al. (2025)) elaborated
from TED data)
Value of Award notices (bn EUR) 2021 2022 2023 2024 Avg
21-24
CAGR
21-24
Cloud & AI value of award notices 2.8 5.1 12.5 9.2 7.4 48%
ICT Value of award notices 77.1 98.7 113.9 124.9 103.6 17%
Total Value of award notices 1 031.5 1 041.1 1 195.0 1 356.1 1 155.9 10%
Share of Cloud & AI in ICT awards 3.66% 5.14% 10.99% 7.38% 7.14% 26%
Share of Cloud & AI in total awards 0.27% 0.49% 1.05% 0.68% 0.64% 35%
2.4.4 Estimated number of distinct cloud and AI computing service providers in
the EU
There is no common registry of cloud and AI computing service providers. Furthermore, the
provision of said services is not categorized under NACE J. Under these constraints and in
order to understand the current number of providers in the EU, it has been performed a desk
search of Cloud Service Providers along with crossing data from various associations, market
reports, TED and other documents. While this approach is limited, it has managed to capture
the most relevant CSPs in the EU, both EU headquartered and non-EU.
The results of this manual search yields in the following:
38 https://docs.ted.europa.eu/api/latest/index.html
58
Table 20. Number of cloud and AI computing service providers. (Source: Technopolis et al. (2025))
Cloud and AI computing service providers Number
Total 410
EU-based 350
non-EU based 60
2.4.5 Audited Sovereign cloud and AI computing service providers
To calculate the number of audited sovereign cloud and AI computing service providers, the
number of services listed on the FedRAMP marketplace with Impact level moderate and high39
has been taken as baseline for the assumptions. FedRAMP is a programme that has been
running since 2012 and, as of October 2025, there are currently 533 services at impact level
moderate and high with the status “FedRAMP authorised” or “FedRAMP in process” or
“FedRAMP ready”.
The baseline situation is that in the first year, 30 cloud and AI computing service providers
shall get audited, reaching up to 600 in 5 years. To reach to this number, this would imply a
CAGR of 82%. The assumptions for the number of positive audits granted the first year are in
line with how many authorisations FedRAMP authorised its first year of the programme,
namely 2013 (10), and cross-referenced with SecNumCloud values40.
2.5. Indirect benefits from cloud adoption
Most of the policy measures contribute to the development of European data centre
infrastructure and cloud computing capacity, a richer European cloud and AI computing service
offering and a more dynamic technological and industrial ecosystem, contributing to a higher
adoption of cloud. This will improve the efficiency in software and IT projects, increasing
productivity of IT staff, ensuring the continuity of business processes (less downtime) and
facilitating migration processes. This impacts not only the efficiency of the IT staff but also the
flexibility and capability of the European private and public sector to adapt in an agile way to
evolving market needs and conditions, and to the dynamic technology landscape, increasing its
competitiveness. This could have been measured in terms of effort saved in IT tasks (FTEs),
but it would have been difficult to allocate the specific contribution of each specific measure,
adding potential risk of double counting their effect. The policy measures related to federation
and joint procurement are making the biggest contributions, while most of the others will have
a more limited impact.
Cloud computing services generally offer a significant number of benefits, among them also
the drive or productivity. The OECD reports strong productivity effects through cloud software
investment, especially for low productivity firms, suggesting that such companies derive
greater productivity benefits from this investment when fixed costs are lower and scaling-up
opportunities are higher. It is implied that cloud adoption has the potential to support the
39 https://marketplace.fedramp.gov/products 40 Source: https://cyber.gouv.fr/sites/default/files/document/catalogue-produits-services-qualifies-agrees-certifies-anssi_0.pdf
. The list of services qualified are listed in Chapter 5.1. This table shows 9 individual cloud service providers qualified but 16
cloud services with the following breakdown: 2 services in 2023, 2 services in 2024 and 12 in 2025.
59
productivity catch up of latecomer firms, while low cloud adoption could still hamper broader
productivity gains41.
3. IMPACTS OF POLICY MEASURES IN TERMS OF COSTS AND BENEFITS
The assumptions presented below and used for the estimation of the impacts of policy measures
on key stakeholders draw on recognised literature sources and observable practices in the
industry and the public sector. To test their robustness, the supporting study led by Technopolis
conducted a series of one-to-one interviews with a diverse set of relevant stakeholders. These
assumptions were further scrutinised in the study’s final validation workshop, which led to an
adjustment of some parameters.
3.1. PM1: Expanding the Alliance for Industrial Data, Edge and Cloud with a
working group on data centres and extending membership to relevant
players
Under this measure, the European Commission would set up an additional Working Group
(WG) under the Alliance for Industrial Data, Edge and Cloud to include data centre operators.
This would facilitate the exchange of best practices on the deployment of data centres with the
objective to foster EU-located capacity, while collectively addressing the obstacles that hold
back data centre investment and expansion in the EU.
Impact on data centre deployment acceleration. This new working group may improve
deployment speed for the participating alliance members. The reduction in data centre
deployment time will likely be very slow due to the limited reach of the initiative, the time
required to produce and disseminate results, and the lack of structural effects.
Adjustment costs for data centre operators. Participation in the new working group is
voluntary at entry but implies expectations of active attendance and contribution once
membership is granted. This is expected to generate adjustment costs for stakeholders willing
to participate. Under the Better Regulation Toolbox, these can be considered consultation-
related burdens, although voluntary. Thus, the recurring source of adjustment costs for data
centre operators arises from their involvement in the newly established Working Group, who
would face additional costs for travel and participation in this new structure. This is based on
2.5 days of effort per participant for each in-person meeting (i.e. 1 day of attendance and 1.5
days of preparation), assuming two General Assemblies (in-person meetings) per year. In
addition, participation in 10 webinars annually has been costed at 0.5 days of effort (or 4 hours)
per participant per webinar, including both attendance and preparatory work. Please see the
formula below. These are estimated at approximately EUR 5,400 per operator each year, made
up of EUR 2 000 for in person engagement, EUR 2 000 for webinar participation and EUR 1
400 for travel expenses (estimated at EUR 700 per trip per participant). The Working Group is
expected to involve 30 new delegates: therefore, the total recurrent adjustment cost stems from
multiplying this total per participant by 30 and discounting it at present value.
Recurrent adjustment cost per participant,year = (Days per in-person meeting × No. of in-person
meetings × labour cost) + (Travel cost × No. of in-person meetings) + (Days per webinar ×
No. of webinars × Labour costs)
41 OECD Compendium of Productivity Indicators 2025
60
Additional effort has been considered for around 40% of the Working Group members, who
would be active in producing documents, i.e. roadmaps or reports, each year, estimated at
around 20 days of additional work per year, for a total of EUR 96 008 each year. This estimate
reflects the contribution of the chairing team and rapporteurs, elected from within the WG, who
are responsible for supporting and implementing the group’s daily activities. Although not
mandatory, these recurrent activities would be undertaken in the operators’ best interest to gain
information and reduce uncertainty in project development. For this, the formula is:
Document production 40% subgroup/year = (40% × Total No. of participants) × (Effort of active
engagement × labour cost)
Total recurrent adjustment costs for data centre operators under PM1 are thus estimated at EUR
2.2 m (NPV, 10-years).
Cost savings for data centre operators. The cost savings for data centre operators are explained
as indirect impacts as the reach of this measure is by design not enough to justify a
monetisation.
No new costs or cost savings have been considered for national public authorities, as they would
not participate in this new Working Group and are already participating in the General
Assembly (as members of the Member State Cloud Cooperation Group), to which this new
working group would report.
Administrative costs for the European Commission. The establishment and onboarding of the
new Working Group entails one-off administrative costs for the Commission. The related costs
are estimated considering the time taken to onboard the new delegates in the Working Group
during the first year, considering 20 days or 0.09 FTE-year in 2027. This would be facilitated
by the existence of an existing working group of cloud and edge providers 42. Recurrent costs
would reflect the additional administrative costs linked to the coordination of the new Alliance
WG, i.e. preparing the work programme, meeting planning, agenda-setting, inviting
stakeholders, circulating minutes, drafting reports, managing documents, dealing with day-to-
day queries, running webinars and preparing them, for which 40 days per year have been
considered. These would include the additional meetings per year, preparation and reporting
towards the Alliance Members. Total administrative costs have been calculated as the one-off
administrative costs in year 1 plus the discounted value of the recurrent annual administrative
costs over 10 years.
Total administrative costs = one-off administrative costs + recurrent administrative costs
= (ΔT × labour cost)year 1 +∑ ( × )
(1+0.03)−1
10
=2
Total (one-off and recurrent) administrative costs for the Commission under PM1 are thus
estimated at EUR 0.1 m (NPV, 10-years).
Indirect impact: The WG will help to foster EU-located capacity, providing advice, guidelines
and recommendations to remove or reduce the obstacles that hold back data centre investment
and expansion in the EU. This contribution will help the European Commission and national
public authorities (NPAs) to orient their policies and data centre operators to fine tune their
deployment plans. However, it is also possible that larger companies will be willing to invest
42 Assumptions behind these numbers have been validated with EC staff currently working on the Alliance for Industrial Data,
Edge and Cloud.
61
more time and resources into the participation compared to SMEs, and that the measure will be
more costly in relative terms for SMEs – despite this is hard to quantify.
Sensitivity analysis: No sensitivity analysis has been conducted on this policy measure as the
amounts at stake are too small.
Results: For PM1, the estimated costs for the implementation of this measure across all
stakeholders are EUR 2 339 297. These costs reflect the recurrent administrative costs borne
by the additional Commission staff for the ongoing coordination, meeting and time to manage
the working group activities (EUR 130 696). On top of these, a small one-off administrative
cost (EUR 7 661) is foreseen for initial setup tasks. The measure also includes the adjustment
costs expected to be incurred by data centre operators and CSPs from participating in the WG.
Table 21. Expanding the Alliance for Industrial Data, Edge and Cloud with a workgroup on data centres
and extending membership to relevant players (PM1)
Cost types
Data centre
operators
(€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - 7 661 7 661 7 661 7 661
recurrent administrative - - 130 696 130 696 130 696 130 696
one-off adjustment - - - - - -
recurrent adjustment 2 200 940 - - 2 200 940 2 200 940 2 200 940
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 2 200 940 - 138 357 2 339 297 2 339 297 2 339 297
Total benefits - - - - -
Net impact 2 200 940 - 138 357 2 339 297 2 339 297 2 339 297
3.2. PM2: Creating a forum for exchanges between relevant public and private
stakeholders involved in the buildout of data centres
Under this measure, the Commission would establish a dedicated forum to facilitate structured
exchanges between the key public and private stakeholders involved in the deployment of data
centres, with the objective of accelerating the deployment of innovative and sustainable
capacity in the EU. These fora would provide a structured platform for dialogue and
coordination, enabling shareholders to share knowledge, identify challenges and coordinate
approaches across the different stages of data centre planning, construction and operations. The
forum would bring together all relevant actors, including transmission system operators
(TSOs), connectivity providers, data centre operators, equipment manufacturers, and local and
regional authorities. The purpose of the forum would be to support the uptake of advanced and
sustainable technologies that can increase capacity while reducing the environmental footprint
of data centres and to identify and address the obstacles currently hindering investment and
expansion. These exchanges would also allow to include discussions on the needed
communication regarding the trade-offs, benefits and potential mitigation strategies for public
contestation of data centres. By fostering early and continuous exchanges, the forum would
enable stakeholders to align on technical requirements, investment priorities, and regulatory
expectations.
Impact on data centre deployment acceleration. This forum is expected to contribute to creating
a more predictable and investment-friendly environment for data centre deployment,
supporting the EU’s objectives of technological leadership, strategic autonomy, and sustainable
62
digital infrastructure growth. This is expected to increase efficiencies in the deployment of a
number of new data centres, especially of those operators participating in the forum. This would
be the consequence of increased exchanges of best practices, and reduced interactions between
economic operators and authorities.
Adjustment costs for data centre operators. Recurrent participation in newly established fora
is classified as a recurrent adjustment cost for organisations choosing to participate on an
ongoing basis and are expected to contribute actively once involved, i.e. in terms of resource
commitments associated with meetings, preparation and coordination. Assuming two in-person
events per year and five online webinars per year, the participation requirement amounts to
approximately 7.5 staff-days per participant per year, plus travel costs per in-person trip.
Participation is expected from 30 data centre operators and 10 CSPs active in the deployment
of data centres, reflecting the forum’s focus on data-centre deployment issues. The cost for
other private sector stakeholders is not factored.
Total adjustment costs for data centre operators under PM2 are thus estimated at EUR 1.5 m
(NPV, 10-years)
Cost savings for data centre operators. The cost savings for data centre operators are described
as indirect costs as the magnitude and reach of the measure is by design not big enough for
these costs to be monetised.
Adjustment costs for national public authorities.Recurrent adjustment costs for public
authorities are primarily linked to their participation in the exchanges. For simplicity, it is
assumed that 27 national authorities take part each year, although in practice representation
may also be delegated to local authorities, depending on their involvement in data centre
deployment. These costs cover attendance and preparation for two in-person meetings annually,
as well as participation in a programme of five webinars per year for each authority.
Total recurrent adjustment costs for national public authorities under PM2 are thus estimated
at EUR 0.8 m (NPV, 10-years).
Recurrent savings for national public authorities.The cost savings for public authorities are
described as indirect costs as the magnitude and reach of the measure is by design not big
enough for these costs to be monetised.
Administrative costs for the European Commission. The Commission would incur one-off
administrative costs to organise the exchanges, covering tasks such as defining the rules of
procedure, issuing invitations and setting up the platforms for interaction. These start-up costs
are estimated at approximately 40 staff days (equivalent to 0.2 FTE) for the first year of
operation. Recurrent costs would arise from the supervisory and organisational role of the
Commission in managing the forum. This is estimated at 40 staff days annually (0.2 FTE) for
coordination and oversight, including the time dedicated to organising and participating in two
in person events and five webinars per year. Furthermore, the cost of hosting one in-person
event per year at the Commission’s premises is estimated at EUR 5 000.
Total administrative costs (one-off and recurrent) for the Commission under PM2 are thus
estimated at EUR 0.2 m (NPV, 10-years)
Indirect impact:For data centre operators cost savings under this measure would stem from
coordination efficiencies due to these discussions focused on data centre development across
the EU. The forum is expected to reduce duplicated outreach and bilateral exchanges with key
63
stakeholders, such as permitting authorities, TSOs and local administrations when building new
data centres.
Recurrent cost savings for national public authorities are expected from the centralisation of
exchanges through the forum, which reduces repeated bilateral interactions with providers.
Sensitivity analysis: As above, no sensitivity analysis has been conducted on this policy
measure as the amounts at stake are too small.
Results: For PM2, the estimated net costs for the implementation of this measure are EUR 2
603 530. These stem mostly from recurrent adjustment costs (EUR 2 372 210) stemming from
the participation in the forum by the national public authorities and the economic operators. A
one-off administrative cost (EUR 15 322) is foreseen for initial setup tasks borne by the
Commission. Recurrent administrative costs of 215 998 are foreseen in relation to coordination
for the management of the forum discussions. No direct savings are expected.
Table 22. Creating a forum for exchanges between relevant public and private stakeholders involved in
the buildout of data centres (PM2)
Cost types
Data centre
operators
(€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - 15 322 15 322 15 322 15 322
recurrent administrative - - 215 998 215 998 215 998 215 998
one-off adjustment - - - - - -
recurrent adjustment 1 553 531 818 679 - 2 372 210 2 372 210 2 372 210
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 1 553 531 818 679 231 320 2 603 530 2 603 530 2 603 530
Total benefits - - - - - -
Net impact 1 553 531 818 679 231 320 2 603 530 2 603 530 2 603 530
3.3. PM3: Adopting guidelines on building sustainable data centres in the EU
Under this measure, the Commission would develop non-binding EU guidelines for the
sustainable planning, design and operation/deployment of data centres, which complement
existing initiatives, e.g. DG ENER’s grid package, the EED, and the simplification of
environmental permitting. The guidelines would prioritise voluntary coordination over top-
down prescription. Their purpose is to address practical bottlenecks encountered by market
participants and national public authorities, notably in terms of permitting, grid access, and
financing, while promoting the uptake of innovative and resource-efficient technologies. They
would also include recommendations on identifying suitable areas for data centre deployment.
The guidelines would be co-created with Member States, relevant regulators, local authorities,
TSOs/DSOs, data centre operators and CSPs. Drafting would also draw on the forum
established under PM2, public consultations, and periodic review to reflect technological
progress and implementation feedback. The text would be adaptable to national or regional
contexts, allowing authorities to reference it if consistent and helpful with respect to domestic
frameworks.
Governance cadence. Initial issuance of the guidelines in 2027, followed by triennial reviews
(2030, 2033, 2036) to reflect technological progress and implementation feedback.
64
Impact on data centre deployment acceleration. The baseline capacity growth presents a 12%
CAGR in data centre capacity (2025-2036). Under this option, it is assumed that the capacity
growth rate (CAGR) increases slightly to 13% (see section 2.3.1 of this Annex). The guidelines
are expected to reduce design, assessment and administrative tasks and applying best practices
in datacentre rollout and operation. The scope of this measure covers data centre operators and
CSPs active in the building of data centres in the EU. Impacts scale with the share of
organisations adopting the guidelines and the share of new projects applying them (see section
2.3.3). Since the use of the guidelines is voluntary, it is assumed that the share of new projects
applying the guidelines increases from 0% in 2027 to 50% by 2036, to consider the differing
views received during the consultations of Technopolis et al. (2025) about the usefulness and
rate of uptake of the guidelines.
Adjustment costs for data centre operators. The impact of the guidelines on businesses is
modelled as adjustment costs using the Standard Cost Model (SCM), which accounts for the
time and resources needed to understand and apply the guidelines. In year 1, each participating
operator incurs one-off adjustment costs for consultation and discussions of the guidelines with
authorities. Participation in the initial consultation process is estimated for 30 data centre
operators and 10 CSPs43, considering around 27.5 staff days per operator per year (equivalent
to 0.125 FTE or around 2-3 full working days per month, spread across the year)44. The
continued involvement in the revisions/updates of the guidelines is estimated at an additional
20 days every three years, but only for the 30 data centre operators and 10 CSPs originally
engaged in the drafting. These days of effort per operator for participating in the discussion of
the guidelines every three years and until 2036 would entail adjustment costs of EUR 1.58 m
in total (NPV), assuming 30 data centre operators and 10 CSPs.
For operators adopting the guidelines, 20 days per operator are considered for
checking/adjusting planned solutions for the roll-out of data centres or of existing processes. It
is assumed, backed by the results of the Final Workshop of the study carried out by Technopolis
et al. (2025), that 50% of the approximately 400 CSPs and data centre operators active in
building data centres in the EU would adopt the guidelines and adjust their processes.
Consequently, adjustment costs are estimated to amount to around EUR 29 317 per operator.
Total adjustment costs (one-off and recurrent) for data centre operators under PM3 are thus
estimated at EUR 7.4 m (NPV, 10-years)
Administrative cost savings for data centre operators. The benefits generated by voluntary
guidelines, e.g. fewer mistakes, reduced rework, faster processes and improved coordination,
are classified as administrative burden reductions. The impact of the guidelines is modelled by
assuming shorter administrative timelines using a SCM approach. The guidelines are expected
to reduce the incidence of mistakes during the pre-operation phase of building a new data
centre. This is modelled as a central 10-days saving (min 5 days; max 15 days), with an
increasing number of data centres benefitting from this measure (subject to sensitivity analysis).
The time saving benefit of the voluntary guidelines is calculated as the time saved per facility
that adopts the guidelines (ΔT) multiplied by the relevant labour cost, the total number of new
facilities, and the estimated adoption rate. This follows the SCM formula:
43 The number of operators engaging with the drafting of the guidelines represents the average number of operators typically
engaging in such processes, i.e. not the total number of operators in the EU. 44 Although this consultation is voluntary and not a legal obligation, operators are likely to engage due to their interest in
influencing a framework that directly affects their operations
65
Total savings = ΔT × labour cost × s × No. new facilities
Total administrative cost savings for data centre operators under PM3 are thus estimated at
EUR 2.1 m (NPV, 10-years)
Adjustment costs for national public authorities. For public authorities that participate in
drafting the guidelines, adjustment costs would be related to their participation in the drafting
of the guidelines and their internal adoption, in terms of familiarisation and optional adaptation
of local checklists and templates, estimated at 30.5 days per year (0.14 FTE) or around 2-3 days
per month of full-time work per national authority spread across the year of adoption. If one
representative from each Member State would participate, this would translate into
approximately EUR 236 575 in year 1. Thereafter, adaptation of the new templates/tools
together with the participation of authorities in the review of the guidelines every three years
are expected to require 20 staff days (0.1 FTE). This would entail recurrent adjustment costs of
EUR 155,131 each update cycle, assuming all Member States would adopt the guidelines.
Total recurrent adjustment costs for national public authorities under PM3 are thus estimated
at EUR 0.6 m (NPV, 10-years)
Administrative cost savings for national public authorities. Harmonised EU guidelines are
expected to limit the need for national authorities to interpret divergent compliance questions
during permitting. Authorities are expected to realise administrative burden reduction once the
guidelines are applied, including reduced errors, faster processing with fewer clarifications, and
improved coordination with operators. These time and resource savings are counted as
administrative cost savings in the impact assessment and are quantified using Standard Cost
Model principles. From 2027 onwards, clearer information requirements are assumed to
generate lasting process efficiencies, reducing administrative efforts per project and lowering
disputes with data centre operators, as well as the need for repeated iterations with applicants.
Given that operators are expected to save 10 days per project, a proportional reduction of
approximately 20% on the authority side is applied, resulting in an estimated saving of 2 days
(min 1 day; max 3 days) per project (ΔT). This reflects reduced requests for clarification, fewer
iterative corrections and smoother coordination thanks to the guidelines. The savings would
largely depend on the voluntary adoption of the guidelines by the new data centre projects,
allowing authorities to save time in terms of decreased interactions with them. Benefits are
estimated considering they would accrue to around 50% of national public authorities (central
estimate) in the EU dealing with data centre projects.
Cost savingsPA = ΔT × labour cost × s × .
× No of MS
Total administrative cost savings for national public authorities under PM3 are thus estimated
at EUR 0.3 m (NPV, 10-years)
Adjustment costs for the European Commission. The Commission would be in charge of
developing the guidelines, requiring one-off adjustment costs covering the drafting,
consultation, translation and dissemination. These tasks are estimated to require around 110
staff days during the first year (0.5 FTE). Thereafter, periodic improvements and reviews,
comprising an update every three years, would require about 55 staff days (0.25 FTE) for each
update cycle.
Total adjustment costs = one-off adjustment costs + recurrent adjustment costs = (ΔT ×
labour cost)year 1 + (ΔT × labour cost)year update
66
Total adjustment costs for the European Commission under PM3 are thus estimated at EUR 0.1
m (NPV, 10-years)
Indirect impact: The guidelines are supposed to support cross-fertilisation of best practices,
especially for those operators with operations across EU27 borders, and by doing that they
should work towards the improvement of the investment climate for green and sustainable
technologies for data centres rollout.
Sensitivity analysis: The key parameters tested for sensitivity include (1) the adoption rate of
the guidelines in terms of the number of projects concerned, (2) time saved per project for
operators and authorities. For such purposes min-max were created for these parameters. The
number of projects benefitting from the guidelines ranges from 30% in the low scenario to 70%
in the high scenario, with a central estimate of 50%. A low case combines a conservative
adoption rate and limited time savings, while a high case combines a high adoption rate with
higher efficiency gains.
Results: The table below presents the aggregated costs for all stakeholders for the
implementation of this measure. Total implementation costs are estimated between EUR 4 948
939 to EUR 15 841 158, while total cost savings range from EUR 1 298 171 to EUR 4 191 108,
with a central value of EUR 2 383 022. The measure is therefore costly in all cases. The min
and max scenarios are largely driven by the adjustment costs required and consequent cost
savings expected.
Table 23. Creating guidelines on building sustainable data centres in the EU (PM3)
Cost types Data centre
operators
Public
authorities
European
Commission
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - - - - - -
one-off adjustment 2 080 146 - 40 907 2 121 053 1 305 468 4 159 557
recurrent adjustment 5 363 245 612 097 51 523 6 026 866 3 653 471 11 681 301
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 7 443 391 612 097 92 431 8 147 919 4 948 939 15 841 158
Total benefits (2 103 643) (279 379) - (2 383 022) (1 298 171) (4 191 108)
Net impact 5 339 748 332 718 92 431 5 768 775 3 650 768 11 650 050
3.4. PM4: Project facilitators for the roll-out of data centres
Under this measure, Member States would simplify the administrative efforts required to roll
out data centres, by designating a project facilitator acting as a central coordinator for the
permitting process at national level, where this does not already exist. Member States would be
free of designating the facilitator as per their own structures, e.g. a service within their central
administration or within an agency. The project facilitator would be responsible for
accompanying the applicants in securing all relevant authorisations (including environmental
assessments, construction permits and/or grid connection permits) and coordinating with other
national agencies/administrative entities (e.g. possible future single points of contact for
environmental assessments) involved in the permitting process on one side, and with the data
centres on the other. It would also provide support to the data centre operators in assessing
whether projects qualify to be recognised as strategic, thereby enabling them to benefit from
access to funding, helping data centres demonstrating readiness when it comes to grid
67
connection, and dedicated administrative support. For environmental authorisations, the project
facilitator would interact directly with the designated authorities or through the national single
point of contact (SPOC), ensuring that there is efficient coordination.
Evidence from Technopolis et al. (2025) covering 12 Member shows that overall permitting
timelines, i.e. including zoning & land allocation, building permit, utilities & grid connection,
environmental permit/s currently range 32 months on average (see section 98 of this Annex for
a more detailed overview).
Impact on data centre deployment acceleration. The baseline capacity growth presents a 12%
CAGR in data centre capacity (2025-2036). Under this option, it is assumed that the capacity
growth rate increases to 15%. This uplift reflects the assumption that shorter, more predictable
permitting procedures reduce barriers to investment, accelerate time to market, and improve
certainty for operators, thereby stimulating additional capacity deployment. The scope of this
measure covers data centre operators and CSPs active in new data centre builds in the EU (see
section 2.3 of this annex for further details). It is assumed that 50% of the new builds will be
adopting this measure, in a staged adoption going from 0% in 2027 until 50% in 2036. This is
a conservative assumption used for modelling purposes only, with the caveat that the detailed
assumptions of number of new builds forecasted per country per year is not considered, and not
knowing how many countries will adopt this measure.
Costs for data centre operators. Data centre operators will engage with the facilitator to submit
necessary documents, attend joint meetings and receive guidance on the permitting processes.
From the operator’s perspective this should replace the numerous, fragmented integrations with
individual authorities that currently occur, resulting in limited additional costs. Although
operators may incur some initial costs to familiarise themselves with the facilitator's
procedures, these are likely to be outweighed by benefits in terms of fewer repeated
submissions and shorter procedures (see below). As the measure is expected to simplify the
process without increasing the overall time spent on permitting, no separate administrative cost
is considered for data centre operators.
Direct economic benefits for data centre operators. The most relevant benefits of this measure
arise from the direct economic benefits of earlier cash-flow realisation for one project linked to
shorter permitting times achieved through the accompanying effects of the facilitator, i.e. the
reduction in administrative processing time and the resulting acceleration of the overall
development timeline for new data centre projects45. These benefits represent a subset of all the
potential improvements generated by the measure, but they are the ones that can be robustly
monetised. To capture them, the analysis applies a project-level Net Present Value (NPV)
approach (discounted cash flow) that compares the baseline permitting duration with the
accelerated scenario and measures the economic value of bringing construction and commercial
operations forward, discounted at the project WACC (6%). Earlier time-to-market gains from
faster permitting are reflected directly in the DCF through a shorter period before revenues are
realised and thus earlier start of cash flows, while reduced operating costs from improved
energy efficiency (declining PUE, baseline scenario) are included as part of the OPEX
assumptions over the life of the project. Risk-related efficiency gains, e.g. better certainty on
45 Although all parameters (PUE, utilisation rate, revenues and costs) are held constant between the baseline and this policy
scenario, the discounted cash-flow model is inherently non-linear in time. Therefore, shortening the permitting phase brings
forward the entire operating cash-flow profile, reducing discounting on the most value-relevant early-operation cash flows.
68
siting, permitting and investment conditions, are considered indirect benefits rather than
appearing directly in the model46.
The simplified process through the use of a project facilitator is estimated, based on interviews
with stakeholders and validated during the final validation workshop, to shorten the average
permitting duration for a new data centre facility by 6 months, from an average baseline of 32
months to around 26 months duration from the development phase to operations. The NPV gain
associated with this reduction in permitting timelines is not driven only by timing effects. The
project’s cash-flow model incorporates several operational and financial parameters that
together increase the economic value of accelerated deployment, resulting in a higher NPV than
would be obtained through time-shifting alone. Consequently, the NPV increase from EUR 58
million to EUR 65 million reflects the combined impact of accelerated commissioning and
improved operational performance, which contribute to the economic benefit of permitting
simplification.
Direct economic benefit = ΔNPV = 4 −
Where each NPV is computed over a 10-year horizon and includes all cash flows: CAPEX
outflows during construction, OPEX, Revenues. The analysis has been carried out in nominal
terms, with both costs and revenues indexed over time and discounted using a nominal WACC.
NPV baseline (EUR m) 58.32 Economic benefits per project (EUR m)
Timeline reduction of 6 months
NPV after PM implementation (EUR m) 64.74 6.42 11%
Beyond the quantified impact, the measure is also expected to generate additional direct
benefits such as improved information for operators and authorities, reduced administrative
friction, and greater market efficiency through more consistent and predictable interactions
with permitting bodies. These impacts can be considered administrative cost savings, improved
information and better market efficiency, under the Better Regulation. They are not monetised
because of their smaller scale compared to the efficiency gains captured through the NPV
differences.
Total direct economic benefits for data centre operators under PM4 are thus estimated at EUR
4.6 bn (NPV, 10-years)
Adjustment and enforcement costs for national public authorities. National authorities are
expected to face one-off adjustment costs to comply with this measure and establish the project
facilitator. These costs arise to design and implement streamlined permitting procedures,
designate and staff a task force and align responsibilities across relevant authorities. For
modelling, a one-off cost of around EUR 0.4 million per Member State has been assumed to
capture the effort to (i) set up the project facilitator, i.e. defining the mandate, governance,
allocation of responsibilities and decisions (1 FTE), (ii) staff new people, i.e. initial team
preparation and training (2 FTEs), (iii) design streamlined procedures and guidelines (1 FTE)
and (iv) coordinate effectively among different authorities, i.e. meet with relevant ministries,
agencies, authorities and other bodies, as well as secretariat support (2 FTEs). This cost
corresponds to a total effort of 6 FTEs to enable the measure’s organisational design, workflow
streamlining, initial staffing and coordination with other authorities. To reflect differences in
territorial complexity and the number of administrative regions involved, this amount has been
46 The same applies to other measures using this modelling approach, i.e. PM5 and PM10.
69
uplifted with a mark-up for each NUTS administrative tier (NUTS 1/2/3), i.e. + EUR 0.052m
per NUTS-1 region, + EUR 0.01m per NUTS-2 region, + EUR 0.005 m per NUTS-3 region47.
The measure would also entail recurrent enforcement costs for national authorities, reflecting
the operations of the project facilitator and task force once established. These include staff time
for coordinating with competent authorities, handling applications, maintaining harmonised
workflows and acting as a central interface for data centre projects. These costs are mainly
staff-time-based, assuming 4 FTEs/year. The total EU-wide costs are obtained by multiplying
Member State–level savings by the total number of Member States.
Total costs (one-off adjustment and recurrent enforcement) for national public authorities under
PM4 are thus estimated at EUR 78.9 m (NPV, 10-years).
It is important to underline that this estimate represents a worst case scenario in which
authorities would need to establish a dedicated team from scratch, whereas in practice Member
States may rely on existing resources under the Gigabit Infrastructure Act, including by
designating a single information point established under Regulation (EU) 2024/1309, with the
relevant functions, procedures and mechanisms applying accordingly.
Administrative cost savings for national public authorities. Recurrent savings for national
public authorities should stem from administrative simplifications given the assumed reduction
of parallel processing and fewer back-and-forth interactions between authorities and with
economic operators. In the baseline, permit applications are processed through multiple parallel
channels, with each authority devoting staff time to separate contacts, repeated information
requests and uncoordinated reviews. Under the new set-up, a dedicated facilitator team would
coordinate input, standardise documentation and organise joint meetings. As a result, and based
on the interviews conducted, the total number of staff days required per project across all
authorities is assumed to decrease, even though some of this effort is reallocated. The figures
used consider that average processing of permitting procedures currently requires around 80
staff days (or 20 weeks) per project by each authority. Under the measure, process redesign and
single-window coordination could reduce this by approximately 20 staff days per new data
centre project (15-25 for sensitivity).
Cost savingsPA = ΔT × labour cost × s × No. new projects × No. of MS
Total administrative cost savings for national public authorities under PM4 are thus estimated
at EUR 110.8 m (NPV, 10-years)
Indirect impact: This measure is also expected to produce wider economic benefits and non-
monetary benefits, such as reduced investment uncertainty, more predictable project planning,
lower likelihood of disputes or appeals, enhanced coordination across authorities, and a more
attractive investment environment for large-scale digital infrastructure. While these indirect
effects are not quantified, they reinforce the overall positive impact of the measure and help
explain the broader efficiency gains supported by the quantified analysis.
Sensitivity analysis: Four key parameters are subject to sensitivity checks because of the
complexity of this policy measure and the number of assumptions implied most affecting the
balance of costs and benefits. The parameters subject to sensitivity are: (i) the time saved in
47 NUTS counts proxy multi-tier coordination and interface complexity (more NUTS-levels/regions implies small regions. The
EU currently lists 92 (NUTS-1), 244 (NUTS-2), 1,165 (NUTS-3) regions, underscoring the heterogeneity across Member
States.
70
permitting processes (ΔT) as this determines the scale of the economic benefits for operators
through time-to-market acceleration (varied between 4 and 8 months), (ii) the number of data
centre projects that would benefit by 2036, (iii) the implementation costs for authorities and
(iv) time savings for permitting by public administrations. These dimensions form the low and
high scenarios to test potential outcomes of this measure.
Results: The aggregated costs and benefits for all stakeholders over the time horizon are
summarised in the table below. While operators do not incur additional direct and experience
substantial economic benefits from time-to-market acceleration, national public authorities face
one-off adjustment and recurrent enforcement costs, with administrative cost savings from
simplification. Even though authorities’ cost savings slightly outweigh their costs in all
scenarios, the large economic benefits accruing to operators dominate the overall balance. The
differences between scenarios are mainly driven by changes in the magnitude of the NPV gains
linked to the expected reduction in permitting timelines.
Table 24. Administrative simplification for the roll-out of data centres (PM4)
Cost types Data centre
operators (€)
Public
authorities (€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - - - - -
one-off adjustment 20 668 601 - 20 668 601 19 011 860 22 325 342
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - 58 225 227 - 58 225 227 29 112 613 87 337 840
Total costs - 78 893 828 - 78 893 828 48 124 473 109 663 182
Total benefits (4 585 666 446) (110 843 709) - (4 696 510 155) (1 908 005 277) (8 742 605 033)
Net impact (4 585 666 446) (31 949 881) - (4 617 616 327) (1 859 880 804) (8 632 941 851)
3.5. PM5: Mechanism for Member States to identify areas to fast-track data
centre deployment
Under this measure, on the basis of a national data centre strategy/plan, Member States would
need to identify areas to fast-track data centre deployment based on shared EU-level criteria
(e.g. on connectivity, energy, water), allowing for the pre-selection of strategic sites where
permitting procedures will be streamlined. Fast-tracking measures would apply to projects that
have a significant social interest. This would be supported by EU level support through a
coordination hub, which will provide guidance, model templates and best practices. The
toolbox established in the new Regulation on speeding-up environmental assessments would
be leveraged for the designated areas, to allow them to benefit from the additional favourable
provisions in environmental assessments (overriding public interest, tacit approval, and dispute
settlement). Within fast-track areas, maximum timelines for data centre permitting beyond
environmental impact assessment should be of 18 months from the date on which a complete
application is acknowledged as duly submitted to the point at which construction activities for
the new facility begin. This implies that Member States would be required to adopt procedural
measures for accelerated data centre deployment.
Impact on data centre deployment acceleration. As above, it is assumed that the data centre
growth rate under this option increases to 15%. This uplift reflects the expectation that faster
and more predictable processes with lower entry barriers, reduce time-to-market and improve
71
investor certainty, thereby stimulating additional deployment compared to the baseline. A
geographic expansion beyond primary and secondary markets is supposed to emerge under this
measure, with new regional hubs emerging in Southern Europe, the Baltics, and Central &
Eastern Europe. The measure would primarily affect data centre operators, who account for
most projects, and cloud service providers (CSPs) that construct and operate their own facilities
(around 40% of projects). See Section 2.3.3 for further details. It is assumed that the share of
new projects benefitting from this measure increases from 5% in 2027 to 50% by 2036.
Administrative and adjustment costs for data centre operators. If operators choose to build in
these areas to benefit from accelerated permitting, they must comply with the technical,
environmental and procedural conditions attached to the areas. This gives rise both to
administrative costs (i.e. information obligations needed to demonstrate compliance) as well as
to adjustment costs (i.e. changes in project design or planning). Even if they are voluntary costs
as operators opt into the scheme, they are considered direct compliance costs once participation
is decided. Administrative costs arise from the additional effort required to access the area, i.e.
preparing a specific application file, familiarising themselves with criteria and national
implementing guidance, including connectivity, energy and water benchmarks, completing
standardised templates/checklists and submitting evidence of compliance with area conditions.
These costs are assessed following the Standard Cost Model, focusing on the time required for
operators to complete these tasks to access fast-track areas. This is estimated as an effort of 20
staff days or 4 weeks (sensitivity range 25-15) per application. This is multiplied by the share
of the data centre operators applying for the areas across the EU, since such costs are only
incurred by operators that apply to use fast-track areas. This is assumed to be 50% of the
companies identified as building data centres in the EU.
One-off administrative costs year t = ΔT × labour cost × No. of applications
Operators are also expected to face one-off adjustment costs if their project design/planning
shall be adapted to meet the area requirements, e.g., to comply with environmental conditions
applicable within the area. For simplicity, adapting workflows/operations to comply with area-
specific requirements and accelerated permitting timelines has been estimated to cost each
operator additional 40 staff days (30-60 days) to modify the project48. After this initial set-up,
recurring costs per project to demonstrate area eligibility are expected not to be incremental
with respect to the baseline and are thus not represented.
One-off adjustment costst = ΔT × labour cost × s × No. new facilities
Total costs (one-off adjustment and one-off administrative) for data centre operators under PM5
are thus estimated at EUR 13.8 m (NPV, 10-years)
Direct economic benefits for data centre operators. This measure aims to provide additional
benefits by reducing uncertainty over where investments should be directed. By creating a
mechanism for Member States to identify fast-track areas, operators avoid wasted effort on
non-viable sites. These changes are expected to reduce the number of iterations required and
shorten internal and external time commitments. Once initiated, each new deployment in a
designated area requires fewer efforts as eligibility is predefined and baseline data on
connectivity, grid, water, land use are available. The areas are also assumed to shorten the
48 CAPEX investments are not considered as part of the adjustment costs as only projects that already comply, or are willing
to comply, with the EED and other relevant requirements under the baseline scenario would be eligible to apply for the fast-
track areas. As a result, no additional CAPEX beyond what is already required under existing legislation is expected to arise
specifically from participation in the areas.
72
average development cycle for a new data centre facility by an additional 8 months (sensitivity
6-10), thus leading to a total duration from end of procurement to operation of 18 months,
assuming the full effect of the above measure (PM4) is also factored in. The improvement in
timeline coming from the implementation of the areas identified for simplification fast-track
deployment, has been estimated in 8 months average, due to:
• Coordinated spatial mapping, identifying brownfield sites, industrial areas near power
infrastructure, and locations with minimal environmental sensitivity (2 months savings).
• Area-level Environmental Impact Assessment (EIA) completed in advance for entire areas
(1 month carried out in parallel to the zoning).
• Standardized building codes, infrastructure requirements, and technical specifications (3
months average for building permit).
• Dedicated infrastructure areas, i.e., pre-planned electrical capacity (12 months for grid
connection).
As above, to capture these direct economic benefits, the analysis applies a NPV approach to
compare the baseline permitting duration with the accelerated scenario and measure the
economic value of bringing construction and commercial operations forward, discounted at the
project WACC. The NPV gain is also associated with the reduction in PUE foreseen under this
measure, as only the most sustainable data centres would be able to benefit from the fast-
tracking. In fact, expected improvements in energy efficiency, modelled as a declining PUE,
also contribute to reducing operating costs over time. The NPV increase from EUR 58 million
to EUR 74 million reflects the combined impact of accelerated commissioning and improved
operational performance, which contribute to the economic benefit of permitting simplification.
The business benefit is measured as the difference in net present value (NPV) between an
accelerated-permitting scenario and a baseline scenario, discounted at project WACC (6%):
Direct economic benefit t = (ΔNPV = 5 − )× s × No. new facilities
Where each NPV is computed over a 10-year horizon and includes all cash flows: CAPEX
outflows during construction, OPEX, Revenues. The analysis has been carried out in nominal
terms, with both costs and revenues indexed over time and discounted using a nominal WACC.
NPV baseline (EUR m) 58.32 Economic benefit per project (EUR m)
timeline reduction of 8 months
NPV after PM implementation (EUR m) 74.89 16.58 28%
As mentioned above under PM4, a small share of this economic benefit (around 1%) is expected
to direct administrative cost savings, i.e. avoided work by staff and consultants (around 1.5
FTE-years) that would otherwise be needed for document preparation, follow-ups with
authorities, and procedural back-and-forth foreseen before building a new data centre project.
Under this policy option, the project’s internal rate of return would consequently increase by
over 80 basis points with respect to the baseline. Taken together, PM4 and PM5 would increase
the project’s IRR from 9.86% to 10.70%, i.e. an economically meaningful shift which is
expected to improve investor appetite. Interviews with investors confirmed that data centre
investments are evaluated across a wide risk-return spectrum, broadly consistent with market
benchmarks positioning core infrastructure targeting around 7-9% IRR and core-plus assets
around 10-13%. Expected returns depend strongly on asset maturity and risk profile: stabilised,
fully built platforms with secured power and long-term contracts are treated as lower-risk
assets, while development-stage projects face higher execution risk related to permitting, power
availability, equipment procurement, and commercialisation. Equity investors typically target
73
“single digit returns plus a risk premium,” with materially higher expectations for projects
exposed to development, energy, or commercial risk. Together, these perspectives illustrate
why moving a project from 9.9% to 10.7% IRR can reposition assets within investor target
bands, improving competitiveness versus alternative infrastructure investments and affecting
bankability. This is particularly relevant for mid-sized European data centres (typically 5–25
MW), which are aligned with institutional investment ticket sizes but face tighter margins and
higher relative development risk than hyperscaler projects.
Total direct economic benefits for data centre operators under PM5 are thus estimated at EUR
11.8 bn (NPV, 10-years)
Adjustment and administrative costs for national public authorities. Member States are
required to draft strategies for national data centre deployment, where these do not already exist
and, as part of these, designate suitable areas for fast-track data centre permitting. This activity
is expected to generate mandatory one-off administrative costs for public authorities. Since at
least 7 MS already have a national data centre strategy or an analogous document ready or in
the making (see footnote 206), 20 Member States are expected to incur additional
administrative costs to draft new strategies. These have been estimated as 3 FTEs (1-5 for
sensitivity) per Member State. Moreover, each Member State would need to reprioritise
resources to set up processes and align national practices with these new rules. This has been
estimated as a one-off adjustment cost equivalent to 6 FTEs (4-8 for sensitivity) per Member
State.
Member States are also expected to face ongoing administrative costs to update the strategies
and the list of suitable fast-track areas for data centre deployment, estimated respectively as 3
FTEs every 5 years and 4 FTEs every year.
Total one-off adjustment, one-off administrative and recurrent administrative costs for national
public authorities under PM5 are thus estimated at EUR 79.9 m (NPV, 10-years).
Administrative cost savings for national public authorities. Recurrent savings for national
public authorities should also stem from administrative simplifications given the assumed
reduction back and forth interactions between authorities and with economic operators, thanks
to the clearer requirements from the areas. As above, given that permitting requires around 80
staff days per project by each authority, process redesign and clearer requirements are expected
to reduce this to 20 staff days per new data centre project. The analysis thus foresees that the
project facilitator and the fast-track approach might reduce overall administrative staff days per
project compared with the baseline.
Administrative cost savingsPA = ΔT × labour cost × s × No. new projects × No. of MS
Total administrative costs savings for national public authorities under PM5 are thus estimated
at EUR 110.8 m (NPV, 10-years).
Adjustment costs for the European Commission. The Commission would face one-off
adjustment costs to set up the coordination hub. This has been estimated at 3 FTE during the
first year. Moreover, recurrent annual adjustment costs are also estimated for the operation of
the coordination hub that supports Member States in identifying and designing fast-track areas.
This is estimated at 3 staff days per month each year (0.11 FTE/year) covering Commission
staff time.
74
Total one-off and recurrent adjustment costs for the European Commission under PM5 are thus
estimated at EUR 0.4 m (NPV, 10-years)
Indirect savings: Similarly to the measure on administrative simplification, this measure has
the potential to realise a broad range of indirect positive impacts on the investments climate,
better compliance and transparency of operators during the permitting process, as well as an
overall improvement of the decision making process.
Sensitivity analysis: Sensitivity analysis focuses on the potential effectiveness of the measure
and the ensuring associated regulatory costs, with parameters modelled to consider a changing
level of effort to comply with the data centres areas requirements, both in terms of
administrative and adjustment costs, as well as to understand how time savings influencing
NPV calculations and adoption rates impact potential direct economic benefits. With respect to
benefits, the reduction in months saved is varied between 6 and 10 (central 8 months saving)
as well as the share of facilities using the areas by 2036, i.e. considering between 30% and 70%
(with a central scenario of 50%). Concerning the costs, the additional time considered for
administrative and adjustment costs is also varied to reflect the uncertainty about the intensity
of the new procedures. The low scenario (min) combines conservative assumptions on
effectiveness (6 months saved, 30% uptake by 2036) with higher costs, while the high scenario
(max) does the opposite. Sensitivities were also considered for national public authorities, to
reflect the uncertainty related to the complexity of setting up the strategies and administering
the new processes, considering respectively minimum 1 and 4 and maximum 5 and 8 FTEs.
Results: The aggregated costs and benefits for all stakeholders over the time horizon are
summarised in the table below. Operators incur limited administrative and adjustment costs,
while they benefit from substantial NPV gains through shorter times to market and reduced
uncertainty, and public authorities face moderate administrative costs. The scenarios are largely
driven by the NPV calculation which is affected by the number of months and PUE savings
that are expected to stem from this new process. This would be expected for simplification
measures, whereby the benefits associated with earlier commissioning and reduced delays are
economically significant while the costs are limited to a small number of administrative and
compliance activities. As participation to use the areas is voluntary, no major regulatory
burdens are imposed.
Table 25. Establishing mechanism for Member States to identify areas to fast-track data centre
deployment (PM5)
Cost types Data centre
operators (€)
Public
authorities
(€)
European
Commissi
on (€)
Total Value (central,
€)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 1 476 159 3 681 647 - 5 157 806 2 057 555 8 903 876
recurrent administrative - 66 321 846 - 66 321 846 46 367 793 86 275 900
one-off adjustment 12 364 761 9 940 446 245 443 22 550 649 12 436 549 39 465 368
recurrent adjustment - - 117 627 117 627 117 627 117 627
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 13 840 920 79 943 939 363 070 94 147 928 60 979 524 134 762 770
Total benefits (11 844 675 172) (110 843 709) - (11 955 518 881) (6 219 100 693) (19 009 669 476)
Net impact (11 830 834 252) (30 899 770) 363 070 (11 861 370 953) (6 158 121 169)(18 874 906 705)
75
3.6. PM6: National funding support for data centres
The introduction of public support measures that target clearly defined market failures in line
with applicable State aid rules, to support data centre deployment would generate cost and
benefits mainly for operators and national authorities. Under this measure, Member States may
grant public support to particularly sustainable and strategic data centres deployed within pre-
identified areas (see PM5), according to commonly defined criteria.
Impact on data centre deployment acceleration. The baseline capacity growth presents a 12%
CAGR in data centre capacity (2025-2036). Under this option, it is assumed that the capacity
growth rate increases to 15%. This uplift reflects the assumption that support measures -
together with the other PM described above implemented as part of this option – would reduce
barriers to investment, accelerate time to market, and improve certainty for operators, thereby
stimulating additional capacity deployment. The scope of this measure covers data centre
operators and CSPs active in new data centre builds in the EU, while impacts scale with the
share of new projects receiving funding. Please see Section 2.3.3 for further details.
Administrative costs for Data Centre Operators. For operators, the one-off administrative
costs of participating in such public support measures are modest. Preparing applications,
providing supporting documentation and fulfilling reporting obligations are estimated at 0.2
FTE, or 45 days, during the first year. These are estimated as the staff time spent on the
application and, where required, verification processes multiplied by the relevant labour costs.
One-off administrative costs year t = ΔT × labour cost × No. of applications
Operators may face incremental costs to meet eligibility requirements. For example, the
inclusion of battery storage, waste-heat reuse, or other technologies to improve grid stability
and sustainability could raise investment costs. However, these would depend on the nature of
the support measure and thus cannot be reasonably quantified.
Total one-off administrative costs for data centre operators under PM6 are thus estimated at
EUR 0.7 m (NPV, 10-years).
Benefits for data centre operators. Operators would benefit from direct financial support, for
example through tax advantages or incentives, which is expected to reduce their capital
expenditure and accelerate or make possible the implementation of data centre projects. These
subsidy amounts would represent direct benefits for operators and corresponding costs for
public authorities. Nevertheless, given the voluntary nature of the funding scheme (as the
Member State must first decide that it wants to implement such a scheme) they have not been
quantified as fiscal transfers.
Adjustment costs for national public authorities. For national authorities, in addition to the
overall fiscal transfer costs, the set-up and management of the scheme generates new
adjustment costs. These are quantified based on the expected effort required by each Member
State, assuming that for the design and legal establishment of the scheme during the first year,
4 FTEs (3-5 for sensitivity ranges) are required and entail one-off adjustment costs. These costs
are scaled by the number of Member States expected to operate the scheme under low, central
and high uptake scenarios (25%, 50% and 75% of MS). Thereafter, processing, evaluation of
applications, and monitoring is estimated at 3 FTEs (2-4 for sensitivity ranges) every two years
as recurrent adjustment costs for the duration of the grant (assumed to be of 7-years).
Total adjustment costs = (one-off adjustment costs + recurrent adjustment costs) × s × MS =
[(ΔT × labour cost) year 1 + (ΔT × labour cost) year 1,3,5,7] × s × MS
76
Total one-off and recurrent adjustment costs for national public authorities under PM6 are thus
estimated at EUR 10.1 m (NPV, 10-years)
Cost savings for national public authorities. Administrative savings may be realised where
harmonised EU criteria and templates reduce duplication of effort, e.g. estimated at around 1
FTE annually compared with purely national schemes. However, they are deemed negligeable
and are not included in the calculations.
Costs for the European Commission. At EU level, modest administrative costs would arise for
the oversight and monitoring of Member State schemes. These costs are not accounted for in
the calculations.
Indirect benefits: Subsidies for particularly sustainable and strategic data centres may generate
public sector benefits through higher tax revenues with positive spillover effects on the
advancement of digital objectives. Tax reductions for data centres can also indirectly generate
benefits, e.g. in terms of increased capital inflows and improved digital capacity that supports
broader economic competitiveness. Such measures would need to be carefully designed to
account for their opportunity costs and ensure that the expected benefits justify the reduced
fiscal intake. The support measures should also aim to foster energy-efficient, environmentally
responsible data centre facilities, contributing to improve digital resilience and service quality
for citizens and businesses. These elements have not been quantified in this section.
Sensitivity analysis: Sensitivity analysis was considered for the total administrative costs for
the number of public authorities administering the scheme and the effort required to set-it up
and manage it in terms of FTEs.
Results: The table below presents the aggregated costs for all stakeholders for the
implementation of this measure. Total costs are estimated at EUR 10.8 m. The minimum and
maximum ranges are driven by variations in the number of MS wishing to operate the scheme
by and the magnitude of staff days involved. Fiscal transfers have no net effect when aggregated
across stakeholders, i.e. financial benefits for operators are mirrored by equivalent fiscal costs
for public authorities. Indirect opportunity costs for operators participating in the scheme are
also illustrated in the table below and vary based on the uncertainty with respect to the number
of applications and days dedicated to proposal/application preparation.
Table 26. National funding support measures for data centres (PM6)
Cost types Data centre
operators (€)
Public
authorities (€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 654 225 - - 654 225 169 614 1 332 680
recurrent administrative - - - - - -
one-off adjustment - 3 313 482 - 3 313 482 1 242 556 6 212 779
recurrent adjustment - 6 830 638 - 6 830 638 2 276 879 13 661 275
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 645 225 10 144 120 - 10 798 344 3 689 049 21 206 734
Total benefits - - - - - -
Net impact 645 225 10 144 120 - 10 798 344 3 689 049 21 206 734
77
3.7. PM7: Set deployment targets and monitor progress
The measure sets EU-level deployment targets for data centre capacity (e.g., MW IT load)
which should be regularly adjusted in light of a monitoring of the compute-capacity gap
identified for the EU. Such monitoring would be carried by the Commission and verified by
relevant national authorities. The measure does not create mandatory obligations for private
operators. The main costs would fall on the Commission, which would be responsible for
developing the methodology and compiling the initial dataset. In so far as possible, the
methodology would build on existing publicly available data and the data collected as part of
the energy efficiency monitoring of the Energy Efficiency directive, and operators are expected
to incur limited costs associated with providing supplementary information, while national
authorities would face small recurrent costs to verify and validate the data submitted.
Impact on data centre deployment acceleration. As above, under this option, it is assumed that
the capacity growth rate increases to 15%.
Administrative costs for Data Centre Operators. Operators are expected to face minor
recurrent administrative costs to provide supplementary information need for the dataset. These
tasks involve supplying data that they already hold for operational or reporting purposes,
notably covering capacity in MW or FLOP49, which would keep the burden at a minimum.
These costs have been modelled as staff time spent participating in surveys and possibly
validating data quality. This is expected to require approximately 1 day per year (0.5-1.5 for
sensitivity ranges). Annual updates are assumed to entail the same cost.
Total recurrent administrative costs (voluntary)= (ΔT × labour cost × No. of operators)
Total administrative costs for data centre operators under PM7 are thus estimated at EUR 0.4
m (NPV, 10-years).
Costs for national public authorities. National authorities would incur limited recurrent
enforcement costs. Their main effort would entail performing periodic checks to verify that the
compiled data is complete and reflects national conditions. In this context, the monitoring effort
should take into consideration the implementation of national data centre strategies and track
progress against their objectives so that any relevant results are also shared with the
Commission. This effort is expected to require approximately 15 staff days per year (10-20 for
sensitivity ranges), resulting in small and predictable administrative burden.
Total recurrent enforcement costs = ΔT × labour cost × No. of MS
Total enforcement costs for national public authorities under PM7 are thus estimated at EUR
0.9 m (NPV, 10-years).
Costs for the European Commission. The Commission would be responsible for carrying the
study and publishing the results at EU level. This would result in one-off adjustment costs to
set up the process and guidance effort, estimated at 1.5 FTEs (1-2 for sensitivity ranges) in year
1, including the procurement of a study to survey capacity periodically over 10-years, with an
estimated budget of EUR 800 000. Recurring enforcement costs for monitoring and
coordinating with Member States’ have then been estimated for at 25 days per year or 0.1 FTEs.
49 Floating point Operations Per Second (FLOPS) is a standard measure of the performance of a computer, indicating how
many calculations on numbers with decimal points it can perform in one second.
78
Total one-off adjustment and recurrent enforcement costs for the European Commission under
PM7 are thus estimated at EUR 1.0 m (NPV, 10-years).
Indirect impacts. The measure generates relevant indirect benefits. For operators, clearer EU-
wide targets are expected to reduce investment uncertainty and effort in oversupplied locations.
It is expected to create better conditions for operators to understand regional demand and supply
gaps and get access to EU-wide benchmarking on capacity utilisation and build-out. In addition,
one harmonised EU reporting would reduce duplication of data reporting at national level.
National authorities would benefit from improved infrastructure foresight and visibility of
national data centre capacity, with better planning for grid reinforcements and targeting of
support schemes. The Commission would ultimately also benefit through reduced duplication
of studies, easier cross-DG coordination, and enhanced credibility when engaging with
international partners. It would benefit from a cross-border, EU-wide analysis of capacity
availability and needs to improve evidence-based policy intervention. Wider economic benefits
include improved sector planning, enhanced competitiveness and potentially better
environmental outcomes.
Sensitivity analysis: Sensitivity analysis has been performed to understand the possible
direction of costs, i.e. FTE resources needed by operators, authorities and the Commission to
participate in this monitoring exercise. Min and max scenarios vary the central assumptions
presented above by approximately ±33%, resulting in conservative, central and optimistic
estimates of total costs.
Results: For this measure, total estimated costs for the implementation are modest at EUR 2
266 361. These costs reflect the recurrent adjustment and recurrent administrative costs mainly
borne by the Commission staff and national authorities for monitoring activities.
Table 27. Set deployment targets and monitor progress (PM7)
Cost types Data centre
operators (€)
Public
authorities (€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative 392 445 - - 392 445 196 222 588 667
one-off adjustment - - 899 421 899 421 858 513 940 328
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - 902 108 72 388 974 496 659 316 1 289 676
Total costs 392 445 902 108 971 809 2 266 361 1 714 051 2 818 671
Total benefits - - - - - -
Net impact 392 445 902 108 971 809 2 266 361 1 714 051 2 818 671
3.8. PM8: EU Funding for R&D and innovation ecosystems for cloud and AI
Under this measure, the Commission would identify key R&D funding challenges for the
development of energy and resource efficient and secure data centre technologies, including
advanced cooling, renewable energy and storage integration and AI-based optimisation tools.
Funding schemes would be used accordingly to develop R&D and innovation ecosystems
focused on cloud and AI data centres and innovative software enabling cloud and AI computing
services across the European cloud-to-edge continuum. The measure would aim to de-risk and
accelerate the uptake of advanced, sustainable and secure technologies, with a focus on data
centre technologies to improve the replicability and scalability of innovative solutions across
79
Member States, especially for smaller providers. The measure assumes that successors to
today’s Horizon Europe exists in the next multiannual financial framework and is provided for
modelling purposes only, with no pre-emption of future budgetary decisions. The time horizon
considered is seven years, with Work Programmes designed every two years. The scope covers
activities at technology readiness levels four to eight, ranging from laboratory-scale research
through piloting and demonstration, and includes sustainable cooling and energy solutions,
reuse of waste heat and AI-enabled operations.
Impact on data centre deployment acceleration. The baseline capacity growth presents a 12%
CAGR in data centre capacity (2025-2036). Under this option, it is assumed that the capacity
growth rate increases to 13%. The scope of this measure covers CSPs and operators active in
new data centre builds in the EU, while impacts scale with the share of new projects receiving
funding. See Section 2.3 for further details. It is assumed that the share of new projects
benefitting from PO1-C increases from 10% in 2026 to 50% by 2036 (see also sensitivity
section).
Administrative costs for data centre operators. Shall they wish to apply for this type of funding
businesses would face administrative one-off costs include proposal preparation and
consortium coordination. These costs (e.g. for preparing applications, providing supporting
documentation and fulfilling reporting obligations) have been estimated at 0.2 FTEs, or 45 days
during the first year for consortium building and design of governance structures. It has been
assumed that 20 proposals would be prepared every 2 years. These are estimated as the staff
time spent on the application multiplied by the relevant labour costs.
One- off administrative costs year t = ΔT × labour cost × No. of applications/proposals
Total one-off administrative costs for data centre operators under PM8 are thus estimated at
EUR 1.1 m (NPV, 10-years).
Operators may also face incremental costs to meet eligibility requirements.
Benefits for Data Centre Operators. Operators would benefit from direct financial support,
which is expected to reduce their return on investments for cloud and AI projects. These subsidy
amounts are considered benefits for operators and corresponding costs for the Commission for
transparency, but they do not contribute to the direct impact assessment as they constitute fiscal
transfers. Savings are calculated as the amount of reimbursable financing provided to the
eligible final recipients. Currently, Horizon Europe Research and Innovation Action (RIA)
applies a 100% funding rate while the Digital Europe Programme funding rate is 50% for
simple action grants and 75% for the SME Support Actions.
Adjustment costs for the European Commission. The Commission bears responsibility for the
management of calls, development of templates and reporting tools, and the evaluation and
synthesis of projects. The main cost would be one-off programme design and administration,
i.e. work programme preparation, call management, evaluation, grant management preparation,
monitoring and results dissemination. This has been estimated as 3 FTEs during the first year
of the programme and recurring every two years for Work Programme design. Recurrent costs
then arise in relation to managing the funding scheme and have been estimated as 1 FTEs. The
assumed portfolio consists of one call every two years funding approximately 10 projects.
Total adjustment costs = (one-off adjustment costs + recurrent adjustment costs) = [(ΔT ×
labour cost) year 1,3,5 + (ΔT × labour cost) year 2,4,6,7]
80
Total adjustment costs for the European Commission under PM8 are thus estimated at EUR 1.1
m (NPV, 10-years).
There could be an impact for those MS that decide to contribute to the funding programme with
national funds in line with applicable State aid rules. As this would depend on the size of this
contribution, it has not been modelled under this policy measure.
Indirect savings: the main indirect impact, which is hard to quantify, is the value generated in
terms of making current product and services more efficient and competitive and of generating
new products and services. The EC estimated that every Euro invested in Horizon Europe
generates 11 EUR in economic value (GDP gains by 2045), i.e. “For every euro of costs to EU
society, the programme is expected to generate up to six euros in benefits for EU citizens by
2045. In terms of economic growth, every euro of EU contribution is estimated to generate up
to €11 in GDP gains by 2045, according to an evaluation of the Commission released today..."
(April 2025)50. The impact of Horizon Europe is directly reflected in the success of funding
various research and innovation projects, in January 2025 for example, more than 15,000
projects in fields such as electric vehicles, new antibiotics, and accessible AI technologies
benefitted from the programme51. A last element has also made Horizon Europe particularly
beneficial to European innovation and research savings, with lump sum grants being estimated
to reduce beneficiaries’ administrative costs from 14% up to 30% over projects’ lifetime52 with
the elimination of financial reporting requirements being of paramount importance in
stimulating industry participation of SMEs and newcomers.
Sensitivity analysis: Sensitivity analysis was considered for the total costs for the Commission
in administering the scheme, i.e. the effort required to set-it up and manage it in terms of FTEs,
also to consider the uncertainty related to the overall magnitude of the scheme.
Results: The table below presents the aggregated costs for all stakeholders for the
implementation of this measure. Total adjustment costs for the European Commission are
estimated at EUR 1.09 m. The minimum and maximum ranges are driven by variations in the
effort levels required to operate the scheme by and the magnitude of staff days involved. Fiscal
transfers have no net effect when aggregated across stakeholders, i.e. financial benefits for
operators are mirrored by equivalent fiscal costs for the Commission. Administrative costs for
operators participating in the scheme are also illustrated in the table below and vary based on
the uncertainty with respect to the number of applications and days dedicated to the preparation
of the proposal/application.
Table 28. EU Funding for R&D and innovation ecosystems for cloud and AI (PM8)
Cost types Data centre
operators (€)
Public
authorities (€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 1 090 375 - - 1 090 375 678 455 1 599 216
recurrent administrative - - - - - -
one-off adjustment - - 694 870 694 870 463 247 926 493
recurrent adjustment - - 440 093 440 093 293 395 586 790
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
50 See: IP_25_1115_EN.pdf 51 ibid. 52 ibid.
81
recurrent enforcement - - - - - -
Total costs 1 090 375 - 1 134 963 2 225 337 1 435 097 3 112 500
Total benefits - - - - - -
Net impact 1 090 375 - 1 134 963 2 225 337 1 435 097 3 112 500
3.9. PM9: EU deployment funding for strategic projects
Under this policy measure, without prejudice to the outcome of the negotiations on the next
MFF proposal, the Commission would provide financial incentives specifically for data centre
projects of European strategic interest and importance according to EU-level eligibility criteria.
The measure assumes is provided for modelling purposes only, with no pre-emption of future
budgetary decisions. Under this policy option, the EU would act as a market-enabler through
financial incentives that de-risk infrastructure investment. By absorbing early capital
expenditures (CAPEX) risk in data centre deployment, these investments would incentivise
private investments where private return of investment is uncertain. This measure is designed
to accelerate the roll-out of secure, sustainable, and cross-border digital services across the
Union, while reinforcing European strategic autonomy.
Impact on data centre deployment acceleration. The baseline capacity growth presents a 11%
CAGR in data centre capacity (2025-2036). Under this option, it is assumed that the capacity
growth rate increases to 14% over the same period. This uplift reflects the assumption that
public funding lowers barriers to investment, reduces financing costs for enterprises and
accelerates the timing of large-scale deployment projects. Over time, CSPs and operators build
more facilities than in the baseline and benefit from public support on a rising share of those
builds. The scope of this measure covers CSPs and operators active in new data centre builds
in the EU, while impacts scale with the share of new projects receiving funding.
Administrative costs for Data Centre Operators. Even though participating in the funding
scheme would be voluntary, it would create administrative costs through financing obligations
and administrative compliance requirements. This would include proposal preparation and
consortium coordination. Operators would need to prepare proposals and adapt processes to
meet eligibility requirements. These would translate into one-off administrative costs incurred
every two years in line with the funding cycle. It has been assumed that operators would prepare
around 15 proposals per cycle, each requiring staff days (approximately 45 days) of work on
average53. The average size of consortia has been considered of 2-4 partners per project.
One-off administrative costs year t = ΔT × labour cost × No. of applications/proposals
Total administrative costs for data centre operators under PM9 are thus estimated at EUR 0.8
m (NPV, 10-years).
Operators may also face incremental costs to meet eligibility requirements.
Benefits for Data Centre Operators. As described above, operators would benefit from direct
financial support, which is expected to reduce their return on investments for data centre
projects. These subsidy amounts are acknowledged as direct benefits for operators and
corresponding costs for the Commission, but they do not contribute to the direct impact
assessment as they constitute fiscal transfers.
53 In this report, based on a survey, it was estimated that lump sum proposals required 50 days effort from ideation to submission
in H2020 (https://ec.europa.eu/info/funding-tenders/opportunities/docs/2021-2027/horizon/other/comm/ls-assessment-report-
2024_en.pdf).
82
Adjustment costs for the European Commission. The Commission bears responsibility for the
management of calls, development of templates and reporting tools, and the evaluation of
projects. The main adjustment cost would be one-off programme design and administration,
i.e. work programme preparation, call management, evaluation, grant management preparation,
monitoring and results dissemination. Aligned with the estimations above, this has been
estimated as 3 FTEs during the first year of the programme and recurring every two years for
Work Programme design. Recurrent costs then arise in relation to managing the funding scheme
and have been estimated as 1 FTEs. The assumed portfolio consists of one call every two years
funding approximately 5 projects.
Total adjustment costs = (one-off adjustment costs + recurrent adjustment costs) = [(ΔT ×
labour cost) year 1,3,5 + (ΔT × labour cost) year 2,4,6,7]
Total adjustment costs for the European Commission under PM9 are thus estimated at EUR 1.1
m (NPV, 10-years).
There could be an impact for those MS that decide to contribute to the funding programme with
national funds in line with applicable State aid rules. As this would depend on the size of this
contribution, it has not been modelled under this policy measure.
Indirect savings: the main indirect impact stems from the value generated in terms of making
current product and services more competitive and from generating new products and services.
Sensitivity analysis: Sensitivity analysis was considered for the total administrative costs for
the Commission in administering the scheme, i.e. the effort required to set-it up and manage it
in terms of FTEs, also to consider the uncertainty related to the overall magnitude of the
scheme.
Results: The table below presents the aggregated costs for all stakeholders for the
implementation of this measure. Total adjustment costs for the European Commission are
estimated at EUR 1.1 m. The minimum and maximum ranges are driven by variations in the
effort levels required to operate the scheme by and the magnitude of staff days involved. Fiscal
transfers have no net effect when aggregated across stakeholders, i.e. financial benefits for
operators are mirrored by equivalent fiscal costs for the Commission. Administrative costs
(EUR 0.79 m) for operators participating in the scheme are also illustrated in the table below
and vary based on the uncertainty with respect to the number of applications and days dedicated
to proposal/application preparation.
Table 29. EU deployment funding for strategic projects (PM9)
Cost types Data centre
operators (€)
Public
authorities (€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 792 583 - - 792 583 446 840 1 235 249
recurrent administrative - - - - - -
one-off adjustment - - 694 870 694 870 463 247 926 493
recurrent adjustment - - 440 093 440 093 293 395 586 790
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 792 583 - 1 134 963 1 927 546 1 203 482 2 748 533
Total benefits - - - - - -
83
Net impact 792 583 - 1 134 963 1 927 546 1 203 482 2 748 533
3.10. PM10: EU-level identification of areas for fast-track data centre deployment
Under this measure, the Commission would directly designate areas to fast-track permitting for
data centre deployment based on EU-level criteria (e.g. on connectivity, energy, water) and in
consultation with Member States experts, allowing for the pre-selection of strategic sites where
permitting procedures will be streamlined. Consultation with a newly created EU Data Centre
Acceleration Board consisting of Member State experts would be required for the identification
and greenlighting of such areas. Permit granting for data centre projects would follow EU-wide
rules and require approval from the Board. For these areas, Member States would be required
to enact a set of acceleration measures prescribed at EU level.
Impact on Datacentre data centre deployment acceleration. Under this option, it is assumed
that the capacity growth rate increases to 14%. This uplift reflects the expectation that faster
and more predictable processes with lower entry barriers, reduce time-to-market and improve
investor certainty, thereby stimulating additional deployment compared to the baseline trend.
It is assumed that the share of new projects benefitting from this policy measure increases from
5% in 2027 to 50% by 2036.
Administrative and adjustment costs for data centre operators. As above, if operators choose
to build in these areas to benefit from accelerated permitting, they must comply with the
attached conditions, which give rise to both administrative and adjustment costs. Costs are
assessed following the Standard Cost Model, focusing on the information obligations and
activities required for operators to access fast-track areas. Reading EU-level guidance, adapting
internal procedures and staff would entail one-off costs for each operator. To access areas,
operators must familiarise themselves with EU-level criteria. As above, this is estimated as an
effort of additional 20 days (sensitivity range 25-15) per operator, which is multiplied by the
share of the operators applying for the areas across the EU. This is assumed to be 50% (central
value; 30%-70% for sensitivity) of the companies identified as building data centres in the EU.
One-off adjustment costs year t = ΔT × labour cost × No. of applications
Operators are also expected to face one-off adjustment costs in adapting workflows and
templates to comply with area-specific requirements and accelerated permitting timelines.
Adapting workflows/operations to comply with requirements has been estimated to cost each
operator additional 40 staff days (30-60 days) to modify the project. This would be multiplied
by a share of the projects benefitting from the areas across the EU (50% by 2036 as the central
scenario; 30%-70% for sensitivity). As outlined above, CAPEX investments are not considered
as part of the adjustment costs as only projects that already comply, or are willing to comply,
with the EED and other relevant requirements under the baseline scenario would be eligible to
apply for the fast-track areas. As a result, no additional CAPEX beyond what is already required
under existing legislation is expected to arise specifically from participation in the areas.
One-off adjustment costs = ΔT × labour cost × s × No. new facilities
After this initial set-up, recurring costs per project to demonstrate eligibility are expected not
to be incremental with respect to the baseline and are thus not represented.
Total costs for data centre operators under PM10 are thus estimated at EUR 11.9 m (NPV, 10-
years).
84
Direct economic benefits for data centre operators. This measure aims to provide additional
benefits by reducing uncertainty over where investments can be facilitated, similarly to PM5
but at European scale. With fast-track areas, operators should be able to avoid much of the
wasted effort on non-viable sites. These changes are expected to reduce the number of iterations
required and shorten internal and external time commitments. These savings are modelled using
NPV, reflecting a time saving from end-procurement to deployment of 3 months in total (2-4
months for sensitivity ranges). This is assumed because, while this measure is wider than the
measure on the fast-track zoning at MS level, the feedback received during consultations with
industry highlighted that a top-level identification would be less effective as it would not be
able to account in the same way, of geographical and local specificities.
As above, to capture these direct economic benefits, the analysis applies a NPV approach to
compare the baseline with the accelerated scenario and measure the economic value of bringing
construction and commercial operations forward, discounted at the project WACC. The strong
NPV gain is also very sensitive to the strong reduction in PUE foreseen under this measure, as
data centres would also be able to benefit from dedicated R&D to optimised energy efficiency
and increase their sustainability. Expected improvements in energy efficiency, modelled as a
declining PUE, greatly contribute to reducing operating costs over time. The NPV increase
from EUR 58 million to EUR 72 million reflects the combined impact of accelerated
commissioning and improved operational and energy performance, which contribute to the
economic benefit of permitting simplification. The business benefit is measured as the
difference in net present value (NPV) between an accelerated-permitting scenario and a
baseline scenario, discounted at project WACC (6%):
Economic benefit t = (ΔNPV = 10 − )× s × No. new facilities
Where each NPV is computed over a 10-year horizon and includes all cash flows: CAPEX
outflows during construction, OPEX, Revenues. The analysis has been carried out in nominal
terms, with both costs and revenues indexed over time and discounted using a nominal WACC.
NPV baseline (EUR m) 58.32 Cost savings per project (EUR m)
Timeline reduction of 3 months
NPV after PM implementation
(EUR m)
72.08 13.76 24%
Total direct economic benefits for data centre operators under PM10 are thus estimated at EUR
8.1 bn (NPV, 10-years).
Adjustment and administrative costs for national public authorities. It is assumed that each
Member State would need to set up processes and align national practices with these new rules
to provide data to the Commission. This has been estimated as a one-off adjustment cost
equivalent to 6 FTEs (4-8 for sensitivity) per Member State as they would need to get
familiarised with and onboard the new platform and adopt templates to provide data to the
relevant EU entity in charge of identifying suitable areas to fast-track data centre deployment.
In addition to this, MS would face one-off administrative costs associated to having to appoint
delegates to the management board of the EU Data Centre Acceleration Board.
Once the mechanism is operational, authorities would incur recurrent administrative
administrative costs to map and upload data on suitable deployment areas onto the EU platform.
For this, 3 FTE annually has been assumed. Recurrent adjustment costs are also expected from
national authorities’ online participation in the EU Data Centre Acceleration Board. For this,
85
participation is estimated at 3 FTE each year per Member State, covering several meetings
annually and additional related work.
Total costs for national public authorities under PM10 are thus estimated at EUR 87.3 m (NPV,
10-years).
Administrative cost savings for national public authorities. By using a central EU-based
database, Member States avoid repeated ad hoc clarifications with operators on deployment
and siting rules for new data centres, thus reducing information obligations and simplifying
tasks. On average, savings are estimated at one-third of time saved in terms of staff days, i.e.
estimated as 1 FTE (0.5-2.0 for sensitivity ranges) each year, reducing authorities’ net
administrative burden. Savings for authorities would also results from a centralised EU-level
structure that reduces duplication in national-level coordination meetings.
Total administrative cost savings for national public authorities under PM10 are thus estimated
at EUR 12.9 m (NPV, 10-years).
Administrative and adjustment costs for the European Commission. The Commission would
face one-off adjustment costs to set up the mechanism and the data centre acceleration board.
This has been estimated at 8 FTE during the first year, plus costs for the procurement and
development of the supporting digital tool (EUR 2.0 m). Recurrent annual administrative costs
are also estimated for the operation of the Board and to map and update the list of suitable fast-
track areas for data centre deployment, also estimated at 6 FTEs per year. Adjustment costs in
terms of maintenance of the tool (EUR 0.8 m) and a budget for procuring periodic external
studies and/or expert meetings (EUR 2 m) are also considered to take place every two years on
average.
Total costs for the European Commission under PM10 are thus estimated at EUR 16.1 m (NPV,
10-years).
Sensitivity analysis: The parameters are very sensitive to the number of benefitting operators
and the assumed time reductions. As above, sensitivity analysis focuses on the potential
effectiveness of the measure and the associated regulatory costs, with parameters modelled to
consider a changing level of effort to comply with the areas requirements, both in terms of
administrative and adjustment costs, as well as to understand how time savings influencing
NPV calculations and adoption rates impact economic benefits. The reduction in months saved
varies between 2 and 4 (central 3 months saving) as well as the share of facilities using the
areas by 2036, i.e. considering between 30% and 70% (with a central scenario of 50%).
Concerning the costs, the additional time considered for administrative and adjustment costs is
also varied to reflect the uncertainty about the intensity of the new procedures. The low scenario
combines conservative assumptions on effectiveness (4 months saved, 30% uptake by 2036)
with higher costs, while the high scenario does the opposite. Sensitivities were also considered
for national public authorities, to reflect the uncertainty related the annual effort for
participating in the new processes. This has been varied considering between 15 days (high-
effort) and 5 days (low-effort) per year and expected administrative cost savings generated by
the mechanism, ranging from 0.5 to 1.5 FTEs saved per year at EU level under the low and
high scenarios. The measure is highly sensitive to assumptions on PUE improvements, which
significantly influence overall OPEX. In the absence of PUE reductions beyond the baseline
scenario, the NPV would fall to EUR 59.82 million, i.e. only around 3% of projected cost
savings. This highlights the critical role of PM8 and PM9 in driving more sustainable and
energy-efficient data-centre operations.
86
Results: The total costs and benefits for all stakeholders are summarised in the table below.
Total costs are estimated between EUR 105.5 m and EUR 131.1 m, including administrative
and adjustment costs for establishing and maintaining the EU-level mechanism. Total savings
are substantial, ranging from EUR 4.8 bn to EUR 12.2 bn, reflecting the expected efficiency
gains from faster permitting, coordinated siting, and reduced duplication across Member States.
The high-low scenarios are largely driven by the NPV calculation which is affected by the
number of months which is perceived to be saved in this process.
Table 30. EU-level identification of areas for fast-track data centre deployment (PM10)
Cost types Data centre
operators (€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 1 476 159 - - 1 476 159 830 339 2 767 798
recurrent administrative - 38 698 697 3 822 094 42 520 790 42 520 790 42 520 790
one-off adjustment 10 375 852 9 940 446 2 596 263 22 912 560 13 728 731 37 475 850
recurrent adjustment - 38 698 697 9 682 334 48 381 031 48 381 031 48 381 031
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 11 852 011 87 337 840 16 100 690115 290 541 105 460 892 131 145 470
Total benefits (8 082 059 633) (12 899 566) - (8 094 959 199) (4 769 247 820) (12 245 115 194)
Net impact (8 070 207 622) 74 438 274 16 100 690 (7 979 668 658) (4 663 786 928) (12 113 969 724)
3.11. PM11: Creating EU-level harmonized criteria for sovereign cloud and AI
computing services
Policy Measure 11 (PM11) establishes a harmonised Union-level framework for ‘sovereign’
cloud and AI computing services. AI systems are not concerned by the measure, as they are
already subject to the AI Act. Acknowledging that different use cases require varying degrees
of ‘sovereignty’, the framework provides for four levels of sovereignty assurance.
To be considered ‘sovereign level 1’, a service must meet the following cumulative criteria:
(i) the service provider must be established in the Union (meaning the EEA); and
(ii) the service must be fully operated from computing infrastructure, personnel and
assets located in the Union; and
(iii) customer data, including metadata and telemetry data, is in the EU unless the
customer explicitly requires otherwise; and
(iv) the service provider demonstrates that it complies with state-of-the-art
cybersecurity standards; and
(v) if technical and operational support is outsourced to third-party providers outside
of the Union, necessary measure are put in place to ensure that would not
compromise the provider’s operational autonomy; and
(vi) there is full transparencey around the use of subcontractors, for which the cloud
service provider assesses that they meet Union legal obligations; and
(vii) where the cloud service provider is subject to the control of a third country or a
third country entity, it must be able to prove that the laws and government practices
in that country do not require the provider to tell that country's authorities about
software vulnerabilities before those vulnerabilities have been publicly discovered
These requirements would allow service providers with a parent company headquartered
outside of the Union to be considered as ‘sovereign level 1’.
87
To be considered ‘sovereign level 2’, a service must meet the following cumulative criteria:
(i) the service provider and subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure, personnel and
assets located in the Union; and
(iii) provide available personnel complying with additional personnel screening and
Union citizenship requirements, if the customer determines that imposing these
additional requirements is necessary; and
(iv) the service provider must be controlled by a legal entity in the Union.
Alternatively, if the service is controlled by a third-country legal entity, it must
demonstrate that it has in place the necessary technical, legal and organisational
measures necessary to prevent third-country governmental access and transfer of
data stored in the Union, to prevent or refuse any request from a third-country
government, ensure that the control of the third-country or third-country entity is
not exercised in a manner that restricts the provider’s ability to deliver the service,
and to prevent the service disruption and/or degradation of the service by a third-
country government54; and
(v) the data generated by using the audited service shall not be re-used for the training
or fine-tuning of an AI system operated by an entity outside the EEA and in
any case are not transferred outside the Union; and
(vi) the customer data, including metadata and telemetry data, remain in the Union
unless the customer explicitly requests otherwise; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at
least at level ‘substantial’ under the European Cybersecurity Certification
Scheme for Cloud Services (EUCS)55; and
(viii) the service provider must demonstrate a high degree of control over the software
components that underpins the service. This notably implies that there exists a list
of identified dependencies related to the provision of the service, and where the
software components are provided by a third-country entity, the relevant code of
the security relevant components of the service stack can be audited, and there
exists a migration plan in the event a vendor fails or a third-country imposes
restrictions; and
(ix) if subcontractors are from a third country or a third country entity, appropriate
measures in place to demonstrate the absence of control; and
(x) operational and technical support, including outsourcing, are initiated and
performed exclusively within the Union; and
(xi) where the cloud service provider is subject to the control of a third country or a
third country entity, it must be able to prove that the laws and government practices
in that country do not require the provider to tell that country's authorities about
software vulnerabilities before those vulnerabilities have been publicly discovered
54 In the absence of a harmonised framework, non-EU service providers attempting to prevent third-country governmental
access and transfer of data stored in the Union and to prevent or refuse any request from a third-country government are using
a diverse technical, legal and organisational measures. This include technical architecture with segregated physical
infrastructure, ensuring that the encryption keys are not accessible to the provider or are held exclusively by the customer,
adding specific clauses in their EU employees’ contract that forbid them from taking instructions from outside of the EU,
setting up independent boards to review extra-territorial data access requests, etc. 55 As part of the ‘One Europe, one market’ roadmap agreed by the Parliament, the Council and the Commission, the co-
legislator have agreed to finalise negotiation for this initiative ed by Q4 2027. Adding one year for the measures to take effect,
this implies CADA entering into force in early 2029. EUCS technical work is finalized and has been adopted by CEN-
CENELEC Technical Specifications. The candidate scheme has therefore reached an advanced stage of development, which
now needs to be transformed into an Implementing Act adopted under the Cybersecurity Act, a process much shorter than
CADA’s interinstitutional negotiations.
88
These requirements would allow service providers controlled by a third-country or third-
country entity to be considered as ‘sovereign level 2’, but on the basis of some
organisational efforts. Service providers owned and controlled by a legal entity in the Union
would face less difficulties in complying with these criteria.
To be considered ‘sovereign level 3’, a service must meet the following cumulative criteria:
(i) the service provider subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure and assets
located in the Union; and
(iii) members of the board, executive team and personnel operating the service are
Union nationals, located in the Union, and are security cleared where
appropriate; and
(iv) the service provider must be owned and controlled by a Union legal entity and
the subcontractors are not subject to the control of a third country or a third-country
entity. A cloud computing service subject to the control of a third country or a
legal entity established in a third-country can still be audited against the audit
criteria where the third country has implemented specific safeguards that ensure
that there is no risk of unauthorised access to Union data or possible disruption of
service quality or continuity; and
(v) the data generated by using the audited service shall not be re-used for the training
or fine-tuning of an AI system operated by an entity outside the Union and in
any case are not transferred outside of the Union, and
(vi) the customer data, including metadata and telemetry data, remain in the Union
unless the customer explicitly requests otherwise; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at
least at level ‘substantial’ under the European Cybersecurity Certification Scheme
for Cloud Services (EUCS); and
(viii) the service provider must demonstrate a high degree of control over the software
components that underpin the service (software stack). This notably implies that
there exists a list of identified dependencies related to the provision of the service,
and where the software components are provided by a third-country entity, the
relevant code of the security relevant components of the service stack can be
audited, and there exists a migration plan in the event a vendor fails or a third-
country imposes restrictions; and
(ix) operational and technical support, including outsourcing, are initiated and
performed exclusively within the Union and by Union citizens, and by third parties
that are not subject to the control of a third country or third country entity; and
(x) where the cloud service provider is subject to the control of a third country or a
third country entity, it must be able to prove that the laws and government practices
in that country do not require the provider to tell that country's authorities about
software vulnerabilities before those vulnerabilities have been publicly discovered
These requirements would not allow service providers whose parent company is
headquartered outside of the Union to be considered as ‘sovereign level 3’.
To be considered ‘sovereign level 4’, a service must meet the following cumulative criteria:
(i) the service provider and subcontractors must be established in the Union; and
(ii) the service must be fully operated from computing infrastructure and assets
located in the Union; and
89
(iii) members of the board, executive team and personnel operating the service are
Union nationals, located in the Union, and are security cleared where
appropriate; and
(iv) the service provider must be owned and controlled by a Union legal entity and
the subcontractors involved in the provision of the service are located in the Union
and owned and controlled by a Union legal entity; and
(v) the data generated by using the audited service shall not be re-used for the training
or fine-tuning of an AI system operated by a third-country legal entity and in
any case are not transferred outside of the Union; and
(vi) the cusomter data, including metadata and telemtry data, remain exclusively in the
Union; and
(vii) the service must demonstrate a high level of cybersecurity by being certified at
least at level ‘high’ under the European Cybersecurity Certification Scheme for
Cloud Services (EUCS); and
(viii) the service provider must demonstrate effective control over the software
components that underpin the service (software stack) by demonstrating that a
third country or a third country entity does not have excessive control over the
software and the software lifecycle. This notably implies that the relevant code of
the service stack can be audited and that the effective control of the code exists by
a Union legal entity; and
(ix) operational and technical support, including outsourcing, are initiated and
performed exclusively within the Union and by Union citizens, and by third parties
that are not subject to the control of a third country or third country entity; and
(x) where the cloud service provider is subject to the control of a third country or a third
country entity, it must be able to prove that the laws and government practices in
that country do not require the provider to tell that country's authorities about
software vulnerabilities before those vulnerabilities have been publicly discovered.
These requirements would not allow providers headquartered outside of the Union to be
considered as ‘sovereign level 4’.
Sovereignty is not only cybersecurity. The criteria above deal solely with sovereignty.
Cybersecurity and sovereignty are closely related and are complementary. However, they do
not focus on the same aspects nor pursue identical objectives.
In this respect, sovereignty goes beyond the technical protection of services. It primarily
addresses who exercises control over the data, and under which legal framework this control is
exercised. A cloud service can be technically very secure and still be exposed to non-technical
risks stemming from the exposure to third country legislation, because it is subject to foreign
laws, or because key operations and decision-making are located outside of the EU. In such
cases, the residual risk stems from legal obligations rather than from technical vulnerabilities.
Hence, non-technical risks related to the potential access of third-country authorities to the data
cannot be fully addressed through technology or cybersecurity alone. Both need to be
understood and looked at as complementary measures, not mutually exclusive and partially
overlapping. Technical cybersecurity and non-technical risks address different threat vectors.
Properly designed security architectures, such as customer-controlled encryption, strict role
segregation or data access minimisation can technically constrain the practical ability of a third-
country authority to access the data, regardless of the provider’s jurisdiction. In this sense, a
technical cybersecurity implementation can serve as a meaningful mitigation strategy of a non-
technical risk.
90
Measures that significantly improve sovereignty (e.g. increasing strategic autonomy and
resilience, decreasing dependence) may have a limited incremental effect on classical
cybersecurity, while they are still justified for regulatory and trust-building reasons.
Conversely, some cybersecurity measures such as stronger encryption mechanisms can be
implemented without drastic changes to who controls the infrastructure and data or which legal
construct governs access, thus resulting in marginal improvements of sovereignty and
operational autonomy.
Even when a service is exceptionally well protected from a cybersecurity standpoint, a legally
binding request from third country authorities may still compel the provider to grant access to
the data, which constitutes a residual risk if robust internal processes are not in place. This risk
exists independently of the technical robustness of the service and the service provider and
stems from the provider’s legal obligations rather than from technical vulnerabilities. For this
reason, the way a service provider constructs its whole architecture becomes crucial and
typically follows a “cannot comply” paradigm. Under this approach, compliance with external
access requests in rendered technically, operationally or legally impossible (see the OVH case
in Canada56). This is achieved through well-defined governance and control mechanisms,
including strict internal processes governing how employees handle access requests, clear
delineated roles and responsibilities regarding data access, as well as more organisational
measures such as segregation of duties or control over the cryptographic keys (e.g. ensuring
that the encryption keys are not accessible to the provider or are held exclusively by the
customer). Although these architectures involve additional upfront and operational costs, the
majority in personnel, they demonstrably increase assurance and trust for customers, as they
reduce the scope for a compelled access to data through the provider and strengthen compliance
with sovereignty requirements. Organizations already have clear roles and responsibilities and
segregation of duties defined as part of their policies and procedures addressing technical
requirements and hence are not considered in the calculations. Only the additional costs such
as the definition of the policies and procedures to protect from the requests of third country
legislations with extraterritorial reach and the modification of controls are to be considered in
the adjustment costs.
Sovereign architectures incorporating measures to address non-technical risks often require
separate environments for jurisdiction, stricter network segmentation and redundant
infrastructure, thereby increasing the design, integration and maintenance effort compared to a
‘single’ cloud deployment, and introducing new misconfiguration risks. Complying with non-
technical requirements on top of cybersecurity obligations adds significant legal and
administrative efforts as providers must demonstrate extensive evidence of compliance and rely
on specialised expertise to do so.
One-off adjustment costs for the EC. In the case of the EC the costs for the definition of the
harmonized criteria are not to be accounted for as the definition is set in the Act. However, one
off adjustment costs to update the definition are accounted for with 0.5 FTE in 2032.
One-off adjustment costs for national public authorities. In the case of National Public
Authorities, one-off adjustment costs relate to the adaptation of their procurement templates
based on the harmonized criteria provided by the act. As in the case of the Commission, effort
has been estimated for both the initial changes and for the ones needed to be done after the
56 https://www.heise.de/en/news/Canadian-Court-OVHcloud-from-France-must-hand-over-user-data-11092029.html
91
revision of the harmonized criteria. In the first case, 3 days have been estimated whereas 1.5
days have been considered for the second activity. These values are per Public Administration
publishing public procurement of cloud and AI computing services and deciding to voluntarily
use the harmonized criteria for sovereignty. Therefore, the formula for the total one-off
adjustment costs must be computed as:
One-off adjustment costs (National Public Authorities) = time needed to adapt the
procurement templates * labour costs * number of procuring authorities using the criteria
Recurrent savings for national public authorities could result from having a clear harmonized
criteria for sovereignty. These are considered negligeable.
One-off adjustment costs in cloud and AI computing service providers. The effort to adhere
and align with the criteria for sovereignty only applies when these are linked to an official
document specifying the request to comply with the criteria for sovereign services for a specific
procedure. Negligeable administrative costs for cloud and AI operators is assumed since
voluntary in nature. This has been contrasted with stakeholders as part of the study led by
Technopolis group.
Recurrent savings in cloud and AI computing service providers. No recurrent savings are
assumed.
Indirect savings: The availability of harmonized criteria would create some certainty on what
is considered sovereign, avoiding different interpretations. Through this measure, organisations
able to adjust their processes to adhere to the criteria will increase trust from cloud consumers
and potentially resulting in larger market revenues for EU providers due to market credibility.
Sensitivity analysis: The data presented above is sensitive notably to the number of procuring
authorities under consideration. This has been set between a minimum of 50% and a maximum
of 100%. The average value is set to 75% of public authorities. Another important aspect is the
estimation of the percentage of procurement procedures that will voluntarily request the
alignment with the harmonized criteria for sovereignty.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. The results point towards low total costs for National Public
Authorities and negligeable costs for cloud and AI computing service providers. The total
estimated costs for the implementation of this measure are EUR 336 077 over the entire period.
These costs entirely stem from one-off adjustment costs. The main driver in the min and max
values is the number of countries that adopt the EU-level definition for their procurements.
Table 31. Creating an EU-level harmonized criteria for sovereign cloud and AI computing services
(PM11)
Cost types Cloud & AI
providers (€)
Public
authorities (€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - - - - - -
one-off adjustment - 336 077 - 336 077 224 051 448 289
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
92
Total costs - 336 077 - 336 077 224 051 448 289
Total benefits - - - - - -
Net impact - 336 077 - 336 077 224 051 448 289
3.12. PM12: Creating EU guidelines for sovereign Cloud AI computing services for
public procurement
The assumptions detailed for PM11 remain valid for PM12, with the following additions:
Whereas PM11 only considers the criteria of what constitutes a sovereign cloud and AI
computing service, PM12 adds detailed guidelines developed by the Commission over PM11’s
criteria to ensure that service providers interpret them in a similar manner. While the
Commission is the sole responsible for the drafting of the guidelines, these would be consulted
with representatives of national public authorities and service providers.
One-off adjustment costs for the European Commission stems from the development and
adoption of the EU guidelines for sovereign cloud and AI computing services. Guidelines entail
detailed explanations and interpretations and therefore the effort estimated to draft and adopt
them is 2 FTEs over 1 year.
One-off adjustment costs for national public authorities include the following items: firstly
the participation of Member States’ authorities in the discussions to draft the guidelines in
collaboration with the European Commission and the cloud and AI computing service providers
and secondly, the adaptation and updates of the bid templates as well as the institutionalisation
of certain elements so that the guidelines can be applied uniformly across the public sector.
The participation in the discussions for the guidelines has been estimated to 27.5 days per public
authority considering one participant per MS, while for the adaptation of the templates the
estimation is 3 days per public authority applying the guidelines, assuming that 904, the
majority of public administrations of the procuring authorities adopt them. A sensitivity is
introduced as a minimum of 3 days and a maximum of 10 days.
One-off adjustment costs (NPA) = labour costs * ((effort in discussions * number of
authorities involved in the discussion) + (effort to adapt the guidelines * number of
authorities adopting the guidelines)
Recurrent savings in national public authorities maystem from a voluntary but systematic
application of the adapted templates in the different procurement procedures. The clarity of the
criteria could result in a lower number of disputes between procuring public authorities and
cloud and AI computing service providers. However, interviews with Public Administrations
do not point to savings as a result of guidelines alone. As part of the sensitivity a min of zero
and a max of 5 days per procured tender is arbitrarily set.
Recurrent savings for reusing the harmonized criteria (national public authorities) =
estimated effort saved per tender * labour cost * number of yearly public tenders integrating
the guidelines
One-off adjustment costs for cloud and AI computing service providers. Under this PM, cloud
and AI computing service providers will participate in the discussions for the guidelines and
the effort estimated for this activity is 20 staff days for 30 cloud and AI computing service
providers.
The second type of one-off administrative costs is related to the alignment to the sovereignty
guidelines. This requires checking the undertakings’ internal organisational and legal
93
procedures. This effort has been estimated in 20 days, and it would apply to all economic
operators.
Recurrent adjustments costs in cloud and AI computing service providers. Under this measure
no adjustments costs for companies to comply with the requirements and modify organisational
or legal procedures, set by the guidelines.
Recurrent savings in cloud and AI computing services providers result from the systematic
application and reuse of the documentation that needs to be handed in for each procurement
procedure. cloud and AI computing service providers will be able to re-use a set of documents
throughout procurements since these will be aligned with the guidelines. These savings are
estimated at 1 day per tender and 10 tenders a year.
Recurrent savings (cloud and AI computing service providers) = effort saved per bid * labour
costs * number of yearly bids per provider * number of cloud and AI computing service
providers aligned to guidelines
Indirect savings: Through this measure, organisations able to adequate their processes to
adhere to the harmonized criteria for sovereign services will increase trust from cloud
consumers and potentially resulting in larger market revenues for EU providers due to market
credibility.
Sensitivity analysis: The items used for the sensitivity analysis are 1) the percentage of public
administrations using the harmonized criteria for sovereign services in their procurement
activities, ranging from 50% to 100%; 2) the number of public administrations procuring cloud
services; 3) time needed by public administrations to adopt the procedures as well as to evaluate
the bids; 4) the number of cloud and AI computing service providers that can participate in
procurement processes, ranging from 244 (min) to 350 (central and max).
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. The average scenario is slightly above cost neutral. While the
min scenario measure points towards benefits not fully offsetting costs the max scenario points
towards net benefits. The main source of difference between the min and max is caused in the
cost savings category and how many national public authorities integrate the guidelines to their
tenders and cloud and AI computing service providers adopt the guidelines.
Table 32. Creating EU guidelines for sovereign cloud and AI computing services for public procurement
(PM12)
Cost types Cloud & AI
providers (€)
Public
authorities
(€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - - - - - -
one-off adjustment 2 975 585 1 252 254 168 538 4 396 376 2 562 268 7 021 802
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 2 975 585 1 252 254168 5384 396 376 2 562 268 7 021 802
Total benefits (6 693 922) (5 693 922) (1 897 974) (16 358 845)
Net impact (2 718 336) 1 252 254 168 538 (1 297 545) 664 294 (9 337 043)
94
3.13. PM13: Annual conference on digital sovereignty
This policy measure aims at creating awareness and foster discussions on digital sovereignty,
sharing latest developments, best practices and solutions among different stakeholders such as
policymakers, cloud and AI computing service providers, researchers, or financial institutions.
It answers the call made by several cloud and AI computing service providers to gain visibility
through political exposure.
The event will be organised on a yearly basis and is expected to last for a week in a venue
outside of the European Commission. The target number of attendees is 300 people from the
private sector, 3 representatives from each member state national public administration and 100
representatives from the EUIBAs, whose costs are for modelling purposes assigned to those of
the Commission.
Recurrent adjustment costs for the European Commission, which result from the organisation
and attendance to the conference. EUR 35 000 per day, are estimated for subcontracting costs
covering organisation, logistics and operational activities. This includes venue rental,
communication with panellists, registration and admission process, badges and catering. For
the organisation, it is estimated that 50 staff days will be needed every year by the Commission
to oversee the activities carried out by the subcontractor. It is assumed that 5 officials from the
Commission will attend all 5 days and require additional 2.5 days for the preparation of their
participation in panels, keynotes or discussions. Travel costs are also considered.
Recurrent adjustment costs for national public authorities. 3 representatives per Member
State will attend and participate with an estimated effort of five days attending the event and
0.5 to 2.5 days for the preparation of their participation in panels, workshops or discussions.
Travel costs of EUR 700 per participant are considered.
Recurrent adjustment costs (NPA) = MS57 * Number of participants (PA) * (labour cost *
(number of days attending the conference + number of days preparing the workshop) + travel
costs per participant)
Recurrent savings in National Public Authorities. Public authorities will potentially achieve
savings staff from a reduction in the effort on bilateral exchanges with other stakeholders.
However, given the limited scale of this effect this has not been monetised.
Recurrent adjustment costs for cloud and AI computing service providers (and other
stakeholders). An effort between 5.5 and 7.5 days for the participation and preparation of the
event are considered for each of the 300 participants from private sector companies and other
stakeholders. Additionally, travel costs are included.
Recurrent adjustment costs (CASP) = Number of cloud and AI computing service providers
participating * ((number of days for the preparation and attendance * labour cost) + travel
costs)
Recurrent savings in cloud and AI computing service providers (and other stakeholders).
Event participants may save staff days from a reduction in the effort on bilateral exchanges.
Given the limited scale of this effect this has not been monetised.
Indirect savings:
57 MS unless stated otherwise is MS = 27
95
• For cloud and AI computing service providers, faster and better understanding of the
technologies and techniques to be secured and sovereign, hence delivering services into
the market that are perceived as more trustworthy and more up to date in terms of the
use of for instance dependency analysis tooling, ultimately leading to an increase in
market sales.
• For national public administrations: a clearer perception of what are the main risks
associated with the use of non-sovereign solutions, potentially reducing research on
attack and threat vectors, and ultimately resulting in an improved business continuity
of the digital services offered.
Sensitivity analysis: The impact of this policy measure depends on the number of events and
participants, but a sensitivity analysis is not relevant given the size of the impact.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. Low costs are assumed for all stakeholders modelled. Savings
are negligeable and hence not monetised.
Table 33. Annual conference on digital sovereignty (PM13)
Cost types Cloud & AI
providers (€)
Public
authorities
(€)
European
Commission (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - 1 972 376 - 1 972 376 1 972 376 1 972 376
one-off adjustment - - - - - -
recurrent adjustment 12 099 496 - 4 594 816 16 694 311 16 694 311 16 694 311
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 12 099 4961 972 3764 594 816 18 666 687 18 666 687 18 666 687
Total benefits --- - - -
Net impact 12 099 496 1 972 376 4 594 816 18 666 687 18 666 687 18 666 687
3.14. PM14: Interoperability flanking measures
Under this policy measure, the goal is to prioritise the support and development of new
harmonised standards under the Data Act in order to cover existing gaps and foster
interoperability among cloud services.
This policy measures therefore focuses on three aspects. Firstly, a contract for supporting the
Commission in continue analysing the gaps in standardisation, notably in PaaS and SaaS and
creating awareness of this lack of harmonised standards. Secondly, on the creation and
monitoring of coordination groups steered by the Commission with MS and industrial players
for potential recommendations on the prioritisation of standard development. Thirdly, on the
development of standards by European Standardisation Organisations (ESO). The conservative
approach is that two standards will be developed and adopted yearly.
The application of the standards is out of the scope of the current policy measure.
One-off adjustment costs for the European Commission are related to the budget reserved for
the contract on carrying out the studies (EUR 200 000) as well as for drafting the specifications
(0.25 FTE). It also includes an external study on portability of AI models, valued at EUR
200000.
96
One-off adjustment costs = external study + (effort for drafting the specifications * labour
costs) + study on portability of AI models
One-off administrative costs for the European Commission stem from the setting up of the
coordination group, which would hold participants coming from the industry and MS,
estimated in 2 FTEs.
One-off administrative costs = effort to set up the coordination group * labour costs
Recurrent administrative costs for the European Commission include, firstly, the
administrative support for the coordination of standards, both in the coordination group and in
the ESOs estimated in 2 FTEs. Secondly, it includes the administration of a support action (2%
of administrative overheads), funded under an R&D programme, fostering the participation of
European stakeholders in ESOs and standardization activities. The budget estimated for this
action is EUR 6 million, within a funding period of 6 years.
Recurrent administrative costs = labour costs * (effort estimated to monitor the coordination
group + effort to monitor the coordination with standardization organizations + effort for the
programme administrations)
One-off administrative costs for national public authorities are the result from MS selecting
the members that would participate in the coordination group set up. This is estimated in 0.1
FTE.
One-off administrative costs (NPAs) = MS * effort to identify members * labour costs
Recurrent administrative costs in national public authorities stem from the participation in
both the coordination groups and standardisation organisations (0.25 FTE for each activity) on
a yearly basis. Since ESOs offer the possibility to participate online in the meetings this has
been the option considered, and travel costs neglected.
Recurrent administrative costs (NPAs) = MS * (effort to participate in ESOs + effort to
participate in coordination groups) * labour costs
One-off adjustment costs for cloud and AI computing service providers are also related to the
setting up of the coordination group, estimated in 0.1 FTE.
One-off administrative (CASP) = number of participants * (effort to set up the coordination
group * labour cost).
Recurrent administrative costs for cloud and AI computing service providers are related to
the participation of the providers both in the coordination groups and ESOs estimated
respectively in 0.25 FTE and 1 FTE respectively. Whereas the meetings for the coordination
groups are to be more opportunistic, the meetings on ESOs occur on a regular basis58.
Recurrent administrative costs (CASP) = number of participants * ((effort to participate in
the coordination groups + effort to participate in the ESOs) * labour cost)
Recurrent administrative costs for European Standardisation organisations occur by
supporting the creation, development and publication of standards, as well as by the
58 Depending on the ESO the meetings of the Joint Technical Committees or Technical Committees are scheduled on a monthly
basis.
97
coordination with the European Commission and the key stakeholders such as the rapporteurs.
This effort is estimated in 1 FTE. This stakeholder category is not accounted for the in the CBA.
Indirect savings: For cloud and AI computing service providers, having clear interoperability
standards facilitates EU players the entry into established markets, lowers transaction and
integration costs and simplifies the management of a multi-cloud / multi-vendor ecosystem.
Finally, it also provides enhanced trust, credibility, transparency and auditability.
Sensitivity analysis: For this policy measure no sensitivity analysis has been considered, given
that only the development of standards is considered while the enforcement and compliance by
cloud and AI computing service providers are out of scope.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. Low costs are assumed for all stakeholders modelled. Savings
are negligeable and hence not monetised. Sensitivity analysis does not apply.
Table 34. Measures to effective interoperability of cloud and AI computing services (PM14)
Cost types Cloud & AI
providers (€)
Public
authorities
(€)
SDOs (€) European
Commission
(€)
Total Value
(central, €)
Total
Value (min,
€)
Total
Value
(max, €)
one-off administrative 11 205 170 644 - 168 538 350 387 350 387 350 387
recurrent administrative 2 134 177 10 917 230 1 437 660 1 546 004 16 035 070 16 035 070 16 035 070
one-off adjustment - - - 368 538 368 538 368 538 368 538
recurrent adjustment - - - - - - -
one-off regulatory fees - - - - - - -
recurrent regulatory
fees - - - - - - -
one-off enforcement. - - - - - - -
recurrent enforcement - - - - - - -
Total costs 2 145 381 11 087 8741 437 660 2 083 079 16 753 994 16 753 994 16 753 994
Total benefits - - - - - - -
Net impact 2 145 381 11 087 874 1 437 660 2 083 079 16 753 994 16 753 994 16 753 994
3.15. PM15: Voluntary sovereign risk assessments for the use of cloud and AI
computing services in the public sector
Policy Measure 15 (PM15) would recommend Member States to carry at least one
sovereignty risk assessment and repeat it at least every four years or more frequently if deemed
necessary. The purpose of the sovereignty risk assessment is to identify which public sector use
cases within a Member State require the use of which sovereignty level as described under
PM11. The sovereignty risk assessment would assess, inter alia, the risks induced by the access
to such data by a third-country authority or third-country legal entity; or the risk of possible
service disruption due to dependence on a single or limited number of third-country services
providers. On the basis of dedicated discussions conducted with 3 different public authorities
representing about 200 NIS 2 Annex 1 contracting authorities operating at regional, national
and European level (out of an estimate of 6 400 NIS2 entities across the EU),, this assessment
assumes that the matching of sovereignty levels to public sector demand follows the following
pattern: 70% of use cases would require a sovereignty level 1; 20% for level 2; 9% for level 3;
and 1% for level 4. Even though the scheme is novel and does not correspond to existing
frameworks, this assessment fits with broad orders of magnitude that can be inferred from
98
existing analyses conducted in several Member States that have introduced risk assessments
for their public sector clouds, such as France, Poland59 or Italy60.
Critical use cases, defined as the use cases whose disruption would affect operational autonomy
or public order, correspond to use cases covered by level 2, 3 and 4. The risk assessment would
have to consider the reality of the supply market to avoid unrealistic outcomes, such as
mandating the use of services that don’t exist (yet) in the market.
To facilitate appropriate and coherent sovereignty risk assessments, the European Commission
would develop guidelines for Member States to conduct such assessments and provide a sample
risk assessment methodology (note that these guidelines concern the conduct of risk
assessments and differ from PM12, which consist in explaining the different levels of
sovereignty). For Member States to have up-to-date information about the market conditions
of cloud and AI sovereign solutions, the Commission would also produce market monitoring
reports that will point Member States to possible gaps in the coverage of some services.
The Member State would be recommended to reflect the outcome of the risk assessment in
applicable public tenders.
While PM11 only puts forward the definition of sovereignty levels, PM15 goes further by
putting forward a framework through which the respective levels of sovereignty can be
assessed.
Cloud service providers shall submit the relevant evidence that demonstrate that they comply
with the sovereignty assurance criteria to the designated competent authority. For assurance
level 1, the service provider will submit the competent authority their self-assessment report
for the service. Cloud computing service providers qualifying as SMEs will not be required to
undergo the validation by the national competent authority. Verifying the compliance of the
service against sovereignty level 2-3-4 would be done through third party’s auditors. In this
case, the service provider will submit the audit report and the ‘positive’ audit opinion to the
competent authority of the country of establishment who shall verify them, along with the
submitted evidence, without undue delay. The result of this activity is an acceptance or rejection
of the audit report and opinion.
If a cloud service provider wants to participate in public procurement procedures across the
Union, the validating Member State shall notify the other Member States for a review of 60
days.
During the period of review there is a recourse possibility for the Member States. If no reasoned
objection has been submitted in the framed period the validation of the audit report and audit
opinion shall be considered accepted by all Member States and the service recognized across
the Union. If continued objections are raised, the Commission shall adopt a binding opinion.
The competent authorities should register the decision in a Union repository, maintained by the
Commission.The repository of sovereign cloud and AI computing services will be a public list
of audited sovereign cloud and AI computing services that verifiably comply with the
sovereignty requirements. The benefits are for providers and users alike: providers will enhance
their visibility and users their market research.
59 See Cloud in Government Services 60 See Strategia Cloud Italia
99
To cater for market evolutions, the sovereignty criteria of all levels and evaluation evidence
proposed, but not limited to, would be modifiable by comitology. This evaluation evidence
would help third party auditors in their assessment of the service and ensure full harmonisation
in the way different auditors conduct their assessment and for Member States to ensure that the
procedures have been followed.
The assessment would be peridoically renewable following the same evaluation methodology.
Policy measure 15 then implies that the Commission sets up and maintains a repository of the
services audited against level 2-3-4.
Finally, policy measure 15 implies the actual transfer cost for public authorities to change from
a non-sovereign service to a sovereign service (cloud porting).
This measure is primarily designed to contribute to the protection of public order by enhancing
the resilience in the public sector, which is SO4. Nevertheless, European providers would face
less costs and efforts to meet sovereignty conditions. When it comes to meeting the criteria to
demonstrate sovereignty level 1 and 2, EU providers can more readily substantiate that they are
not affected by third-country policies affecting data access or limiting service continuity. As
well, level 3 and level 4 sovereignty can only be served by service providers owned and
controlled by EU entities. This implies that PM15 will also contribute to decreasing the overall
reliance on non-European cloud and AI computing services, which is SO3.
One-off adjustment costs for the European Commission
Setting up and maintaining the repository of sovereign cloud and AI computing services. These
costs include the drafting of the tender specifications for an external provider to build, develop
and maintain the repository plus additional time for the evaluation and the definition of their
governance procedures. This has been estimated to 120 staff days. The repository will be
developed by an external contractor with a contract value of EUR 500 000.
One-off adjustment costs (EC) = time to draft tender specifications for the repository *
labour costs + external contract to develop the repository
Guidelines for MS to perform the sovereignty risk assessments. In order to support Member
States in the implementation of their sovereignty risk assessments, the Commission will publish
guidelines. These guidelines would include the process to carry out the risk assessment, the
criteria that would allow them to identify the scope, for instance by means of use cases and / or
applications, the data classification process, where the market surveillance or information on
sovereign services can be found. The estimated effort for the development and publication of
these guidelines is 2 FTEs over 1 year.
Recurrent adjustment costs for the European Commission
Maintenance of the repository for audited services. Another important activity is related to the
maintenance of the repository of audited services and related activities, including:
• Yearly hosting costs of the repository, estimated to be of EUR 200 000. The repository
will be deployed on a sovereign cloud at the Commissions’ premisses and telemetry
and FinOps procedures will be put in place to control expenses.
100
• Yearly maintenance of the repository, performed by an external contractor, includes the
upgrades of the software libraries amidst identified vulnerabilities, correction of errors
and bugs and additional features. The estimated effort assumed for this is 2 staff days
on a weekly basis.
• Oversight of the repository, including the follow-up of the contractors’ activities,
estimated in 2 staff days per month.
Recurrent adjustment costs (EC) = costs of the yearly hosting + external contract for the
yearly maintenance of the repository + (number of days for the oversight of the repository *
labour costs)
One-off adjustment costs for public authorities.
Conduct the risk assessments. National public authorities and regional or federal authorities in
the case of decentralized Member States shall perform at one sovereignty risk assessment, with
the aim of mapping sovereignty assurance levels to cloud and AI computing services used in
the public sector, taking into consideration aspects such as sensitivity and criticality of the data
processed, the risks associated to the potential unlawful access to data by third country
legislation, potential disruption of the services, and the existence of computing services audited
under that sovereignty level, among others. The estimated number of authorities is 267,
considering the NUTS-0 and NUTS-1 distribution. Given that it is a voluntary measure, it is
considered that the uptake will be only by in 25% of the authorities, that is, 67.
The output of this risk assessment will be a classification of applications or use cases categories
mapped to the minimum sovereignty assurance level permitted for the procurement of cloud
and AI services serving those cases.
The estimated effort for this sovereignty risk assessment is of 10 FTEs. This is a new activity
for which there is yet few expertise in the Member States. It is considered that in the initial
iteration, Member States will need to understand well the sectors, applications and type of data,
among other aspects, that will drive their risk analysis to decide one sovereignty assurance level
or another. It is expected that initially this will be an activity encompassing various ministries
and agencies and therefore an intensive collaboration is expected. The guidelines from the
Commission, however, should alleviate this effort.
The risk assessments will start in 2029, once EUCS and the sovereignty framework is in place.
Designation of the competent authority, which has an estimated of 0.2 FTE per MS, which will
be responsible to verify the audit reports received from the providers that had their services
audited at sovereign level 2 at least.
Cloud-to-sovereign-cloud migration costs for the public sector. Detailed information is
provided in Annex 12 on the methodology and calculation of the costs for the migration and
porting of cloud applications to sovereign cloud services.
Due to the voluntary nature of the measure and given that transition to a sovereignty service is
recommended and not mandatory, only 25% of the cloudified and the cloudifiable IT systems
would be ported or migrated to a sovereign solution.
Both the migration of applications from legacy-to-sovereign cloud or from cloud-to-sovereign
cloud will be performed in a linear distributed manner in a time frame spanning 5 years, starting
in 2029, once there is a set of services audited as sovereign.
101
The average cost of porting a single application to a new cloud varies from EUR 30 000 and
EUR 600 000 depending on the size of the application, and not much on the type of cloud at
origin or destination, as they are mainly based on human effort (400 to 8000 hours of work).
See Annex 12 for a explanation of these figures.
These costs have not been quantified in the cost-benefit analysis, although they are described
in detail, and illustrated with the real use cases in Section 6.1.2 of the main text.
Recurrent administrative costs for national public authorities
Sovereignty risk assessment. This risk assessment shall be renewed every 4 years. The effort
estimated for this reassessment of the sovereignty risks is estimated to be 5 FTEs, a lower effort
than implementing the risk assessment from scratch, given that the activity is already ramped
up and the Member States would already have a good understanding of their situation. 67
authorities will carry out these assessments and therefore will have to renew them.
Revision of the audit reports by the competent authorities, per service audited at sovereign level
2 at least, estimated at 5 days per service. After verifying the audit report, opinion and evidence
and consulting with the other competent authorities, the evaluating competent authority will
adopt a decision that would allow a service provider to participate in public procurement
activities.
Recurrent savings in National Public Authorities
Running applications on the cloud. Operating applications on the cloud, if performed well, with
a continuous FinOps verifying that services are not overprovisioned can lead to significant
savings. Literature sources estimate that the total cost of ownership (TCO) savings can amount
to 20-50% in the case of legacy-to-cloud migrations61. No savings are here accounted for in the
operation of applications migrated from cloud-to-sovereign cloud, as the applications would
have already been benefitting from the total cost of ownership savings. See Annex 12 for a
further qualitative discussion on savings.
Simpler public procurement. Savings in this respect stem from the simplification of voluntarily
using procuring audited services, which are part of the repository of sovereign cloud and AI
computing services. This shall allow public administrations to save time in the verification of
the documentation for the evaluation of the offer. This will result in 2 staff days per bid, and
assuming that there are 3.5 potential contractors presenting bids in each tender procedure.
Recurrent savings = Average number of yearly service bids requiring sovereignty assurance
level 2 – 4 * staff costs savings per bid
Sovereign services. The savings from using sovereign services cannot be quantified due to the
intangible nature of sovereignty and resilience.
Costs for service providers to develop sovereign services
Assessing the cost and benefits for providers to provide sovereign services is a complex task
that involves many parameters and differs greatly from provider to provider. As well, in the
absence of an established market for sovereign services, data sources are rather anchored in
providers’ business plans, not in ex post analysis of established businesses. Such data is
61 Chatzithanasis & Michalakelis (2018) estimates 24% to 50% The Benefits of Cloud Computing:. Ali Khajeh-Hosseini, David
Greenwood, Ian Sommerville
102
confidential to companies, and the below considerations are based on targeted discussions with
stakeholders that requested to remain anonymous. A first consideration is that the consulted
companies unanimously indicated that, in developing the business plan for these new services,
they count on new large critical use cases that are today not in the cloud; in other words, they
see sovereign services to generate a new source of income, but not to substitute existing.
Cost wise, new costs notably include the amortisation of all one-off adjustment costs such as
the additional cost induced by using EU-located infrastructure, the additional compliance costs
induced by the audits, the additional costs of being certified under EUCS; and proper recurrent
costs such as the higher salaries of employing EU workforce. As an illustration, speaking under
the condition of anonymity for business secrecy reasons, one of the largest EU service providers
of sovereign services speaks of an overall investment of EUR 1.5 bn, including hardware, for
a broad range of IaaS and PaaS services (for an unspecified computing capacity). Conversely,
another EU service provider with an established range of non-sovereign services speaks of an
overall investment in the range of EUR 20-40 m to adjust existing hardware and software to
the stricter norms that a sovereign service entails, with plans to invest progressively should the
market develops.
The benefits for service providers to develop sovereign services are covered under section
6.1.5.
Benefit wise, cloudifying legacy on-premises services means new revenues. As to the move
from traditional cloud to sovereign cloud, today’s few available sovereign services come with
a mark-up which is still too uncertain to draw conclusions (see discussion under Problem Driver
4 in 2.3.4). To this additional income, consulted EU service providers see also sovereign
services as niche market where they have a demonstrable added value that could spill over to
other market segments. The consulted companies unanimously indicated that it is too early to
confirm whether, over the span of the next ten years, the incremental cost and benefits would
reach the same balance (i.e. margins) as with equivalent non-sovereign services, which is today
commonly assumed to be around 30%62; to cater for this uncertainty, this assessment assumes
that stakeholders’ margins vary from 25% to 35%.
One-off adjustment costs for cloud and AI computing service providers
Ultimately, the decision of a provider to audit one or more services is their own business
decision. However, in order to have a sense of magnitude of how many providers could decide
to do so, extensive desk research conducted as part of the preparatory study (Technopolis et al.,
2025) identified that services from 59 non-EU headquartered cloud service providers likely
meet Level 1 requirements and would be able to qualify to Level 2 should they decide so. 226
EU headquartered providers could qualify their services as Level 2 and would be able to qualify
their services as Level 3/4 should they decide so. If only large companies are considered and
SMEs are excluded, the figure goes down to 59 EU headquartered large companies.
The assumptions are as follows: 150 services will be audited from 2029 until 2034, with the
assumption of 30 audits the first year, and an annual growth of 38%. This builds over the
number of services currently qualified in comparable national qualification and certification
schemes63, and the FedRAMP mechanism. As of September 2025. the FedRAMP repository
contains 530 authorised services under level moderate and high.
62 Amazon Web Services profits squeezed as AI arms race drives spending surge – GeekWire 63 For this impact assessment ES, IT, and FR have been considered as baseline as their data is public.
103
Services may be audited under different levels, provided that the criteria mentioned above (see
PM11) are met. For the purpose of this impact assessment, it is estimated that the proportion
of audited services in the different assurance levels will be mirroring the needs of the public
authorities (see below). This would then make that out of the estimated 150 audited cloud
services:
• 70% of the services will be Sovereignty Assurance Level 1, making 105 services
• 20% of the services will be Sovereignty Assurance Level 2, resulting 30 services
• 9% of the services will be Sovereignty Assurance Level 3, amounting to 13 services
• 1% of the services will be Sovereignty Assurance Level 4, resulting in 2 service.
The effort needed to get audited under sovereignty assurance level 2-4 is estimated to be 15
FTEs. This includes the definition and implementation of the necessary legal, organisational
and technical measures to reach sovereignty assurance level 2 – 4 and the auditing procedure
carried by the third party auditor. This effort was validated in the Final Validation workshop
held as part of the study led by Technopolis.
One-off adjustment cost of new audits = (effort to get sovereignty assurance level 2 – 4
audited * labour costs) * number of audited services assurance level 2 – 4
One-off regulatory fee for cloud and AI computing service providers
New audits, assumed to be EUR 20 000, for the estimated number of audited services at
sovereignty assurance level 2– 4.
One off regulatory fee for new audits (CSAP) = regulatory fee * number of audited services
assurance level 2 – 4
Recurrent regulatory fee for cloud and AI computing service provider
Renewal of audits, assumed to be EUR 14 000.
Recurrent regulatory fee for new audits (CSAP) = regulatory fee * number of audited
services assurance level 2 – 4
Recurrent administrative costs for cloud and AI computing service providers
Audits: These compliance costs stem from the audits to maintain the audit certificate, where the
goal is to verify that the conditions under which it was first obtained remain valid (effort
estimated as 3 FTE/year per service). Compared to the effort needed to achieve it initially, the
effort dedicated to the intermediate audits has been assumed to be lower than obtaining it
initially, in spite of the annual reporting obligations set out above.
Recurrent administrative compliance cost of renewal of audits = (time to be audited * labour
costs) * number of audited services
Recurrent savings for cloud and AI computing service providers.
Audits.A crucial matter in the analysis of the recurrent savings for cloud and AI computing
service providers is the single market effect. Providers will be able to be audited in one MS and
operate in all EU27 at once. Additionally, the analysis considers that in the business-as-usual
scenario one cloud service provider would seek to obtain the audit in 5 MS in average. Given
the costs of compliance, a cloud and AI computing service provider would carefully prioritize
in which MS(s) it would seek to obtain the audit, usually taking into consideration aspects such
as the market share or other commercial opportunities. With this policy measure, a single audit
104
provided by a MS will enable a provider to present offers including the audited service to public
tenders in all 27 MS.
Saving costs audits valid throughout the EU = Number of MS where the cloud and AI
computing service providers would get certified in the absence of an EU approach * cost to
get audited * number of audited services
Public procurement. Another relevant saving stems from the fact of not having to collect and
resubmit all the evidence necessary to demonstrate that the criteria are fulfilled for each bid,
given that the audited services are listed in the repository and accessible for the contracting
authority. The assumption is that undertakings will save 5 staff days / year per service and
bid. The number of yearly bids amounts to 50 a year.
Business costs saved: = Businesses time saved per evidence submission in a tender procedure
* number of submitted bids
Recurrent administrative costs for the private sector (Auditors)
Third party audits. In this PM, as mentioned above, the target number of cloud and AI
computing services that will be in possession of the positive audit report is estimated to reach
150 by 203264, whereas 30 of these will be audited sovereign level 2-4 during the first year.
The assumed CAGR is 38%. The assumption under this policy measures follows a linear
progressive uptake of the certification scheme.
The estimated average effort for auditing a service under Sovereignty assurance level 2-4 is
110 days / audit. The same effort is considered for the renewal process..
Indirect savings: Through this measure, cloud and AI computing services able to comply with
the sovereignty audit scheme will increase trust from public administrations, and notably for
the most critical cases. This will result in more trustworthy, credible and very specialized as
well as very advanced offers responding to challenges such as resilience and autonomy.
Sensitivity analysis: The most affected parameters for this Policy Measure are as follows: 1)
number of FTEs required to get audited and to maintain the audit (min and max value are 10
and 20 respectively). For the maintenance of the audit this varies between 2 and 5; 2) the
number of MS that designate the competent authorities to carry out the sovereignty risk
assessments. This varies from 5 MS (min) to 27 (MS), with 16 being the central value: 3) in
terms of savings, the common audit process will allow to save audit efforts in some of the
markets where the provider operates, considered to be a minimum of 3 MS and a maximum of
10, with a central value of 5.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. For all scenarios including the average, min and max the
benefits outweigh costs. What causes the range between the min and the mix is dependent on
the number of Member States which take up the scheme.
Table 35. Voluntary sovereign risk assessment for the use of cloud and AI computing services in the
public sector (PM15)
Cost types Cloud & AI
providers (€)
Public
authorities (€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
64 Assumption taken considering the current valid authorised cloud services in FedRAMP moderate and high, which amounts
to 531 as of 26 September 2025.
105
one-off
administrative
- - - - - -
recurrent
administrative
25 161 510 17 314 082 42 475 592 25 383 464 2 763 406 090
one-off adjustment 51 370 998 39 378 999 693 692 91 443 689 59 940 678 247 443 239
recurrent adjustment - - 2 117 274 2 117 274 2 117 274 44 128 193
one-off regulatory
fees
773 977 - - 773 977 773 977 773 977
recurrent regulatory
fees
1 047 966 - - 1 047 966 1 047 966 1 047 966
one-off enforcement. - - - - - -
recurrent
enforcement
- - - - - -
Total costs 78 354 450 56 693 081 2 810 965137 858 497 89 263 359 232 817 048
Total benefits (451 348 550) (3 280 022) -(454 628 572) (172 234 292) (566 620 181)
Net impact (372 994 099) 53 413 060 2 810 965 (316 770 075) (82 970 933) (333 803 133)
This policy measure would trigger an acceleration of porting of some applications to sovereign
cloud services of assurance levels 2 to 4, requiring an anticipation of expenses of EUR 620 m
to 4 bn. These costs are estimated and reported under section 12.4 of Annex 4. Since they
represent planned expenditure that would be incurred in the future as part of the regular cloud
contract renewal cycles and independent of the present policy measure, they have been scoped
outside of the summary cost table so as not to conflate structural renewal costs with the
incremental financial impact of the measure itself.
The benefits accruing from the risk reduction, notably an increased autonomy and strengthened
operational resilience through the adoption of sovereign cloud and AI computing solutions are
intangible and cannot be expressed in quantitative terms. While these benefits cannot be
quantified, they are acknowledged as a significant consideration underpinning the policy
rationale.
3.16. PM16: Non-mandatory specific award criteria for the procurement of cloud
and AI computing services
This policy measure defines non-mandatory award criteria for the procurement of cloud and AI
computing services, rewarding their specific characteristics in terms of EU added value.
With a view to enabling an enhanced EU value added of public procurement, this measure
establishes a set of voluntary non-price award criteria to be included, voluntarily, by contracting
authorities in the procurement of cloud and AI computing services. These criteria will earn
additional points to tenderers that demonstrate:
(i) Their contribution to reinforcing the digital technology supply chain in the Union,
including the use of software or hardware designed or manufactured in the Union
(ii) Integrated Union technologies, including the uptake of research and development
results stemming from EU-funded research and development programs;
(iii) That the innovation required to deliver the service being procured is conducted
in the Union or in a third country that contributes to the development of a European
cloud and AI ecosystem;
(iv) That the service is delivered, to the greatest extent feasible with regard to market
availability or technical requirements, through critical [computing, storage and
networking] hardware components designed in the Union or manufactured in the
Union, or both, or, where this is not feasible, through hardware components from a
106
country or countries that contribute to strengthening security of supply and
developing a European cloud and AI ecosystem;
Next, the baseline data used for this policy measure is detailed.
Number of public tenders for cloud and AI computing services. The data is taken from the TED
portal65 where all public tenders open by MSs are recorded. The list of tenders is filtered by the
CPV related to IT services, software and computing, i.e. CPV codes under the families of
72000000, 30000000, and 48000000, and by keywords “cloud” and “AI”, which results in
1.043 tenders for 2024. This is expected to grow until 2030 using the same CAGR as the one
observed for these tenders in the period 2021-24 (8.6%). From 2031 until the end of the period,
the CAGR is reduced to half (4.3%) as a result of the improvement in cloud and AI adoption.
It is assumed that one fourth of the tenders use the voluntary criteria.
The assumption is that all national public authorities participate in this measure even if not
mandatory and each of them receives 3 to 4 offers for each public tender they organise.
Cloud and AI computing service providers that participate in public procurement every year.
An estimate of 350 providers is used in the calculation. The estimate is based on a combination
of TED and desk research.
One-off administrative costs in national public authorities. Administrations that choose to
adhere to these criteria will need to adapt their procurement templates by integrating the
specific award criteria, with a low effort of up to 10 staff days for each public authority. This
is translated into the corresponding cost by multiplying by the average salary and the number
of public authorities introducing the specific award criteria. It is assumed that 25% of the
authorities decide to use the award criteria in their tenders and hence make the corresponding
changes.
One-off administrative costs (NPA) = number of national public authorities introducing the
non-price award criteria
(labour cost * effort to adapt templates)
Recurrent savings in national public authorities. Public authorities will use the standard
award criteria and guidelines instead of drafting those separately. This represents a saving and
is estimated at an assumption of 1.5 days per public tender.
Recurrent savings (NPA) = number of tenders
(labour cost * effort saved)
One-off adjustment costs for cloud and AI computing service providers. Providers who
choose to participate in procurement processes including the specific award criteria need to
align with the public procurement procedures for cloud and AI computing services. This
includes adopting and aligning their processes and improving them to meet these criteria, with
an estimated one-off effort of up to 20 staff days.
Recurrent administrative costs for cloud and AI computing service providers, are the costs
needed to adequate the offers, assumed to be 2 staff days per bid. These days were validated
during the interviews as part of the supporting study. They assume only administrative work
for companies who would be eligible. The cost is calculated multiplying this effort by the
65 EU-Tenders search APIs: https://docs.ted.europa.eu/api/latest/search.html
107
number of bids presented in a year and the average salary of the cloud and AI computing service
providers.
Recurrent administrative cost = number of bids presented * effort / bid * labour costs
Recurrent savings for cloud and AI computing service providers. Revenue gains are possible,
especially for providers who can differentiate due to the award criteria. These are however not
modelled.
Indirect savings: There are notable indirect benefits for cloud and AI computing service
providers that are able to comply with the different non-specific award criteria. For instance,
additional revenue gains, increase in reputation and the possibility to enlarge the scope of their
solutions which in turn may lead to an increase of sales.
Sensitivity analysis: The impact of this policy measure is sensitive to the number of public
authorities procuring cloud and AI computing services, the number of tenders and offers per
year and the number of cloud and AI computing service providers and the costs to adapt the
offers.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. Total costs are estimated between EUR 6.0 million and EUR
31.5 million, consisting mainly of recurrent administrative costs related to coordination,
reporting, and compliance activities, alongside smaller one-off administrative and adjustment
costs. Total cost savings are limited, ranging from EUR 1.1 million to EUR 3.3 million. Costs
outweigh benefits across the average, min and max set of parameters. The difference in the min
and max is mainly caused by the number of CSP and AI providers participating in public
procurement every year across EU.
Table 36. Non-mandatory specific award criteria for cloud and AI computing service procurement
(PM16)
Cost types Cloud & AI
providers (€)
Public
authorities (€)
European
Commissio
n (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - 1 731 580 - 1 731 580 865 790 2 597 370
recurrent administrative 13 486 479 - - 13 486 479 3 371 620 23 601 338
one-off adjustment 3 560 007 - - 3 560 007 1 780 004 5 340 011
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 17 046 486 1 731 580 -18 778 066 6 017 413 31 538 719
Total benefits -(1 630 135) -(1 630 135) (1 086 757) (3 260 270)
Net impact 17 046 486 101 445 - 17 147 931 4 930 657 28 278 449
3.17. PM17: Public Sector Cloud Federation
PM17 establishes a European public sector cloud federation for voluntary participation
European and national public authorities at all levels to allow them to share their own existing
data centre, cloud and AI computing services. This policy measure requires the establishment
of the technical back end that operates the necessary brokering services.
108
Cloud Capacity in the Public Sector. The public sector in Europe is running a data centre
capacity of 1.5 GW66 in 2025 and shows an annual increase of 8.9% due to the growing
adoption of PaaS and SaaS services67, that might be impacted in the first years with this policy
measure. As the measure is voluntary, it is assumed that only a part (two thirds) of the public
authorities will join the federation, reducing the capacity available in the federation to 1 GW.
Assuming a power usage effectiveness (PUE) of 1.29, a consumption of 500 W per server and
32 virtual machines (VMs) per server, the 1 GW of datacentre capacity could deliver around
50 million mid-sized VMs of 4 vCPU and 16 GB.
Number of VMs = ((1 000 000 000 W/ PUE) / 500 W per server) x 32 VMs per server
Market price of a VM is set at 0.23 EUR per hour based on a baseline configuration of 4 vCPUs
(processors) and 16 GB (memory)68. Considering the high margins in commercial cloud
services, beyond 60%, and a cost-based chargeback among participants to the federation, a
conservative assumption of a 30% price advantage of the cloud federation services with respect
to commercial cloud services is made, setting the cloud federation’s VM price at 0.16 EUR per
hour. This assumption was confirmed in the Final Validation workshop held as part of the study
carried out by Technopolis et al. (2025)).
Over time, VM configurations will evolve to higher performance and higher share of AI
processing units (GPUs, TPUs, etc). Prices are considered to remain relatively constant as these
more sophisticated configurations will get commoditized over time.
A target utilization of 60% is assumed for the cloud capacity based on own infrastructure. This
is based on a report from Berkeley Lab on data centre energy usage69, considering an evolution
of public sector data centres from current low occupation in traditional compute (below 20%),
towards higher occupation levels as the load of AI inferencing and AI modelling grows
(datacentres with this kind of loads present an average of 40% and 80% time operational,
respectively). A middle point (60%) is set as the target.
Applying a conservative approach only 25% of the idle capacity, 10% of the total capacity, is
considered to be shared with other members of the cloud federation. On the other hand, cloud
capacity extensions in each MS are assumed to be reduced by 10% by using idle capacity from
other federation members.
With regards to the number of public authorities that own and run a datacentre infrastructure
and could be members of the public sector cloud federation, in the absence of a precise number,
66 Based on calculations by Technopolis et al. (2025). The study estimates a total computing capacity consumed by the public
sector in EU27 in 2025 of 2.9 GW. The CATI surveys shows that 51% of this computing capacity (1.5 GW) is served by
datacentre infrastructure own by the public administrations. This is the capacity that is shared in the cloud federation. 67 Source: Technopolis et al. (2025) 68 Example taken from a basic AWS EC2 product using AWS configuration tool: https://calculator.aws/#/createCalculator/ec2-
enhancement. Parameters used: <Tenancy:Dedicated Instances>, <Region:Frankfurt>, <Operating system:Linux>,
<vCPUs:4>, <Memory:16 GiB>, <Network Performance:Up to 12500 Megabit>. Price calculated at 12th Sep 2025 for
<Instance name:m6i.xlarge>: 0,253 USD/hour (0,227 EUR/h at the average exchange rate for 2025 by Sept 2025: 0,8986
EUR/USD)69 Shehabi, A., Smith, S.J., Hubbard, A., Newkirk, A., Lei, N., Siddik, M.A.B., Holecek, B., Koomey, J., Masanet, E., Sartor,
D. 2024. 2024 United States Data Center Energy Usage Report. Lawrence Berkeley National Laboratory, Berkeley, California.
LBNL-2001637. https://escholarship.org/uc/item/32d6m0d1
109
the assumption made is that there is one in each NUTS1 and NUTS2 area70, resulting in a figure
of 300 public authorities in Europe.
One-off adjustment costs for the European Commission. This policy measure entails one-off
adjustment costs to establish the public sector cloud federation platform. These costs include 4
persons half time in 2027 (2 FTE) from the EC to run the procurement process for the federation
platform (define requirements and prepare the call, assess and award the contract), that will be
tasked to a subcontractor for an estimated cost of EUR 20 000 000. This includes the technical
aspects of the platform, as well as the development of the service level agreements and of the
technical common specifications.
One-off adjustment costs = external contract to develop the platform + (effort to run
the procurement process * labour costs)
Recurrent administrative costs for the European Commission. On a yearly basis, the
operation of the federation platform requires 6 FTEs.
Recurrent adjustment costs for the European Commission relate to hosting of the platform in
a cloud, estimated at 1 million EUR per year, and operating, maintaining and evolving the
federation platform, with an effort of 60 contractor FTEs per year, including the integration of
APIs and connectors to aggregate the NPA cloud resources to this platform.
Recurrent adjustment costs = hosting of the platform + maintenance of the platform
One-off and recurrent adjustment costs for National Public Administrations. Member States
dedicate an effort of 55 staff days (0.25 FTEs) in 2027 to participate in the setup of the public
sector cloud federation.
On a recurrent basis during the first years, up to 2031, each NPA will have to apply efforts to
connect national infrastructures and ensure interoperability to join the federation (2 FTEs), to
onboard cloud resources and services on the federation platform (30 staff days) and to update
and delete them later on (20 and 50 staff days respectively), making a total yearly effort of 2.45
FTEs.
Recurrent savings for National Public Administrations. With the measure and thanks to the
federation platform, NPAs will save 0.5 FTEs in the coordination with other Member States
for the sharing of cloud capacity.
In addition, the sharing of 10% of the idle VM capacity on a pay-per-use mode (with capacity
assumed to be consumed during 10% of the time) will represent a charge back of around EUR
690 million EUR inz the first year (2027).
Savings = VM capacity x %shared x price/hour x number of hours
The acquisition of 10% of the cloud capacity demand from the cloud federation will derive in
around EUR 296 million EUR savings in 2027, due to the 30% price difference between
federation and commercial cloud services.
70 NUTS1 and NUTS2 areas in Europe are close to 340 (https://ec.europa.eu/eurostat/web/nuts). An adjustment has been made
to slightly reduce the number of NUTS2 areas with a public authority owning and running datacentre infrastructure in the
bigger markets where the administrative distribution at that level is more granular.
110
Savings = VM capacity x %extension avoided x price difference x usage
These savings grow over the years with the expected increase in cloud capacity demand (8.9%
annually).
Indirect savings: Through this measure, the public sector achieves additional data centre
efficiency and improves the return on investment from their infrastructure and assets in cloud
and AI. The increase in resource utilization achieved by pooling resources reduces the
investment risk and may stimulate the increase in their investment in data centre, cloud and AI
capacity.
Sensitivity analysis: The values to be challenged in this policy measure include the capacity
shared by each MS with the federation ranging from 5% to 15% which impacts the savings in
terms of idle capacity reduction.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. The main difference stem from the percentage of idle resources
shared and particularly by the Member States participating in the federation.
Table 37. Public Sector Cloud Federation (PM17)
Cost types Cloud &
AI
providers
(€)
Public
authorities (€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - - - - -
recurrent administrative - - 4 312 980 4 312 980 4 312 980 4 312 980
one-off adjustment - 4 613 563 19 540 197 24 153 761 23 231 048 25 076 473
recurrent adjustment - 213 669 483 99 241 949 312 911 432 269 386 167 356 436 698
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs - 218 283 046 123 095 126341 378 173 296 930 195 385 826 151
Total benefits - (12 533 772 955) - (12 533 722 955) (8 175 327 110) (16 892 218 800)
Net impact - (12 315 489 908) 123 095 126 (12 192 394 782) (7 878 396 915) (16 506 392 649)
3.18. PM18: Vendor-neutral EU cloud/AI skill certificates
The goal of this policy measure is to create a vendor neutral cloud and AI computing services
training programme and curricula for the re-skilling and upskilling of workers in key digital
technologies for the single market. Targets of these certification and training programmes are
individual workers and students, not organisations.
Nowadays, many cloud service providers offer training and certification programmes (see for
instance AWS71, Microsoft Azure72 or Google73) but they are vendor specific and tailored to
their technologies. And this, despite the commonalities that exist across different cloud
71 https://aws.amazon.com/certification/ 72 https://learn.microsoft.com/en-us/credentials/browse/?credential_types=certification&products=azure 73 https://grow.google/intl/europe/courses-and-tools/?category=career&topic=cloud-computing
111
platforms: architectural patterns, cloud security controls, and development lifecycle
management methodologies, among others.
There are few vendor-neutral cloud skills certification programmes in the market, and they are
majorly focused on security74, leaving aside important aspects such as the design and
architecture of cloud native applications, AI services, DevOps, AIOps or FinOps75. University
degrees include some of these subjects, but they are typically integrated in long studies, target
undergraduate students and are not usually available as vocational training for professionals.
Under this policy measure, it will be created a curriculum of vendor neutral courses, as well as
their content, both for the materials and the exams. In principle, obtaining these certificates will
be a voluntary mechanism intended for developers whereas companies may require them in
certain job offers at their discretion.
One-off adjustment costs for the European Commission. The adjustment costs modelled
result from the subcontracting of a study under a public tender procedure that will
1) define the gaps on the current skills panorama to develop and deploy cloud and AI computing
services prioritising the most relevant areas; and
2) develop the materials and content covering the prioritized areas, including also the exams.
Substance-wise, as hinted previously, the main focus is to ensure that existing cloud
professionals upskill themselves, from knowing how to use the products from a single vendor
to a vendor-agnostic / multi-vendor skill set.
This contract has been estimated at EUR 750 000 and running for two years (2027-2028) with
the aim of developing 15 courses. For the drafting of the specifications as well as for the follow
up of the implementation of this study, it has been estimated 1 FTE of the Commission. An
additional cost at EUR 252 806 was estimated to establish a training accreditation mechanism.
One-off adjustment costs = external contract + (effort to draft the specifications * labour
cost)
One-off administrative costs for the European Commission. Here the costs to set up the
mechanism to ensure the quality of the training materials and the skills certifications are
ensured as well as the infrastructure for the training. This is estimated at 3 FTEs spanning for
a year.
Recurrent administrative costs for the European Commission stem from the operational costs
of running the training certification programme. This is estimated at 1 FTE.
Recurrent savings for the European Commission are the result of having a coordination
mechanism that may avoid duplication in the creation of a skills framework or training
materials. This has been estimated at 0.25 FTEs.
74 The most certificates recognized in the market include ISC2 CCSP (https://www.isc2.org/certifications/ccsp) and CompTIA
Cloud+ (https://www.comptia.org/en-us/certifications/cloud/) 75 DevOps is a philosophy that aims to close the gap between the development and operation teams involved in the development
and deployment of software applications, including cloud native and focusing on automating the procedures such as integration
and deployment. AIOps is a similar concept focusing on the development and operation of AI applications automating also the
data pipelines. Finally, FinOps is the means in which financial and development teams can jointly manage the expenditure
resulting from deploying services onto the cloud. The goal is to minimize the expense but maximize performance, quality and
velocity.
112
Recurrent administrative costs for cloud and AI experts: Training fees are assumed at EUR
100 per training.
Recurrent savings for cloud and AI computing service providers stem from the vendor-
neutrality of the proposed certificates. A person aiming to become cloud developer or a cloud
architect can obtain one vendor-neutral certification, containing the main principles and
patterns that can be applied to every technology with little effort in customisation or learning
the specificities of said technologies, instead of getting the certificates from all technologies
used. The aim is to avoid the multiplication of costs that currently developers incur by needing
to get certified in multiple technologies for their professional development (savings of 3 days
per course).
The vendor – neutral certification “Certified Cloud Security Professional CCSP”76, manages to
certify 3,000 – 3,500 people globally per year. AWS reports to have 1.05 million unique
individuals certified77. There is no similar available data for Google or Azure. The number of
potential beneficiaries of these certificates assumed for this analysis varies per year as follows:
1 000 experts in the first year and increasing at 10 000 after three years after which 300 experts
per year.
The assumption is that each certified person typically obtains around three certifications. This
reflects the average number of certificates an individual holds across different technologies or
providers.
Recurrent savings = number of people to get certified * (days saved per course * labour
costs) * number of certification courses followed by a person
Indirect savings: the effects over the economy of having a more cloud savvy and vendor-
agnostic workforce are indirect and discussed in sections 6.1.4 and 6.1.5.
Sensitivity analysis: The values to be challenged in this policy measure are as follows: 1) the
number of experts that can potentially be certified on a yearly basis ranging from 100 to 500
experts; 2) the certification fees ranging from EUR 50 to EUR 150; 3) the effort saved by opting
for a vendor-neutral certification i,e., the number of certifications from other technologies and
days saved; 4) the yearly coordination savings for the EC from avoiding duplication of effort
ranging from zero to EUR 21 067 per year and 5) the total reduced costs in the certification of
similar technologies from different providers stemming from the sensitivity introduced on the
number of experts that could potentially get certified.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. Benefits outweigh costs systematically across the average, min
and max scenarios. The min max ranges are characterised by the difference in the cost savings,
and more specifically in the certification of similar technologies from different providers
accounts for, as well as the number of people that can potentially be certified.
Table 38. Vendor neutral EU cloud/AI skill certificates (PM18)
Cost types Cloud & AI
providers (€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - - 252 806 252 806 252 806 252 806
76 https://destcert.com/resources/ccsp-certification-statistics 77 https://aws.amazon.com/certification/
113
recurrent administrative - - - - - -
one-off adjustment - - 834 269 834 269 834 269 834 269
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory
fees
285 906 - - 285 906 47 651 714 765
one-off enforcement. - - - - - -
recurrent enforcement - - 2 301 224 2 301 224 2 301 224 2 301 224
Total costs 285 906 - 3 388 299 3 674 205 3 435 950 4 103 064
Total benefits (78 001 031) - - (78 001 031) (18 812 153) (122 871 064)
Net impact (77 715 125) - 3 388 299 (74 326 825) (15 376 203) (118 786 000)
3.19. PM19: Mandatory specific award criteria for the procurement of cloud and
AI computing services
This policy measure has different components. First, with a view to reducing critical
dependencies, this measure makes the use of the non-price award criteria under PM16
mandatory.
Public procurement represents a significant share of the EU GDP (14%) and constitutes one of
the most powerful instruments available to public authorities to shape market dynamics.
Traditionally, procurement decisions in the cloud sector have been governed predominantly by
technical and financial criteria, such as different supported features, performance benchmarks,
level agreements, and price. While these remain important, they do not present a complete
picture of the sheer complexity of a cloud service. The conditions under which a service is
delivered, its supply chain, hardware and software, the innovation behind it and where it is
conducted, and who ultimately controls the whole chain of entities, are equally relevant,
particularly for the public sector.
The competitive effect of non-technical award criteria in cloud and AI computing services
procurement operates primarily through the demand side. When major public contracting
authorities, notably national authorities, and large public bodies, award points for supply chain,
innovation, security, interoperability, portability, and transparency, they send a clear market
signal about the criteria that matter for public sector contracts. This helps to create strong
incentives for providers to invest in the capabilities that score favourably, irrespective of any
regulatory mandate.
Non-technical award criteria in public procurement are not just a technical adjustment to
tendering practices. Used strategically and at scale, they can be an instrument of industrial and
digital policy, capable of redirecting public purchasing power towards outcomes that the
market, left to its own dynamics, would not spontaneously produce. In the cloud sector, where
market concentration, exposure to third country laws, and strategic dependency are pressing
concerns, the proposed criteria offer the mechanisms to strengthen European competitiveness,
build sovereign digital infrastructure, and ensure that the public sector leads, rather than
follows, the transition to a more open, resilient, and European digital ecosystem.
All the assumptions under PM16 remain valid, except in the adoption of the measure by
contracting authorities that in this policy measure applies to all of them and to all tenders of
cloud and AI computing services.
Administrative costs for national public authorities.
114
Plans. National public authorities would be required to draft and adopt a Cloud and AI
development plan, covering highly critical use cases involving the purchase of sovereign
services, including on how they intend to use public procurement to increase the uptake of
highly innovative cloud and AI computing services. The plans shall (i) assess the value of the
procurement contracts of sovereign cloud and AI computing services for critical uses cases; (ii)
identify priority areas where strategic autonomy is required; (iii) monitor the procurement of
innovative services, including those provided by SMEs and small-cap European providers; (iv)
ensure consistency with other existing or planned national strategies. To limit administrative
burden, the plan should remain concise and non-binding, focusing on priority areas, intended
procurement approaches and objectives. This activity is expected to generate one-off
administrative costs for public authorities, estimated as 3 FTEs per Member State (1-5 for
sensitivity purposes). Member States are also expected to face ongoing administrative costs to
update the strategies, estimated as 3 FTEs every 5 years.
As a follow-up, the uptake of such services, would be monitored through a centralised,
Commission-led monitoring exercise (see PM21). This would create synergies with other
measures within this Policy Option that are aimed at assessing the market presence of EU
providers and/or existing dependencies on third country providers with respect to specific
services (cfr. PM21). This activity would produce recurrent administrative costs for public
authorities estimated at approximately 5 FTEs (3-7 for sensitivity) per authority.
One-off adjustment costs for cloud and AI computing service providers. Providers who
choose to participate in procurement processes including the specific award criteria need to
align with the public procurement procedures for cloud and AI computing services. This
includes adopting and aligning their processes and improving them to meet these criteria, with
an estimated one-off effort of up to 20 staff days.
Recurrent administrative costs for cloud and AI computing service providers, are the costs
needed to adequate the offers, assumed to be 2 staff days per bid. These days are arbitrary but
were validated during the interviews as part of the supporting study. They assume only
administrative work for companies who would be eligible.
The cost is calculated multiplying this effort by the number of bids presented in a year and the
average salary of the cloud and AI computing service providers.
Recurrent administrative cost = number of bids presented * effort / bid * labour costs
Recurrent savings for cloud and AI computing service providers. Revenue gains are possible,
especially for providers who can differentiate due to the award criteria. These are however not
modelled.
Indirect savings: There are notable indirect benefits for cloud and AI computing service
providers that are able to comply with the different non-specific award criteria. For instance,
additional revenue gains, increase in reputation and the possibility to enlarge the scope of their
solutions which in turn may lead to an increase of sales.
Sensitivity analysis: The impact of this policy measure is sensitive to the number of public
authorities procuring cloud and AI computing services, the number of tenders and offers per
year and the number of cloud and AI computing service providers and the costs to adapt the
offers.
115
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. Total costs are estimated between EUR 104 million and EUR
226 million, consisting mainly of recurrent administrative costs related to coordination,
reporting, and compliance activities, alongside smaller one-off administrative and adjustment
costs. Total cost savings are limited, ranging from EUR 4 million to EUR 13 million. Costs
outweigh benefits across the average, min and max set of parameters. The difference in the min
and max is mainly caused by the number of CSP and AI providers participating in public
procurement every year across EU.
Table 45. Mandatory specific award criteria for cloud and AI computing service procurement (PM19)
Cost types Cloud & AI
providers (€)
Public
authorities (€)
European
Commissio
n (€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - 8 333 909 - 8 333 909 3 389 039 13 480 600
recurrent administrative 53 945 915 80 878 153 - 134 824 068 94 364 632 177 304 770
one-off adjustment 3 560 007 - - 3 560 007 3 560 007 13 708 048
recurrent adjustment - - - - - -
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 57 505 922 89 212092 - 146 7170984 101 313 678 204 493 419
Total benefits - (6 520 539) - (6 520 539) (4 347 026) (13 041 078)
Net impact 57 505 922 82 6912 523 -140 197 445 100 049 519 213 032 412
3.20. PM20: Boosting open source in public administrations
Open source software represents a strategic policy response to vendor lock-in and the excessive
dependence on vertically integrated non-EU providers. By broadening the available offer of
digital solutions, open source can bridge some of the gaps left by European solution providers
and reduce the public sector's reliance on large foreign technology companies for the delivery
of critical services. A wider adoption of open source would also create tangible competitive
advantages for EU companies, enabling them to leverage existing software building blocks and
recombine them efficiently to develop and bring new solutions to market at scale.
Open source solutions can equally reduce the costs associated with digital service delivery in
the public sector through use, reuse, and interoperability. When a solution is developed once
and reused across multiple public administrations throughout the EU, its fixed development
costs are amortised over a significantly larger user base. The creation of the EU Digital COVID
Certificates illustrated this principle clearly: it demonstrated both the critical importance of
interoperability for cross-border public services and the capacity of open source to accelerate
deployment at EU scale. That same case also showed that open source can be a powerful vector
of trust for public authorities and citizens alike.
Under this measure, public sector organisations shall take the appropriate measures to promote
the use of open standards and open source software when it is equivalent or superior to
proprietary one, in terms of functionalities, total cost, user-centricity, cybersecurity or other
relevant duly justified objective criteria.
Under the share and reuse of open source software principle, public sector organisations shall
consider the release of their software assets available for public use in an open source repository
connected to the EU Open Source (OSS) catalogue. To be able to do so, the public sector
116
organisation will have to hold the intellectual property rights. As part of the due diligence
required when releasing and deploying open source software, the terms of reference should be
expected to request both static and dynamic security analysis of the code, in order to mitigate
risks of interference by malicious actors exploiting the publicly accessible nature of the
codebase. This analysis shall be carried out by the service provider.
Public sector organisations shall not have to make their code publicly available in cases where
it is restricted for reasons of national security or defence, intellectual property rights held by
third parties, or due to technical or economic infeasibility.
These two principles are mutually reinforcing and should be read as complementary.
Encouraging the use of open source software reshapes how digital services are planned and
delivered, promoting the secure reuse of existing solutions and fostering collaboration across
institutional boundaries. “Share and Reuse” is expected to expand the pool of reusable,
shareable and secure software components. Together, these principles are expected to bring
about a gradual, structural transformation in the way software is commissioned, produced, and
maintained in the public sector.
The measure also includes the organisational and technical arrangements required to implement
these principles, such as the establishment of Open Source Programme Offices (OSPOs),
repositories for public-sector code, and shared infrastructure for collaboration, compliance, and
training. OSPOs, as part of their role in their technical, legal, procedural and strategic – related
tasks, may provide contracting authorities with advice on licenses, security, or maintenance
aspects. The creation and nurturing of a Network of OSPOs would also help in further
promoting, collaborating, and exchanging practices related to the use, reuse and sharing of open
source, through guidance, templates or recommendations thereby sharing expertise costs across
the public sector rather than requiring each OSPO to develop this capacity independently.
Overall, this measure could contribute to reduce dependencies on proprietary software products
and vendors, increase public-sector use, reuse and contribution to open source, and enable
efficiency and innovation gains across the European digital ecosystem. Furthermore, this
measure would stimulate broader market participation, encourage new entrants, reduce vendor
lock-in and improve the overall competitiveness of offers received by public authorities.
Finally, it would contribute to the growth of the open source services sector in Europe, which
is well positioned from increasing public sector demand for interoperable, auditable and
sovereignty solutions
Open source is already the backbone of cloud services and is used in all the different layers.
Without them, cloud services would not have reached the current level of technical maturity.
Some examples of open source used in cloud services include:
- Infrastructure as a Service: RedHat’s Open Stack is the base technology used by many
European providers for the provision of Infrastructure as a Service, which is later on
adjusted to accommodate to the needs of the CSPs. Open Stack allows to manage a pool
compute, storage, and networking resources to build and control private and public
clouds, acting as a cloud operating system.
- Data storage: Open source data bases such as MySQL, PostgreSQL or MongoDB are
heavily used by cloud service providers both for relational and NoSQL databases.
- Orchestration: An orchestrator allows to automate and coordinate complex tasks across
cloud environments, such as provisioning servers and storage, handling application
deployments in an automated way ensuring consistency, efficiency and security. The
117
golden standard solution used nowadays by most cloud service providers is Kubernetes,
originally developed by Google, but released as open source.
- Horizontal functionalities such as Identity and Access management as well as telemetry
are also heavily implemented through the use of open source solutions. For the first,
Keycloak is a relevant solution used, whereas for telemetry several solutions exist such
as Grafana (for visualization), Prometheus and the Elastic Stack (ELK) and are of
common practice in the industry.
- Development management lifecycle: A cloud service is mainly composed of software
components that need to be architected, developed, tested and deployed. The use of
open source solutions to manage the lifecycle of the service is heavily spread. For
instance, for the code management and versioning, the open source GitLab is steadily
growing against GitHub, owned by Microsoft, which dominates the market. Continuous
integration (CI) allows companies to automate the integration of the changes and new
developments made resulting in several release cycles a day. Several open source
solutions for this activity exist with Jenkins CI, Travis CI and Gitlab CI being the most
used. OpenAPI Swagger is also a good example of an open source solution that allows
to create Application Programming Interfaces (API) to foster interoperability within the
components of the application and with other applications.
The above are just some examples of how important open source is for the development and
provision of cloud services. Hyperscalers also use open source software as baseline for their
own solutions. As an example, AWS uses Open Telemetry78 as the underlying technology for
their solution AWS Distro79. Furthermore, cloud service providers, both non-EU (AWS,
Oracle, RedHat) and European ones (e.g. SAP, Telefónica, Aruba, Atos, Deutsche Telekom) are
large contributors to existing open source communities such as the Cloud Native Computing
Foundation, which encompasses relevant projects for cloud and edge.
One-off adjustment costs for national public authorities. National public authorities will incur
setup costs to establish open source repositories and the governance mechanisms defining
contribution, review, and acceptance procedures. This has been estimated at 2 FTE per MS.
Setting up repositories and understanding the challenges and governance mechanisms calls for
a fruitful and engaging community, the assumption is that 5 repositories per MS will be initially
created with an initial effort of 10 FTEs.
Nine Member States already maintain a national Open Source Programme Office or equivalent
coordination structure. The measure therefore focuses on strengthening governance and
coordination, aligning existing offices and governance mechanisms, and supporting those
Member States that do not yet have one. For modelling purposes an assumption of additional
12 OSPOs created is taken, with a set-up cost of approximately EUR 2m. The estimates assume
consistent institutional capacity across Member States, although in reality, OSPO maturity and
reuse potential vary widely.
One-off adjustment costs (NPA) = (number of OSPOs * set-up cost) + ((labour cost * effort
to set-up the repositories and guidance)) * MS
Recurrent adjustment costs for national public authorities. As software solutions developed
by the public authorities are released as open source, new repositories or projects in repositories
78 https://opentelemetry.io/ 79 https://aws-otel.github.io/
118
will have to be created, operated and maintained. The flagship initiative in this respect is
Developers Italia that has created 408 repositories in a GitHub repository since 201780,
averaging 50 a year but with different breadth and scope. One new repository will be created
every other year. This is a very conservative assumption.
Reusing and releasing source code open source incurs into additional costs as extra quality
assurance procedures must be put in place such as DevOps CI/CD pipelines, Static Application
Security Testing (SAST), Dynamic Application Security Testing (DAST), Software
composition analysis (SCA), vulnerability testing, as well processes continuously automated,
and generation of the software bill of materials (SBOM) for each piece of software released
and reused. This is estimated in 10 FTEs yearly per MS. The repositories shall have to be
continuously maintained and governed. The assumption is that the effort dedicated to these
activities increase in 1 FTE every other year, in a similar manner to the number of repositories.
For this impact assessment it is assumed that all repositories created along the timespan under
study will remain active. However, this is often not the situation, where repositories become
deprecated due to various circumstances such as low uptake and engagement.
Recurrent adjustment costs (NPA) = ((number of new repositories every other year * effort *
labour cost) * MS) + (cost of releasing open source code * MS)
One-off administrative costs for national public authorities stem from the adaptation of the
procurement templates to the notion of promoting the use of open source software, reusable
across the board for all bids. This is estimated in 30 days.
One-off administrative costs (NPA) = (effort to adapt templates * labour costs) * MS
Recurrent administrative for national public authorities will involve the operation of the
additional OSPOs accounted for (12, as a reference number), for a recurrent cost of
approximately EUR 1m per OSPO. However, certain institutional costs (e.g., existing OSPOs)
are considered as marginal coordination expenses rather than recalculated in full to avoid
double counting.
Recurrent administrative costs (NPA) = number of OSPOs * cost to maintain an OSPO
Recurrent savings for national public authorities.By reusing and adapting existing software
rather than commissioning bespoke solutions, administrations can reduce duplication of effort
and lower the total cost of ownership (TCO) of ICT projects. Empirical evidence81 82 and
national case studies suggest that reuse can reduce costs by 10–40 %, depending on maturity
and governance capacity.
Savings stem from the avoidance of duplication, the possibility to reuse existing solutions
increasing project efficiency and the decrease of the total cost of ownership. Reusing open-
source software solutions, both from other public authorities and available in open source
communities results in savings mainly through the avoidance of duplication of efforts but also
because the time-to-delivery of a software solution is also shortened. These have been estimated
as a percentage of an average contract value of custom-built mid-software development
projects, estimated at EUR 500 000. For public administrations it is assumed a cost decrease of
80 https://github.com/orgs/italia/repositories? 81 European Commission (2021). The impact of Open Source Software and Hardware on technological independence,
competitiveness and innovation in the EU economy 82 Frank Nagle (2019). Government Technology Policy, Social Value, and National Competitiveness. Harvard Business School
Working Paper No. 19-103.
119
25% due to avoidance of duplication. The rate of adoption is set to 5% for 2027 and then
increasing at 2% yearly until 2036.
Recurrent savings (NPA) = (Total cost of ownership * MS) + (savings for not duplicating the
software * (Percentage of public procurement development projects addressed with OS
solutions + Δ%) * number of tenders)
Strengths and limitations of the assumptions in this policy measure
There is only a small number of reputable studies and case examples that provide inputs for the
detailed parameters underpinning the assumptions, and where such evidence exists, it is often
too coarse or anecdotal to support a full analysis. Consequently, the main quantitative
assumptions draw on stakeholder consultations with industrial practitioners, open source
communities and public administrations, conducted through interviews in the supporting study.
Despite these constraints, open source has a clearly documented economic impact. At macro
level, a 2021 study for the European Commission estimated that in 2018 open source software
and hardware contributed between EUR 65 bn and EUR 95 bn to EU GDP. Stakeholder
consultations in the course of the study highlighted vendor lock-in as a major cost driver, with
public bodies becoming dependent on single, typically US based suppliers using de facto anti-
competitive practices, and one source estimating the resulting cost for European public
administrations at around EUR 1.1 bn per year.
Where cost savings are concerned, the literature and case material mainly qualify effects in
terms of total cost of ownership, reuse and productivity, which are reflected in the modelling
of this policy option, albeit public, detailed quantifications remain scarce.
As part of the supporting study, these assumptions were cross checked with established open
source organisations familiar with the costs of setting up OSPOs and the effort required to
release code under an open source licence, and with public administrations that are already
actively publishing open source software. In addition, the final validation workshop with
software integrators, public authorities and cloud service providers was used to test and confirm
several of the key assumptions.
Given the aforementioned uncertainties, this option is accompanied by a more attentive
sensitivity analysis than other measures, explicitly illustrating the range of possible outcomes
resulting from the variability in the main assumptions.
Sensitivity analysis: The parameters used for the sensitivity analysis include: 1) the number
of the initial repositories that MS put in place at an initial stage, ranging from 5 (min and central
value) to 25; 2) the effort to maintain and govern those repositories, with values moving
between 0.6 FTE to 1.2 FTE, with the central value being 1 FTE; 3) the costs to set up and
maintain the OSPOs, ranging respectively from EUR 2m to EUR 3m (central value is EUR 2m)
and EUR 8m to EUR 28m (central value is EUR 12m); 4) the effort and contract values of the
OS projects developed in and for the public administrations. Whereas the effort varies from 2
FTEs to 5 FTEs, the average contract value ranges from EUR 400 000 to EUR 600 000; 5) the
total cost of ownership (TCO) and 6) the savings for not duplicating software, ranging from
EUR 80 000 to 4500 00, depending on the size of the project
Indirect savings:
Open Source can have broader economic effects. When public administrations increase their
procurement and use of open source software, they create additional demand for open and
120
interoperable solutions across the wider economy. Evidence from France83 shows that a change
in public procurement rules led to measurable increases in open source activity, firm-level
adoption, and software-related employment. The same mechanism, applied at EU scale, would
be expected to stimulate market growth and innovation in related sectors. These spillover
effects are particularly relevant at regional level, where increased public sector use of open
source can strengthen local digital ecosystems and create opportunities for SMEs to participate
in public procurement through open interfaces and shared repositories.
Innovation and productivity. Increased public-sector participation in open source development
contributes directly to the European open source knowledge base. Studies 84,85 associate higher
open source contribution rates with stronger productivity growth, faster diffusion of digital
innovation, and broader SME participation in the digital economy.
Interaction model and mechanisms of impact.The policy measure operates through both
endogenous and exogenous mechanisms. Endogenously, the adoption of open source first
practices enhances the technological capability of public administrations. Reuse of existing
code reduces project complexity and risk, while exposure to open development practices
increases staff proficiency and institutional knowledge. This leads to lower operational costs
and greater autonomy in managing digital systems, strengthening technological sovereignty
and reducing dependency on specific vendors. Exogenously, wider public-sector use and
release of open source software generates spillover effects across the domestic IT ecosystem.
Local suppliers are better able to engage with open codebases and participate in collaborative
development, which can lower market entry barriers, increase competition, and improve the
quality of public-sector digital solutions. These dynamics reinforce each other: a more capable
public administration interacts with a more dynamic local IT environment, creating a self-
sustaining cycle of capability and innovation.
Results for all stakeholders aggregated and for the years covered by this measure are
summarized in the table below. Total implementation costs are driven largely by recurrent
adjustment costs related to setup of the OSPOs and the repositories as well as adaptation efforts.
Estimated cost savings range widely, from EUR 1.6 bn to EUR 12.0 bn, depending on the
realization of efficiency gains. The variation in the min and max costs from the perceived costs
of the proprietary software development tools along with the savings for not duplicating the
software needed.
Table 40. Boosting open source in public administrations (PM 20)
Cost types Cloud &
AI
providers
(€)
Public
authorities (€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative - 10 389 481 - 10 389 481 6 926 321 13 852 642
recurrent administrative - 90 711 949 - 90 711 949 60 474 632 211 661 213 one-off adjustment - 108 623 131 - 108 623 131 56 488 617 184 117 928
recurrent adjustment - 1 654 037 343 - 1 654 037 343 1 050 849 399 2 373 048 085
one-off regulatory fees - - - - - -
recurrent regulatory
fees
- - - - - -
one-off enforcement. - - - - - -
83 Orientations pour l'usage des logiciels libres dans l'administration 84 "Free and Open Source Software and Hardware." Harvard Business School Technical Note 724-380, September 2023 85 European Commission (2021). The impact of Open Source Software and Hardware on technological independence,
competitiveness and innovation in the EU economy.
121
recurrent enforcement - - - - - -
Total costs - 1 863 761 904 - 1 863 761 904 1 174 738 970 2 782 679 868
Total benefits - (5 251 290 954) -(5 251 290 954) (1 653 692 346) (11 955 210 083)
Net impact - (3 387 529 050) - (3 387 529 050) (478 953 376) (9 172 530 215)
3.21. PM21: Mandatory sovereign risk assessments for the use of cloud and AI
computing services in the public sector
Policy Measure 21 (PM21) consists of two elements. First, it turns what is only a
recommendation under policy measure 15 into a binding requirement for Member States.
Member States shall carry out at least one sovereignty risk assessment and repeat it at least
every four years or more frequently if deemed necessary. The purpose of the sovereignty risk
assessment is to identify which public sector use cases within a Member State require the use
of which sovereignty level as described under PM11. The sovereignty risk assessment would
assess, inter alia, the risks induced by the access to such data by a third-country authority or
third-country legal entity; or the risk of possible service disruption due to dependence on a
single or limited number of third-country services providers. On the basis of dedicated
discussions conducted with 3 different public authorities representing about 200 contracting
authorities, this assessment assumes that the matching of sovereignty levels to public sector
demand follows the following pattern: 70% of use cases would require a sovereignty level 1;
20% for level 2; 9% for level 3; and 1% for level 4. Even though the scheme is novel and does
not correspond to existing frameworks, this assessment fits with broad orders of magnitude that
can be inferred from existing analyses conducted in several Member States that have introduced
risk assessments for their public sector clouds, such as France, Poland86 or Italy87.
Critical use cases, defined as the use cases whose disruption would affect operational autonomy
or public order, correspond to use cases covered by level 2, 3 and 4. The risk assessment would
have to consider the reality of the supply market to avoid unrealistic outcomes, such as
mandating the use of services that don’t exist (yet) in the market.
To facilitate appropriate and coherent sovereignty risk assessments, the European Commission
would develop guidelines for Member States to conduct such assessments and provide a sample
risk assessment methodology (note that these guidelines concern the conduct of risk
assessments and differ from PM12, which consist in explaining the different levels of
sovereignty assurance). For Member States to have up-to-date information about the market
conditions of cloud and AI sovereign solutions, the Commission would also produce market
monitoring reports that will point Member States to possible gaps in the coverage of some
services.
The Member States would have to determine which public authorities are required to procure
specific levels of sovereignty and make this mandatory at national level, ensuring that
procurement aligns with the risk assessment, unless duly justified.
While PM11 only puts forward the definition of sovereignty levels, PM21 goes further by
putting forward a framework through which the respective levels of sovereignty can be
assessed.
Cloud service providers shall submit the relevant evidence that demonstrate that they comply
with the sovereignty assurance criteria to the designated competent authority. For assurance
86 See Cloud in Government Services 87 See Strategia Cloud Italia
122
level 1, the service provider may issue a ‘Statement of conformity’ where they show that they
comply with the cumulative criteria of this level. The cloud provider will make the statement
of conformity publicly available, and shall submit it, along with the necessary evidence, to the
designated competent authority. Cloud computing service providers qualifying as SMEs will
not be required to undergo the validation by the national competent authority.
The assessment of the compliance of the service against the criteria for sovereignty level 2-3-
4 will be performed through independent third party’s auditors. In this case, the service provider
will submit the audit report and the ‘positive’ audit opinion, along with the audited evidence,
to the competent authority of the country of establishment who shall verify them, without undue
delay and within 60 days. The result of this activity is a draft acceptance, request for more
information or rejection decision.
• In case the assessment is positive, the decision will be notified to the other Member
States, who have a 60 day review period to confirm the initial acceptance conclusions
by the MS of the country of establishment. During the period of review there is a
recourse possibility for the Member States. If no reasoned objection has been submitted
in the framed period, the validation of the audit report and audit opinion shall be
considered accepted by all Member States and the service shall be recognized across
the Union as a service provider that can participate in public procurement procedures
across the Union. On the other hand, if reasoned objections exist, the competent
authority shall assess them and decide on the maintenance or revocation of their original
decision. In cases of continued objection, the Commission shall assess it. This may
entail a request for additional information from the competent authorities. The
Commission decision shall be binding.
• In case the assessment yields an initial rejection, the competent authority will contact
the cloud provider and give the opportunity to the organization to provide comments on
the evaluation. The competent authority will take into consideration these comments.
• In the case where the competent authority assesses that the submitted evidence is not
sufficient, it will request the cloud service provider additional information to be
provided within a certain time limit, in order to be able to issue a decision.
The competent authorities should register the decision in a Union repository, maintained by the
Commission. The repository of sovereign cloud and AI computing services will be a public list
of audited sovereign cloud and AI computing services that verifiably comply with the
sovereignty requirements. The benefits are for providers and users alike: providers will enhance
their visibility and users their market research.
To cater for market evolutions, the sovereignty criteria of all levels and evaluation evidence
proposed, but not limited to, would be modifiable by comitology. This evaluation evidence
would help independent third party auditors in their assessment of the service and ensure full
harmonisation in the way different auditors conduct their assessment and for Member States to
ensure that the procedures have been followed.
The assessment would be periodically renewable following the same evaluation methodology.
123
Policy measure 21 implies that the Commission sets up and maintains a repository of the
services audited against level 2-3-4, and that public authorities assume the actual transfer cost
to change from a non-sovereign service to a sovereign service (cloud porting).
This measure is primarily designed to contribute to the protection of public order by enhancing
the resilience in the public sector, which is SO4. Nevertheless, European providers would face
less costs and efforts to meet sovereignty conditions. When it comes to meeting the criteria to
demonstrate sovereignty level 1 and 2, EU providers can more readily substantiate that they are
not affected by third-country policies affecting data access or limiting service continuity. As
well, level 3 and level 4 sovereignty can only be served by service providers owned and
controlled by EU entities. This implies that PM15 will also contribute to decreasing the overall
reliance on non-European cloud and AI computing services, which is SO3.
Second, private sector essential entities listed under Annex I of NIS 2 are encouraged to
integrate into the cybersecurity risk assessment they already conduct, the assessment of the
risks stemming from their use of cloud and AI computing services. This includes an analysis
of the laws applicable to the computing service and the extraterritorial reach of such laws; the
risks associated to the possible unauthorised third country government access to and transfer
of data; the risks associated to service continuity and quality; the operational dependency and
possible loss of autonomy.
One-off adjustment costs for the European Commission
Setting up and maintaining the repository of sovereign cloud and AI computing services under
the sovereignty levels. These costs include the drafting of the tender specifications for an
external provider to build, develop and maintain the repository plus additional time for the
evaluation and the definition of their governance procedures. This has been estimated to 120
staff days. The repository will be developed by an external contractor with a contract value of
EUR 500 000.
One-off adjustment costs (EC) = (time to draft tender specifications for the repository) *
labour costs + external contract to develop the repository
Guidelines for MS to perform the sovereignty risk assessments. In order to support Member
States in the implementation of their sovereignty risk assessments, the Commission will publish
guidelines. These guidelines would include the process to carry out the risk assessment, the
criteria that would allow them to identify the scope, for instance by means of use cases and / or
applications, the data classification process, where the market surveillance or information on
sovereign services can be found. The estimated effort for the development and publication of
these guidelines is 2 FTEs over 1 year.
Recurrent adjustment costs for the European Commission.
Maintenance of the repository for sovereign services. Another important activity is related to
the maintenance of the repository of audited services and adjacent activities, including:
• Yearly hosting costs of the repository, estimated to be of EUR 200 00088. The repository
will be deployed on a sovereign cloud at the Commissions’ premisses and telemetry
and FinOps procedures will be put in place to control expenses.
88 Estimation based on the values of Simpl
124
• Yearly maintenance of the repository, performed by an external contractor, includes the
upgrades of the software libraries amidst identified vulnerabilities, correction of errors
and bugs and additional features. The estimated effort assumed for this is 2 staff days
on a weekly basis.
• Oversight of the repository, including the follow-up of the contractors’ activities,
estimated in 2 staff days per month by the Commission staff.
Recurrent administrative costs (EC) = costs of the yearly hosting + external contract for the
yearly maintenance of the repository + (number of days for the oversight of the repository *
labour costs)
Market monitoring. In order to assess the status of the cloud and AI computing services market,
the Commission will run a study for the first three years after the adoption of CADA. The goal
is twofold. On one hand, it will provide input for the dependency analysis to be carried out and
on the other, it will provide a deeper knowledge of the status and evolution of the cloud and AI
computing services to serve for the procurement of sovereign cloud and AI services (see
PM19). The limited timespan of the study is due to the following rationale: while initially this
will be of importance due to, in some cases, low knowledge of existing offers, the repository
and the catalogue of sovereign services would allow for a wider and deeper knowledge of the
market by the contracting authorities.
This study is estimated to cost EUR 300 000 / year for three years.
One-off adjustment costs for national public authorities.
Risk assessments. National public authorities and regional or federal authorities in the case of
decentralized Member States are mandated to perform a sovereignty risk assessment, with the
aim of mapping sovereignty assurance levels to cloud and AI computing services used in the
public sector, taking into consideration aspects such as sensitivity and criticality of the data
processed, risk associated to the potential unlawful access to data by third country legislation,
potential disruption of the services, the service type based on a taxonomy that the Commission
will publish as part of their guidance, existence of cloud and AI computing services audited
under that sovereignty level, among others. The estimated number of authorities is 267,
considering the NUTS-0 and NUTS-1 distribution.
The output of this result assessment will be a classification of applications or use cases
categories mapped to the minimum sovereignty assurance level permitted for the procurement
of cloud and AI services serving those cases.
An important element of the sovereignty risk assessment is that the existence of sovereign
commercial offers will be considered as part of the exercise. This will allow to not request
offers that are not available in the market or that would cause disproportionate costs to the
contracting authorities as part of the procurement process.
The estimated effort for this sovereignty risk assessment is of 10 FTEs. This is a new activity
for which there is yet few expertise in the Member States. It is considered that in the initial
iteration, Member States will need to understand well the sectors, applications and type of data,
among other aspects, that will drive their risk appetite to request one sovereignty level or
another. It is expected that initially this will be an activity encompassing various ministries and
agencies and therefore an intensive collaboration is expected. The guidelines from the
Commission, however, should alleviate this effort.
The risk assessments will start in 2029, once EUCS and the sovereignty scheme are in place.
125
Designation of the competent authority, which has an estimated of 0.2 FTE per MS, which will
be responsible to verify the audit reports received from the providers that had their services
audited at sovereign level 2 at least.
Costs of migration to sovereign cloud by the public sector. Detailed information is provided in
Annex 12 on the methodology and calculation of the costs for the migration and porting of
cloud applications to sovereign cloud services.
Both the migration of the legacy-to-sovereign cloud and cloud-to-sovereign cloud applications
will be performed in a linear distributed manner in a time frame spanning 5 years, starting in
2029, once there is a set of sovereign audited services.
The average cost of porting a single application to a new cloud varies from EUR 30 000 and
EUR 600 000 depending on the size of the application, and not much on the type of cloud at
origin or destination, as they are mainly based on human effort (400 to 8000 hours of work).
See Annex 12 for an explanation of these figures.
These costs are illustrated with the real use cases in Section 6.1.2 of the main text but have not
been quantified in the cost-benefit analysis, although they are described in detail in Annex 12.
Since they represent planned expenditure that would be incurred in the future as part of the
regular cloud contract renewal cycles and independent of the present policy measure, they have
been considered outside of the cost-benefit so as not to conflate structural renewal costs with
the incremental financial impact of the measure itself.
Recurrent administrative costs for national public authorities.
Sovereignty risk assessment. This risk assessment shall be renewed every 4 years. The effort
estimated for this reassessment of the sovereignty risks is estimated to be 5 FTEs, a lower effort
than implementing the risk assessment from scratch, given that the activity is already ramped
up and the Member States would already have a good understanding of their situation.
Revision of the audit reports by the competent authorities, estimated at 5 days per service
audited at sovereign level 2-4. After verifying the audit report, opinion and evidence and
consulting with the other competent authorities, the evaluating competent authority will adopt
a decision that would allow a service provider to participate in public procurement activities.
Recurrent savings for national public authorities
Running applications on the cloud. Operating applications on the cloud, if performed well, with
a continuous FinOps verifying that services are not overprovisioned can lead to significant
savings. Literature sources estimate that the total cost of ownership (TCO) savings can amount
to 20-50% in the case of legacy-to-cloud migrations89. No savings are here accounted for in
the operation of applications migrated from cloud-to-sovereign cloud, as the applications would
have already been benefitting from the total cost of ownership savings.
Simpler public procurement. Savings in this respect stem from the simplification of
systematically using procuring audited services, which are part of the repository of sovereign
cloud and AI computing services. This shall allow public administrations to save time in the
verification of the documentation for the evaluation of the offer. This will result in 2 staff days
89 Chatzithanasis & Michalakelis (2018) estimates 24% to 50% The Benefits of Cloud Computing:. Ali Khajeh-Hosseini, David
Greenwood, Ian Sommerville
126
per bid, and assuming that there are 3.5 potential contractors presenting bids in each tender
procedure.
Recurrent savings = Average number of yearly service bids requiring sovereignty assurance
level 2 – 4 * staff costs savings per bid
Sovereign services. The savings from using sovereign services cannot be quantified due to the
intangible nature of sovereignty and resilience.
Costs for service providers to develop sovereign services
Assessing the cost and benefits for providers to provide sovereign services is a complex task
that involves many parameters and differs greatly from provider to provider. As well, in the
absence of an established market for sovereign services, data sources are rather anchored in
providers’ business plans, not in ex post analysis of established businesses. Such data is
confidential to companies, and the below considerations are based on targeted discussions with
stakeholders that requested to remain anonymous. A first consideration is that the consulted
companies unanimously indicated that, in developing the business plan for these new services,
they count on new large critical use cases that are today not in the cloud; in other words, they
see sovereign services to generate a new source of income, but not to substitute existing.
Cost wise, new costs notably include the amortisation of all one-off adjustment costs such as
the additional cost induced by using EU-located infrastructure, the additional compliance costs
induced by the audits, the additional costs of being certified under EUCS; and proper recurrent
costs such as the higher salaries of employing EU workforce. As an illustration, speaking under
the condition of anonymity for business secrecy reasons, one of the largest EU service providers
of sovereign services speaks of an overall investment of EUR 1.5 bn, including hardware, for
a broad range of IaaS and PaaS services (for an unspecified computing capacity). Conversely,
another EU service provider with an established range of non-sovereign services speaks of an
overall investment in the range of EUR 20-40 m to adjust existing hardware and software to
the stricter norms that a sovereign service entails, with plans to invest progressively should the
market develops.
The benefits for service providers to develop sovereign services are covered under section
6.1.5.
Benefit wise, cloudifying legacy on-premises services means new revenues. As to the move
from traditional cloud to sovereign cloud, today’s few available sovereign services come with
a mark-up which is still too uncertain to draw conclusions (see discussion under Problem Driver
4 in 2.3.4). To this additional income, consulted EU service providers see also sovereign
services as niche market where they have a demonstrable added value that could spill over to
other market segments. The consulted companies unanimously indicated that it is too early to
confirm whether, over the span of the next ten years, the incremental cost and benefits would
reach the same balance (i.e. margins) as with equivalent non-sovereign services, which is today
commonly assumed to be around 30%90; to cater for this uncertainty, this assessment assumes
that stakeholders’ margins vary from 25% to 35%.
90 Amazon Web Services profits squeezed as AI arms race drives spending surge – GeekWire
127
One-off adjustment costs for cloud an AI computing service providers
Ultimately, the decision of a provider to subject to the sovereignty assessment one or more
services is their own business decision. However, in order to have a sense of magnitude of how
many providers could be affected by the sovereignty scheme and based on extensive desk
research conducted as part of the preparatory study (Technopolis et al., 2025), it was identified
that services from 59 non-EU headquartered cloud service providers meet Level 1 requirements
and would be able to qualify to Level 2 should they decide so. 226 EU headquartered providers
could qualify their services as Level 2 and would be able to qualify their services as Level 3
should they decide so. If only large companies are considered and SMEs are excluded, the
figure goes down to 59 EU headquartered large companies.
The assumptions are as follows: 600 services will be audited from 2029 until 2034, with the
assumption of 30 audits the first year, and an annual growth of 82%. This builds over the
number of services currently qualified in comparable national qualification and certification
schemes91, and the FedRAMP mechanism. As of September 2025. the FedRAMP repository
contains 530 authorised services under level moderate and high.
Audit: Services may be audited under different assurance levels, provided that the criteria
mentioned above (see PM11) are met. For the purpose of this impact assessment, it is estimated
that the proportion of services that get a positive audit report across the different assurance
levels will be mirroring the needs of the public authorities (see below). This would then make
that out of the estimated 600 sovereign cloud services 92:
• 70% of the services will be Sovereignty Assurance Level 1, making 420 services
• 20% of the services will be Sovereignty Assurance Level 2, resulting 120 services
• 9% of the services will be Sovereignty Assurance Level 3, amounting to 54 services
• 1% of the services will be Sovereignty Assurance Level 4, resulting in 6 services.
The effort needed to get audited as sovereign under the sovereignty assurance level 2 -4 is
estimated to be 15 FTEs. This includes the definition and implementation of the necessary
legal, organisational and technical measures to reach sovereignty assurance level 2 – 4 and the
auditing procedure carried by the third party auditor. This effort was validated in the Final
Validation workshop held as part of the study led by Technopolis.
One-off adjustment cost of new audits (CSAP) = (effort to get audited sovereignty assurance
level 2 – 4 * labour costs) * number of audited services assurance level 2 – 4
One-off regulatory fee for cloud and AI computing service providers
New audits, the fee is assumed to be EUR 20 000, for the estimated number of services awarded
sovereignty assurance level 2– 4.
One off regulatory fee for new audits (CSAP) = regulatory fee * number of services
sovereignty level 2 – 4
Recurrent regulatory fee for cloud and AI computing service providers
Renewal of audits: fee is assumed to be EUR 14 000..
91 For this impact assessment ES, IT, and FR have been considered as baseline as their data is public.
(92) see footnote 90
128
Recurrent regulatory fee for new audits (CSAP) = regulatory fee * number of audited
services assurance level 2 – 4
Recurrent administrative costs for cloud and AI computing service providers
Audits: These compliance costs stem from the audits . to maintain the positive audit report,
where the goal is to verify that the conditions under which the sovereignty level was obtained
remain valid (effort estimated as 3 FTE/year per service). Compared to the effort needed to
achieve the award of a sovereignty level, the effort dedicated to the intermediate audits has
been assumed to be lower than obtaining the initial qualification, in spite of the annual reporting
obligations set out above.
Recurrent administrative compliance cost of sovereignty level renewal = (time to be audited
* labour costs) * number of services sovereignty level 2 - 4
Recurrent savings for cloud and AI computing service providers
Audits.A crucial matter in the analysis of the recurrent savings for cloud and AI computing
service providers is the single market effect especially considering the mandatory nature of this
policy measure. Providers will be able to audit their services in one MS and be able to offer
them in public procurement procedures in all EU-27. The analysis assumes that in the business-
as-usual scenario one cloud service provider would seek to run the audit for one service in 10
MS to be on the conservative side. Given the costs of compliance, in a business-as-usual case,
a cloud and AI computing service provider would carefully prioritize in which MS(s) it would
seek to obtain a similar audit of a service usually taking into consideration aspects such as the
market share or other commercial opportunities.
Saving costs audit valid throughout the EU = Number of MS where the Cloud and Service
Providers would get audited * cost to get audited * number of audited services assurance
level 2 - 4
Public procurement. Another relevant saving stems from the fact of not having to collect and
resubmit all the evidence necessary to demonstrate that the criteria are fulfilled for each bid,
given that the certified services are listed in the repository and accessible for the contracting
authority. The assumption is that undertakings will save 2 staff days / year per service and
bid. The number of yearly bids amounts to 50 a year.
Business costs saved: = Businesses time saved per evidence submission in a tender procedure
* number of submitted bids
Recurrent administrative costs for the private sector (Auditors)
Third party audits. In this PM, as mentioned above, the target number of cloud and AI
computing services audited as sovereign is estimated to reach 600 by 203293, whereas 30 of
these will reach sovereign level 2-4 during the first year. The assumed CAGR is 82%. The
assumption under this policy measure follows a linear progressive uptake of the audit scheme.
The estimated average effort for auditing a service under Sovereignty assurance level 2-4 is
110 days / audit. The same effort is considered for the renewal process.
93 Assumption taken considering the current valid authorised cloud services in FedRAMP moderate and high, which amounts
to 531 as of 26 September 2025.
129
Recurrent administrative costs for the private sector (Private sector essential entities
operating in sector listed under NIS2 Annex I)
Risk assessment. These are the costs incurred by essential private sector entities listed as under
annex I of NIS2 to sovereignty related risks in their current risk assessments. The effort
estimated is 60 staff days per essential entity for the first year and 12 staff days for the following
ones to account that the novelty generates larger work that re-doing it.
The SWD published as part of the proposal for the Digital Omnibus 94 estimates that the number
of all entities, essential and critical, from the private and public sector, affected by NIS2 Annex
I amounts to 160 000. France's NIS2 available data, indicates that 20% of its entities are
classified as essential entities under Annex I, with the remaining 80% designated as important
entities under Annex II95. Re-using the same ratio for the EU implies 160 000 * 20% = 32 000.
To which the private sector is assumed to represent 80%, or 32 000 * 80% = 25 600.
Indirect savings: Similarly to PM15, through this measure, cloud and AI computing services
able to adequate their processes to comply with the audit scheme will increase trust from public
administrations, and notably for the use cases identified as need-to-be sovereign based on the
sovereign risk assessments performed. This will result in more trustworthy, credible and very
specialized as well as very advanced offers responding to challenges such as resilience and
autonomy. Given the mandatory nature of this PM, the indirect savings are expected to be larger
in this measure than in PM15.
Sensitivity analysis: The most affected parameters for this Policy Measure are as follows: 1)
the effort needed by a cloud and AI computing service provider to obtain the audit; 2) the
number of MS where cloud and AI computing services would seek the audit in the absence of
a single market; 3) the initial number of cloud and AI computing services audited during the
first year and the annual growth; 4) the number of submitted bids per year; 5) effort by the
private sector for the risk assessments, 6) effort by public authorities for the risk assessments.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. The main source of variation in the min and max comes from
the cost savings, and notably the savings from the EU-wide validity per service for cloud and
AI computing service providers and the number of countries where providers would
audited/validated, in the event the scheme would not exist.
Table 41. Mandatory sovereign risk assessment for the use of cloud and AI computing services in the
public sector (PM21)
Cost types Private sector Cloud & AI
providers (€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off
administrative
- - - - - - -
recurrent
administrative
1 439 144 597 80 856 542 562 624 1 520 563 763 540 444 577 2 763 406 090
one-off adjustment - 254 272 125 154 624 928 693 692 409 590 745 247 443 239 571 738 249
recurrent
adjustment
- - 68 603 894 2 965 857 71 569 751 44 128 193 44 128 193
one-off regulatory
fees
- 3 022 890 - - 3 022 890 3 022 890 3 022 890
94 SWD(2025)836 final 95 Cybersécurité : la transposition de la directive NIS 2 examinée le 11 mars au Sénat
130
recurrent
regulatory fees
- 3 367 639 - - 3 367 639 3 367 639 3 367 639
one-off
enforcement.
- - - - - - -
recurrent
enforcement
- - - - - -
Total costs 1 439 144 597 341 519 196 223 791 446 3 659 5492 008 114 787 838 406 538 3 385 663 062
Total benefits - (3 349 440 466) (2 507 623) -(3 351 948 089) (1 676 992 882) (5 085 107 859)
Net impact 1 439 144 597 (3 007 921 270) 221 283 823 3 659 549 (1 343 833 301) (838 586 344) (1 699 444 797)
This policy measure would trigger an acceleration of porting of some applications to sovereign
cloud services of assurance levels 2 to 4, requiring an anticipation of expenses of EUR 3 to 15
bn. These costs are estimated and reported under section 12.4 of Annex 4. Since they represent
planned expenditure that would be incurred in the future as part of the regular cloud contract
renewal cycles and independent of the present policy measure, they have been scoped outside
of the summary cost table so as not to conflate structural renewal costs with the incremental
financial impact of the measure itself.
The benefits accruing from the risk reduction, notably an increased autonomy and strengthened
operational resilience through the adoption of sovereign cloud and AI computing solutions are
intangible and cannot be expressed in quantitative terms. While these benefits cannot be
quantified, they are acknowledged as a significant consideration underpinning the policy
rationale.
3.22. PM22: EU-level Procurement of cloud and AI computing services
PM22 builds on top of PM17, establishing in addition a framework for the public procurement
at EU-level of data centre, cloud and AI computing services.
Cloud Capacity in the Public Sector. Same assumptions as in PM17 are taken for all impacts
related to the public sector cloud federation. Same number of public authorities at all levels
participate in the public sector cloud federation.
Joint Procurement. The Union procured 2,5 trillion EUR in services in 2024, from 250.000
public authorities96. Section 2.4.3 (“Public procurement contract values for cloud and AI
computing services”) of this annex shows that cloud and AI computing services represented
0,68% of the total services procured by the public sector in Europe in 2024. The value of cloud
and AI contracts procured by the public sector in that year (0.68% x EUR 2,5 trillion = EUR
17 bn) is projected to the future using the CAGR observed in the period 2021-24 (20%) for the
first years up to 2030 and a half of that from then on.
A progressive roll-out of the EU-level joint procurement is considered, starting at 2% of the
procured service volume (EUR 340 m) and increasing by 2 additional percentage points every
year, reaching 22% at the end of the 10-year period (EUR 21 bn). The level of savings will also
increase over time as higher volumes are aggregated, and experience is gained, starting at 10%
and growing annually by 4 additional percentage points till 2032, reaching 30% in that year.
One-off adjustment costs for the European Commission include those related to set up the
platform for federation, procurement, brokering and service aggregation. These costs include 4
persons half time in 2027 (2 FTE) from the Commission to run the procurement process for the
96 From Public Procurement Data Space: https://www.public-procurement-data-space.europa.eu/en
131
platform (define requirements and prepare the call, assess and refine the proposals and award
the contract), that will be tasked to a subcontractor for development and deployment at an
estimated cost of EUR 40 000 000. This figure represents additional EUR 20 000 000 with
respect to the federation platform considered in PM17 to cover the service repository, procured
service catalogue and brokering and service aggregation functionality.
One-off adjustment costs = Platform costs + EC Staff costs for platform procurement
Recurrent adjustment costs for the European Commission relate to hosting the platform in a
cloud, estimated at 1.5 million EUR per year, operating, maintaining and evolving the
federation platform, with an effort of 100 contractor FTEs per year (40 additional to those
required in PM17 to cover the extended functionality). The Simpl programme has been used as
a reference for the underlying costs assumption97 .
Recurrent adjustment costs = Hosting + Maintenance & Integration contractor costs
Recurrent administrative costs for the European Commission, stem from the costs to manage
the cloud federation and the operation of the joint procurement process. In addition to the
resources of PM17 to set up the public sector cloud federation, a further 12 FTEs are dedicated
in 2027 to design and setup the joint procurement framework.
On a yearly basis, in addition to the resources estimated in PM17 to manage the cloud
federation, a further 20 FTEs are dedicated to run the EU-level procurement processes.
Recurrent administrative costs for the European Commission =
costs to manage the federation + operation of the joint procurement process
One-off and recurrent adjustment costs for national public authorities. The Public
Authorities dedicate the same effort as in PM17 to participate in the setup of the public sector
cloud federation, and to connect the infrastructures and keep updated the characteristics of the
shared resources in the federation platform.
Recurrent savings for national public authorities. Recurrent savings from federation, coming
from reducing idle capacity, getting cheaper computing resources and reducing the
coordination effort for sharing cloud, are the same as in PM17.
With regards to the procurement of cloud and AI computing services, it is assumed that each
Public Authority dedicates an average of 3 FTEs. Thanks to joint procurement, 0.5 FTEs are
assumed to be saved in the procurement process in each of the Public Authorities. The demand
aggregation and service composition provided by the service platform will save additional 0.5
FTEs per Public Authority, coming from buying more efficient pre-integrated solutions.
The participation in the joint procurement framework and the aggregation of demand in the
common EU-level procurement process will produce 2 effects: a reduction in the FTEs required
for procuring cloud and AI computing services (see above) and a reduction in service prices
coming from higher scale, purchase power.
97Simpl: Cloud-to-edge federations empowering EU data spaces
132
Recurrent savings = saved effort in PAs x NPA staff cost x number of PAs + cloud and AI
computing service procurement volume x % procured at EU-level x % discount
The share of cloud and AI computing services procured at EU-level and the discount achieved
by aggregating demand at EU-level grow over the years at the rate described above in this
section.
Indirect savings: The increase in public procurement of cloud and AI sales volume that is
addressable by European providers may help to increase their scale and competitiveness and
their share in the European and global cloud and AI markets. Joint procurement and federation
can also help in accelerating cloud and AI adoption in public administrations thanks to the
reduction in service costs.
Sensitivity analysis: The results have been challenged in terms of savings for national public
authorities namely: 1) Yearly administrative time saved thanks to fewer coordination needs for
national public authorities; 2) FTEs saved thanks to joint procurement – yearly; 3) Savings in
terms of effort incurred by participating PAs due to the possibility to aggregate demand and
compose services - yearly; Considering IaaS + PaaS + SaaS, deployment automation. Midsized
application. All values were set to range between a min of 0.1 to a max 0.5 FTEs corresponding
to a min of 22 days to a max of 110 days; 4) capacity shared by each MS with the federation
ranging from 5% to 15% which impacts the savings in terms of idle capacity reduction.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below. In all scenarios including the average, min and max savings
outweigh costs. The main source of variation in the min and max comes from the cost savings,
and notably the savings in terms of idle capacity reduction.
Table 42. EU-level Joint cloud and AI Procurement (PM22)
Cost types Cloud &
AI
providers
(€)
Public authorities
(€)
European
Commission (€)
Total Value
(central, €)
Total Value (min,
€)
Total Value (max,
€)
one-off administrative - - - - - -
recurrent administrative - 18 689 579 18 689 579 18 689 579 18 689 579
one-off adjustment - 4 613 563 40 009 806 44 623 369 44 623 369 44 623 369
recurrent adjustment - 213 669 483
163 981 548 377 651 032
334 125 766 421 176 297
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement - - - - - -
recurrent enforcement - - - - - -
Total costs - 218 283 046
222 680 933 440 963 980 397 438 715 484 489 245
Total benefits - (34 199 254 362) - (34 199 254 362) (19 170 613 237) (49 227 895 486)
Net impact - (33 980 971 315) 222 680 933 (33 758 290 382) (18 773 174 523) (48 743 406 241)
3.23. PM23: Financial support for SMEs to adopt cloud and AI
This policy measure puts forward a targeted scheme to provide financial support to SMEs for
adopting cloud and AI computing services to increase their productivity and competitiveness. A
yearly budget of EUR 40 050 000 is considered over the 10-year period. This is done for
modelling purposes only and does not pre-empt the next MFF:
133
• Most of it, EUR 38 500 000, is dedicated to supporting the design and planning phase of
cloud and AI-based transformation projects for SMEs.
o The grants are fixed amounts that the SMEs can spend in consultancy services to design
digital transformation projects based on cloud and AI technology.
o The amount will depend on the size of the SME: small SMEs (10 to 50 employees) will
receive EUR 10 000 and midsize SMEs (50 to 250 employees) EUR 25 000.
o The grants can be spent in services from a catalogue of self-published organizations
maintained by the Commission that will include consultancy service providers from the
IT sector in all MS. The catalogue would be populated based on an objective and
transparent process.
o The objective is reaching 2% of the small (10 to 50 employees) and midsize (50 to 250
employees) SMEs over the 10-year period.
• Every year, the SMEs that have designed the most innovative cloud and AI-based
transformation project plans will receive additional support to fund their implementation:
o small SMEs would receive EUR 100 000 each.
o midsize SMEs would receive EUR 250 000 each.
o Total yearly budget for awards would be EUR 1 050 000.
o A jury will assess the project plans designed every year and select the ones to be
awarded.
• The rest, EUR 500 000, serves to communicate and create awareness of the programme,
and disseminate the results, giving visibility to the most successful and innovative project
to serve as relevant references that may inspire SMEs across the different sectors and MSs:
o The awards will be communicated and made visible through reports, social media, and
events, and the project descriptions will be made available in a web site to serve as
references.
o The awards will reward innovation and its transformation impact in terms of efficiency,
productivity and competitiveness.
o There will be a balanced distribution of the awards among sectors and MSs, so that
relevant references are created across all markets.
o An SME project assessment team will define specific criteria for awarding the grants
and select the awarded SMEs out of the presented proposals. The calculation of the
evaluation efforts considers that there are 5 proposals presented for each awarded grant.
The SME Performance Review report shows that there are 26.1 million SMEs in Europe, out
of which 24.5 million are micro enterprises (less than 10 employees), 1.4 million are small
enterprises (10 to 50 employees) and 214,000 are midsize companies (50 to 250 employees).
The cloud adoption scheme applies to small and midsize enterprises, with the objective of
reaching 2% of each category in the 10-year period. This means a total of 32 200 companies
engaged, 28 000 small ones and 4 200 midsize ones, will be direct beneficiaries.
The awarded SMEs will use the award to contract consultancy services from a catalogue
provided by the Commission that will contain service providers from different MS and diverse
profiles. These providers will be mostly local IT SMEs with skills in cloud and AI technology,
that will be indirect beneficiaries from this policy measure.
Eurostat reports 1 281 000 businesses in the ICT sector2, out of which 820 800 are dedicated
to computing, consultancy and related services. From Eurostat’s database3, the number of micro
companies (less than 10 employees) in the ICT sector is 1 220 000 (as of 2020), 53 000 are
small (10 to 50 employees) and 12.000 are medium size (50 to 250). The catalogue of digital
consultancy service providers will contain some of the European small and midsize IT
134
enterprises (66.8% of the ICT sector, 43 420 companies in EU) that have the right skills and
resources. Assuming that each service provider manages 5 consultancy projects along the 10-
year period, the policy measure could indirectly benefit 6 440 IT SMEs that would support the
awarded SMEs in their project design and planning.
One-off and recurrent administrative costs in the European Commission. The Commission
will apply an effort of 1 FTE for the design and set up of the cloud and AI adoption programme,
requiring consultation with stakeholders (Member States, associations and SME
representatives) and the elaboration of the programme description. The infrastructure required
to run the programme, manage the calls and monitor its execution would re-use existing
infrastructure and is accounted for here. The cost of the catalogue and the onboarding of IT
consultancy service providers are considered negligible.
On a recurrent basis, the Commission will need 100 staff days to administer the cloud and AI
adoption programme, including the preparation, launch, management and closing of the
different calls, and 0.6 staff days to assess each of the proposals presented. The grants will
amount for EUR 10 000 and EUR 25 000 for small and mid-sized SMEs respectively.
One-off administrative costs in SMEs. The effort to prepare a project proposal is estimated to
be 2 persons for 10 days (20 staff days). A sensitivity check for this parameter is accounted for.
Cost savings in SMEs. The benefits stem from the design of the projects supporting the AI-
based transformation. The percentage of small and mid-sized companies benefitting
respectively from the EUR 10 000 and EUR 25 000 grant is 2%.
Indirect savings: the expected increase in the adoption of cloud and AI computing services in
European SMEs from different sectors resulting from this initiative is discussed in section 2.5.5
of this annex.
Sensitivity analysis: The policy measure depends mainly on the targeted reach (% of benefitted
SME).
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below.
Table 43. Financial support for SMEs to adopt cloud and AI (PM23)
Cost types SMEs (€) Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value
(max, €)
one-off administrative 55 373 982 - 84 269 55 458 251 27 725 295 83 443 215
recurrent administrative - - - - - -
one-off adjustment - - - - - -
recurrent adjustment - - 339 317 286 339 317 286 171 954 564 509 704 787
one-off regulatory fees - - - -
recurrent regulatory
fees
- - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs 55 373 982 - 339 401 554 394 775 537 199 679 859 593 148 002
Total benefits (328 412 809) - - (328 412 809) (164 206 405) (495 604 785)
Net impact (273 038 827) - 339 401 554 66 362 727 35 473 454 97 543 217
135
3.24. PM24: Tools to enrich EU cloud and AI computing services offering
This policy measure sets up actions to improve the competitiveness of European cloud and AI
computing services by facilitating tools to improve their service offer and meet the demand for
cloud and AI computing services in different sectors. The aim is to a) promote the development
of open source tools that allow to combine, compose and integrate diverse functionality to
deliver more complex services and ease their consumption, deployment and operation (the
“Toolbox”), b) make visible these tools and other open source software components for cloud
and AI computing services in a repository at Union level (the open source Cloud and AI
“Repository”).
The Repository will improve their visibility and accessibility to the cloud and AI service
provider community, and the Toolbox will facilitate smaller providers to enrich and enable a
more integrated and competitive offer of cloud and AI services, reducing the functionality gap
with the rich marketplaces provided by the dominant players.
One-off administrative costs for the European Commission. The Commission will dedicate
0.5 FTEs during the first year to design the programme(s) for the development of Toolbox
components.
Recurrent administrative costs for the European Commission. 0.5 FTE is estimated to
manage every year the repository, review and monitor the programme(s) and run
administrative, compliance, financial, communication and community engagement work
associated to the Toolbox development and the open source Repository.
One-off adjustment costs for the European Commission. The Commission will allocate 500
000 EUR as one-off cost for building the cloud and AI service toolbox repository and will
dedicate 0.5 FTEs to procure it and set it up.
Recurrent adjustment costs for the European Commission. The EC will engage an external
contractor to maintain the service repository with each year 1 FTE. The EC will host the
platform with a yearly cost of 100 000 EUR.
A community of cloud and AI computing service providers, software and technology providers,
cloud and AI consumers from different sectors, research organizations and public sector
organizations will be set up to help in defining requirements and objectives for the development
programmes and for the repository, and provide feedback about the results to steer them
properly.
Indirect savings: the expected benefit is an qualitative improvement of the offer of Cloud and
AI computing services from European providers that will increase the competitiveness of their
portfolios and help them gain market share and scale. The open source nature of the
developments will also contribute to the compatibility, interoperability and portability of
solutions among European cloud and AI service providers, improving the freedom of choice
for customers and competition. Due to the community nature of the development, it will also
increase innovation by allowing contributions and reach for small companies and individuals.
Sensitivity analysis: The benefits are only presented from a qualitative perspective.
Results for all stakeholders aggregated and for the years covered by this measure are
summarised in the table below.
Table 44. EU cloud and AI toolbox (PM24)
136
Cost types Cloud & AI
providers (€)
Public
authorities
(€)
European
Commission
(€)
Total Value
(central, €)
Total Value
(min, €)
Total Value (max,
€)
one-off administrative - - - - - -
recurrent administrative - - - - - -
one-off adjustment - - 542 134 542 134 542 134 542 134
recurrent adjustment -. - 2 364 883 2 364 883 1 938 373 2 791 393
one-off regulatory fees - - - - - -
recurrent regulatory fees - - - - - -
one-off enforcement. - - - - - -
recurrent enforcement - - - - - -
Total costs -- 2 907 017 2 907 017 2 480 507 3 333 527
Total benefits - - - - `- -
Net impact - - 2 907 017 2 907 017 2 480 507 3 333 527
4. STRENGTHS AND LIMITATIONS OF THE ANALYSIS
This impact assessment uses as main basis the results and the modelling stemming from the
study carried out by Technopolis Group. In spite of extensive efforts to collect relevant data
through surveys, interviews and workshops certain gaps remain. There are a number of points
for which projections and estimates had to be considered. For instance, there is no clear
objective value on the deployed data centre capacity in the EU. The study carried out by
Technopolis et al. (2025) has developed a first-of-a-kind methodology to estimate current
capacity in the EU, which has then cross-referenced with reported literature sources. As for the
international comparison, several reports on US data centre capacity exist, but the numbers for
China are not so conclusive.
The estimation on the use of energy and water currently used by data centres present similar
issues. The reported values stem from literature sources. Both the data centre capacity figures
as of today as well as the growth scenarios have been thoroughly assessed with one-to-one
interviews, CATI interviews and in the two workshops organized along the timeframe of the
study.
Determining the market size for cloud and AI computing services remains methodologically
heterogeneous. Sources differ notably because they don’t look at same layers (IaaS, PaaS or
SaaS) or don’t aggregate them the same way (e.g. IaaS and PaaS, or IaaS-PaaS-SaaS), resulting
in partial overlaps and limited comparability across studies. As a consequence, the figures
presented here should not be read as precise point estimates. To mitigate these limitations, a
triangulation of these estimates across multiple sources was performed, validating the ranges
to establish a robustness check.
There is also no hard evidence with respect to the number of public authorities procuring cloud
and AI computing services, nor on the contracts awarded or the number of IT systems that are
either cloud-aware or cloud-native. The data base from Tenders Europe Daily (TED) has been
taken as the baseline since it contains all the notices published in the supplement of the Official
Journal of the European Union. However, not all contract services are published on TED – only
those surpassing a certain value, namely EUR 143 000 for central governments and EUR 221
000 for local and regional governments are required to be published. Furthermore, the absence
of a distinct code under the common procurement vocabulary for cloud and AI computing
services required a complex query using combinations of keywords. This poses some
limitations in the completeness of the data. Other aspects considered in the impact assessment
137
such as the federation of resources or the sovereignty risk framework, for which little hard
evidence exists due to their novelty, were based on existing pilots (e.g. Simpl98) along with
interviews to public administrations and cloud and AI computing service providers to verify
the validity of the assumptions. These were further verified in the final validation workshop
organized the Technopolis Group with stakeholders from different profiles.
It is also important to note that the direct cost and cost savings/ benefits identified in the analysis
for the Policy Measures under Section 3 above do not incorporate wider economic impacts,
which have been excluded due to limitations in available data for monetising these effects.
Therefore, the results above, do not represent a comprehensive assessment of total policy value,
but should rather be interpreted as partial estimates of efficiency of each policy measure for the
stakeholders under consideration.
Finally, wherever there was a gap that could not be covered through interviews, CATI surveys,
or literature, the study team drew reasonable assumptions. These assumptions have been tested
in the final validation workshops, which helped provide elements to confirm or adjust these
assumptions. In the event these unknown parameters played a pivotal role in the assumptions,
the contracting team applied a min-max approach to demonstrate the robustness of the data.
5. SENSITIVITY ANALYSIS
A scenario-based sensitivity analysis was run to understand the CBA model output behaviour
in response of changes of its inputs to evaluate its overall stability and reliability. It was
conducted to verify the uncertainty range (confidence interval) of the values estimated in the
CBA for the most impactful policy measures. A min and max variation for the parameters was
introduced to create word and best case scenarios for each measure. The parameters that were
varied have been selected according to the following logic:
• When the formulation of the parameter involved a high degree of arbitrary assumptions,
e.g. time savings due to administrative simplification.
• Due to the parameters’ highly volatile nature: for example, the number of proposals
submitted per year or the cost for a software application. To be noted that this variability
does not only reflect the size of the company / public authority affected (that is modelled
separately), but captures elements that are not easily separable, such as different
organisational or business model choices.
• When stakeholders reported during interviews minor to no costs or costs savings the min.
value is set to zero, while the maximum value is the parameter identified according to
other assumptions (i.e. stemming from the literature or other expert assumptions).
It is important to underline that the sensitivity analysis presented here provided a bounded
variation assessment, through which each policy measure has been tested under a minimum
and maximum scenario. As the measures differ considerably in their design, target audience,
and the type of impacts generated, the variables are not always uniform or comparable.
Therefore, the min-max range is not to be interpreted as a consistent sensitivity interval across
all measures. The objective of the exercise was to understand how their costs and benefits
would have varied under minimum and maximum conditions, when the most influential
assumptions are varied within plausible ranges. This has allowed to assess the stability of each
98 The second specific contract for Simpl includes a technical feasibility analysis to assess a federation of cloud resources
across several Member States. More information: https://ec.europa.eu/info/funding-
tenders/opportunities/portal/screen/opportunities/tender-details/14888
138
measure’s performance across a realistic range of assumptions. The variation was based on
evidence coming from interviews, observed practice and feasibility considerations. The Table
below presents the scenario-based ranges for each policy measure, which capture the
uncertainty stemming from changes in key parameters and assumptions varied jointly under
best and worst case scenarios.
Individual net benefits/costs of the measures expressed in NPV over 10 years can be seen under
each Policy Measure in Section 3, as the main indicators of the net value of each measure. The
results and interpretation, presented in the table below, show that most PMs exhibit strong
robustness across all tested ranges, with net positive impacts even under pessimistic conditions.
Table 45. Interpretation of the parameters used for the sensitivity analysis
Policy Measure Key Parameters varied
Central NPV
and scenario
range (€ m)99
Result and interpretation
PM3: Adopting
guidelines on
building
sustainable data
centres in the EU
• Adoption rate of guidelines by DC
projects (30%-70%)
• Time saved per project for
operators (5-15 days)
• Time saved per project for
authorities (5-15 days)
• Adoption effort by operators (10-
30 days)
5.8 [-3.6 – -11.7]
PM3 delivers net costs in both min
and max scenarios, as administrative
efficiencies and reduced
operator/authority workload are not
sufficient to match the adoption effort
by operators and authorities. Results
are sensitive to assumptions, with
outcome ranges exceeding 100% of
the central estimate, indicating
uncertainty around magnitude.
PM4: Project
facilitators for the
roll-out of data
centres
• Time saved in permitting
processes (4-8 months)
• No. of DC projects benefitting
from this PM by 2036 (30%-70%)
and proportionately the previous
years.
• Implementation costs for
authorities (3-5 FTEs)
• Time savings by authorities (10-
20 days)
4 617.6 [1 859.9 – 8
632.9]
PM4 generates net savings, which
increase significantly under the max
scenario, where more projects
benefit, and time savings are larger.
This shows that the measure becomes
increasingly efficient when
permitting time reductions scale.
Estimates vary widely across
assumptions, with a relative spread of
at least 100%, implying moderate
confidence in direction but limited
precision.
PM5: Mechanism
for Member States
to identify areas to
fast-track data
centre deployment
• Time saved in permitting
processes (6-10 months)
• No of DC projects using the areas
by 2036 (30-70%) and
proportionately the previous
years.
• National authorities’ effort to set
up and administer the new process
(1-8 FTEs/year)
• Energy efficiency (PUE, i.e.
amount of data centre power
dedicated to IT).
11 861.4 [6 158.1 – 18
874.9]
PM5 provides large net benefits,
which grow considerably when more
new DC facilities use fast-track areas
or when more months are saved. All
scenarios show net benefits, although
large differences between best and
worst-case outcomes (≥100%) point
to relevant scenario dependence.
PM6: National
funding support
for data centres
• National authorities’ effort for
setting up and administering the
funding scheme (2-4 FTEs)
• Number of MS administering the
scheme (25%-75%)
-10.8 [-3.7 – -21.2]
PM6 represents an adjustment and
administrative cost for public
authorities and operators. Given the
voluntary nature of the scheme, no
benefits are available or quantifiable.
The costs are expected to increase
99 NPV refers to the net present value of benefits minus costs. Positive values indicate net benefits; negative values indicate
net costs.
139
Policy Measure Key Parameters varied
Central NPV
and scenario
range (€ m)99
Result and interpretation
• Business efforts to respond to the
calls/prepare an application (35-
55 working days)
• No. of proposals submitted every
two years (6-18)
moderately in the max scenario due to
higher efforts and MS administering
the scheme. Considering potential
policy impacts, these cost variations
could be considered not very
sensitive to changes in assumptions.
Best- and worst-case results diverge
(≥50%), suggesting moderate
robustness with scenario-sensitive
magnitude.
PM7: Set
deployment targets
and monitor
progress
• Time needed per operator to
participate in the monitoring
exercise (0.5-1.5 days/year)
• Time needed by authorities to
participate in the verification of
data (2-4 working days)
• Time and resources needed by the
Commission to set up and manage
the monitoring (1-2 FTEs and 20-
30 days per year)
- 2.3 [-1.7 – -2.8]
PM7 reflects only administrative and
adjustment costs, with benefits being
largely indirect (e.g. improved
monitoring, market transparency).
Costs increase slightly with higher
participation or expanded monitoring
roles. Sensitivity is limited and cost
impacts remain modest when
compared with other PMs. Net costs
are consistently observed across
scenarios with limited variation,
indicating high robustness.
PM8: EU funding
for R&D and
innovation
ecosystems for
cloud and AI
• Total effort for the Commission to
administer the scheme, i.e. set-it
up (2-4 FTEs) and manage it
periodically (1-2 FTE)
• Business efforts to respond to the
calls/prepare an application (35-
55 working days)
• No. of proposals submitted every
two years (16-24) - 2.2 [-1.4 – -3.1]
PM8 generates adjustment costs for
the Commission to set up and manage
the scheme and administrative costs
for businesses applying to receive
possible funding. Given the uncertain
nature of the possible funding,
benefits have not been monetised.
Thus, this PM appears only as a cost
measure. Overall costs remain
relatively low with respect to other
measures. All scenarios indicate net
costs, but differences between
outcomes exceed 50%, yielding high
confidence in direction and moderate
confidence in magnitude. The range
reflects the variation in staff needs
and effort, based on eventual
proposal and funding volumes.
PM9: EU
deployment
funding for
strategic projects
• Total effort for the Commission to
administer the scheme, i.e. set-it
up (2-4 FTEs) and manage it
periodically (1-2 FTE)
• Number of proposals every two
years
• Business efforts to respond to the
calls/prepare an application (35-
55 working days)
• No. of proposals submitted every
two years (11-19)
-1.9 [-1.2 – -2.7]
Similarly to PM8, PM9 includes only
administrative and adjustment costs
for key stakeholders. As in PM8,
sensitivity is relatively low to the
varied parameters. Net costs occur
across scenarios, with moderate
dispersion (≥50%) between
outcomes. Costs increase with higher
proposal volumes and staffing
assumptions.
PM10: EU-level
identifications of
areas to fast-track
data centre
deployment
• Time saved in permitting
processes (2-4months)
• No of DC projects using the areas
by 2036 (30-70%) and
proportionately the previous
years.
7 979.7 [4 663.8 – 12
114.0]
Similarly to PM5, PM10 generates
large net savings due to the benefits
brought by earlier commissioning
and operational efficiencies,
multiplied by the new DC projects
that would benefit from this measure,
under the specific growth scenario.
140
Policy Measure Key Parameters varied
Central NPV
and scenario
range (€ m)99
Result and interpretation
• National authorities’ effort to
participate (5-15 days/year)
• Cost savings for authorities (0.5-
1.5 FTEs/year)
• Energy efficiency (PUE, i.e.
amount of data centre power
dedicated to IT).
Net benefits are robust across
scenarios, though assumption-driven
variability above 50% reduces
precision, yielding medium–high
confidence overall. The PM is
sensitive to months saved and PUE
levels. The latter significantly
influences overall OPEX. NPV
increases by around €1.1 m per month
of time saved and by €0.7 m for every
0.01 improvement in PUE.
PM11: Creating a
EU-level
harmonised
criteria for
sovereign cloud
and AI computing
services
• Number of procuring authorities
(50% - 100%)
• Share of procurement procedures
that will voluntarily request the
alignment with the definition of
sovereignty (5% - 10%) 0.3
[0.2 – 0.4]
PM11 shows a moderate level of
sensitivity with the max scenario
double the min one. Net benefits are
observed consistently across
scenarios and assumptions, with
limited dispersion, indicating high
robustness. However, the absolute
scale remains small, indicating that
even if assumptions vary the overall
impact of this PM is limited in
economic terms.
PM12: Creating
EU guidelines for
sovereign cloud
and AI computing
services for public
procurement
• Share of public administrations
using the definition in their
procurement activities (50%-
100%)
• Number of public administrations
procuring cloud services
• Time needed by public
administrations to adopt the
procedures
• Time for public administrations to
evaluate bids
• Number of cloud and AI
computing service providers that
can participate in procurement
processes aligning with the
guidelines (244-350)
1.3 [-0.6 – 9.4]
PM12 shows high variability, as the
outcome shifts from a small net cost
in the min scenario to a very large net
saving in the max one, which
represents a considerable shift in
magnitude and direction of impact.
Therefore, the results of this measure
depend heavily on the underlying
assumptions. Results alternate
between net costs and net benefits
across scenarios, combined with large
dispersion, implying low robustness
and high sensitivity to modelling
choices. The variability is driven by
mainly two parameters. First, the
average number of procured contracts
every year for highly critical services
in the EU public sector, spanning
from 67 in the central and min values
to 133. While the maximum value
stems from the projections calculated
considering the existing TED data
multiplied by the estimated value of
highly critical use case (5%, based on
input from stakeholders), the
interviews as part of the study led to
the conclusion that a more
conservative value for the central
value was needed. Secondly, on the
number of cloud providers that could
participate in the bids. The study
yielded the result that more than 400
cloud and AI computing service
providers offer their services in the
EU, while the estimates of providers
that could comply with the
141
Policy Measure Key Parameters varied
Central NPV
and scenario
range (€ m)99
Result and interpretation
sovereignty criteria for their services
range in the values of over 300.
PM15: Voluntary
sovereignty risk
assessments for
the use of cloud
and AI computing
services in the
public sector
• No. of FTEs required to get
audited (10-20 FTEs)
• No. of FTEs to maintain the audit
(2-5 FTEs)
• No. of MS that designate the
competent authorities to carry out
audit (5 - 27)
• No. of MS that adopt and apply
the voluntary sovereignty scheme,
(5-27 MS)
316.8 [83.0 – 333.8]
PM15 generates high net savings in
both scenarios, with outcomes
improving greatly under the max
assumptions..
PM16: Non-
mandatory specific
award criteria for
the procurement of
cloud and Ai
computing
services
• No. of public authorities
procuring cloud and AI computing
services
• No. of tenders and offers per year
• No. of cloud and AI computing
service providers
• Providers’ effort to align to the
award criteria
-17.1 [-4.9 – -28.3]
PM16 results in net costs in both min
and max scenarios. While the PM
generates some savings, these are not
sufficient to outweigh costs. As
above, this could be drive by the no.
of estimated authorities procuring
cloud and AI computing services and
to the administrative effort needed by
the providers to adapt to the new
criteria.
PM17: Public
Sector cloud
federation
• Capacity shared by each MS with
the federation (5%-15%)
• No. of Member States
participating in the federation
• Energy efficiency (PUE, i.e.
amount of data centre power
dedicated to IT). 12 192.4
[7 878.4 – 16
506.4]
PM17 delivers high net cost savings
across all scenarios. Given the
difference between min-max
scenarios, the PM looks sensitive to
the underlying assumptions, e.g. MS
participating in the federation or even
capacity shared. This assumption-
driven variability above 50% reduces
precision, yielding medium–high
confidence overall. As additional
capacity is shared among MS,
benefits increase substantially, as one
could expect.
PM18: Vendor-
neutral EU
cloud/AI skill
certificates
• No. of experts that can be certified
on a yearly basis (100-500)
• Certification fees ranging (EUR
50-150)
• Effort saved by opting for a
vendor-neutral certification, i.e.
the number of certifications from
other technologies and days saved
• Coordination savings for the
Commission from avoiding
duplication of effort (EUR 0-21
067)
• Reduced costs in the certification
of similar technologies from
different providers
74.3 [15.4 – 118.8]
PM18 generates net cost savings in
both min and max scenarios, with
benefits increasing significantly
under the latter. The large dispersion
(≥100%) reflects strong sensitivity to
assumptions. Despite this, it
consistently delivers a positive
impact and becomes more efficient as
cost savings scale.
PM19:
Mandatory
specific award
criteria for the
procurement of
cloud and AI
computing
services
• No. of services audited on a
yearly basis (NUTS 0:100- 500;
NUTS 1: 50-150; NUTS 2: 50-
150; NUTS 3: 20-30; min 220
services and max 830)
• No. of hours saved due to
automated dependency analysis
per year (5 - 300 hours)
- 152.6 [-100.0 – - 213.0]
PM19 shows net costs across
scenarios with medium-low
variability.
142
Policy Measure Key Parameters varied
Central NPV
and scenario
range (€ m)99
Result and interpretation
PM20: Boosting
open source in
public
administrations
• No. of initial repositories that MS
put in place (5 – 25)
• Effort to maintain and govern
repositories (0.6 - 1.2 FTE)
• Costs to set up the OSPOs (EUR
2m-3m)
• Costs to maintain the OSPOs
(EUR 8m-28m)
• Effort of OS projects developed in
the public administrations (2-5
FTEs)
• Contract values of OS projects
developed for the public
administrations (EUR 400 000-
600 000)
• Savings for not duplicating
software, (EUR 80 000 – 450 000)
3 387.5 [478.9 – 9 172.5]
PM20 produces substantial net
savings across all scenarios, making
it one of the highest impact measures.
Given the increase in savings from
the min to the max scenario, the PM
can be considered highly sensitive to
the underlying assumptions. Indeed,
extreme sensitivity to assumptions
significantly weakens confidence in
both magnitude and stability. Despite
this large variation, it remains a
robust measure generating net
savings even under conservative
assumptions.
PM21: Mandatory
sovereignty risk
assessments for
the use of cloud
and AI computing
servicesin the
public sector
• Effort to obtain the audit
• No. of MS where CSPs would
seek the audit/validation in the
absence of a single market (5-27)
• No. of cloud and AI computing
services in possession of the audit
• No of submitted bids per year.
1 343.6 [838.6 – 1699.4]
PM21 delivers consistent savings
across all scenarios, with a modest
level of sensitivity to the underlying
assumptions. Assumption-driven
uncertainty above 50% implies
moderate confidence in scale.
PM22: EU-level
procurement of
cloud and AI
computing
services
• Time saved thanks to fewer
coordination needs for national
public authorities (0.1 - 0.5 FTEs)
• FTEs saved thanks to joint
procurement (0.1 - 0.5 FTEs)
• Savings in terms of effort incurred
by participating PAs due to the
possibility to aggregate demand
and compose services (0.1 - 0.5
FTEs)
• Capacity shared by each MS with
the federation (5% - 15%)
33 969.5 [18 940.9 – 48
998.2]
PM22 delivers large net cost savings
across all scenarios. The increase
from min to max indicates that the
PM may be quite sensitive to the
underlying assumptions. This large
dispersion limits quantitative
certainty. Despite this, it remains
robust as even in the min scenario,
cost savings are significant.
PM23: Financial
support for SMEs
to adopt cloud and
AI
• Resources dedicated by SMEs for
applications (10-30 working days)
-66.4 [-35.5 – -97.5]
PM23 presents net costs across all
scenarios, although the overall scale
is modest with respect to the other
measures. Net costs stem also from
the non-quantified strategic benefits
generated by this measure.
Assumption sensitivity introduces
uncertainty (>50%) around central
estimates.
Policy measures affecting permitting timelines and data centre deployment (PM4-PM5 and
PM10) deliver substantial net savings/benefits in both minimum and maximum scenarios.
Although the magnitude of savings is sensitive to adoption rates, months saved, deployment
trajectories and PUE reductions, the impact of the measures remains positive even when such
parameters change, demonstrating the efficiency of interventions that accelerate data centre
project delivery. PM6-PM9 represent measures highlighting only costs for operators, national
public authorities or the Commission as their benefits were either non-quantifiable or indirect.
These measures exhibit low sensitivity overall, with modest variations across min-max
scenarios and relatively small impacts with respect to the other measures. Results for PM11-
143
PM24 show a wide range of impacts. Most measures generate net savings under the different
scenarios, even under conservative assumptions, thus demonstrating the robustness of results.
At the same time, several PMs show high sensitivity to the underlying parameters. PM17,
PM20, PM21 and PM22 stand out for their high impact and strong robustness, with large
savings, while PM12 shows a mixed performance, shifting from net costs to net cost savings
across the min and max scenarios. The variability of PM12 is driven by mainly two parameters.
First, the average number of procured contracts every year for critical services in the EU public
sector, spanning from 67 in the central and min values to 133. While the maximum value stems
from the projections calculated considering the existing TED data multiplied by the estimated
value of highly critical use case (5%, based on input from stakeholders), the interviews as part
of the study led by Technopolis led to the conclusion that a more conservative value for the
central value was needed. Secondly, on the number of cloud providers that could participate in
the bids. The study led by Technopolis yielded the result that more than 400 cloud and AI
computing service providers (see Annex 4) offer their services in the EU, while the estimates
of providers that could comply with the sovereignty criteria for their services range in the values
of over 300. These values were cross validated in the final validation workshop from the study
led by Technopolis.
A sensitivity assessment was also performed for the measures evaluated using a discounted
cash flow approach. While several variables influence project value, the analysis focused on
those parameters that could be directly influenced by the policy intervention, i.e. 1) Power
Usage Effectiveness (PUE), which affects operating costs and consequently the project’s NPV
and 2) Time savings, which impact the timing of revenues and discounting of future cash flows.
In the model, reducing the project development timeline from 32 months to 26 months (6-
month acceleration), increases NPV by around EUR 6.4 m, implying a sensitivity of around
EUR 1.1 m of NPV per month of time saved. PUE affects operating costs by determining the
electricity needed per unit of IT load. Reducing PUE levels while keeping the other
assumptions unchanged, leads to around EUR 0.7 m of NPV for every 0.01 improvement in
PUE.
While the model is sensitive to additional variables, these cannot be directly tackled by the
class of policy measures under review. Utilisation and revenues are mostly influenced by
market demand and commercial strategies. CAPEX levels are determined by construction
market dynamics and technology costs, unless specific subsidies are applied. The WACC is
predominantly affected by financial market conditions and sovereign risk, necessitating macro-
financial instruments rather than operational regulations to influence it. Taxation and
depreciation are dictated by national fiscal policies. Power prices are shaped by wholesale
market rates and grid tariff structures. Therefore, the analysis concentrates on timeline and
PUE, as both can be technically managed through targeted policies and hold significant
economic importance.
This analysis will be particularly useful to better understand how to tailor future policy design.
Given the available evidence and heterogeneity of measures evaluated, it has been instrumental
to understand where the outcomes of each policy measure where more assumption-dependent
or stable.
6. MULTI-CRITERIA DECISION ANALYSIS
The multi-criteria analysis has been carried out by Technopolis et al (2025) under the study
“Study: cloud and AI”. The models used, the criteria, weights of the impacts as well as the
attained results are detailed next.
144
6.1. Models used for the multi-criteria analysis
The multi-criteria decision analysis (MCDA) has been implemented using three
complementary models:
1) A simple additive model based on Multi-Attribute Utility Theory (MAUT)
2) Preference Ranking Organisation Method for Enrichment Evaluation (PROMETHEE
II)
3) ELimination Et Choix Traduisant la REalité (ELECTRE III)
The use of multiple models allows to triangulate the results. Firstly, MAUT provides
transparent aggregation, while PROMETHEE II and ELECRE III introduce preferences and
outranking logics that test the robustness of the findings. The three methods provide three
different categories of models: full compensatory (MAUT), partially compensatory
(PROMETHEE II), and non-compensatory (ELECTREE III).
The three models are designed to work in combination:
1) MAUT provides a clear and transparent baseline;
2) PROMETHEE II is an outranking method and incorporates preference thresholds,
capturing the strength of differences, allowing partial compensation;
3) ELCETREE III tests robustness and non-compensatory consistency, identifying cases
where trade-offs between criteria are not acceptable. It also belongs to the outranking
methods, but it incorporates preferences, indifferences and veto thresholds.
All models use the same normalised and weighted option-level dataset, enabling consistent
interpretation across approaches.
6.1.1 Multi-Attribute Utility Theory (MAUT)
The MAUT approach100 is a classical multi-criteria decision-making method based on utility
theory. It uses utility functions to evaluate and compare alternatives based on multiple criteria.
The process in MAUT typically involves the following approach:
1. Definition of the relevant criteria;
2. Assignation of weights to the criteria to reflect their relative importance;
3. Normalisation of the performance for each alternative; and
4. Computation of the utility function per policy option, as a weighted sum of its scores
across all criteria, capturing the overall desirability of each option.
This approach makes it a fully compensatory model, meaning that a strong performance in one
impact can offset weaker performance in another. The benefits of this model are simplicity and
transparency. It provides a clear overall ranking of options while allowing the decomposition
of results by impact area, making it easy to identify which dimensions drive performance
differences.
In the current analysis, each policy option is assigned a normalised score for every sub-impact
area based on the CATI surveys carried out with external stakeholders101, and on the estimated
100 Jansen, S.J.T. (2011). The Multi-attribute Utility Method. In: Jansen, S., Coolen, H., Goetgeluk, R. (eds) The Measurement
and Analysis of Housing Preference and Choice. Springer, Dordrecht. https://doi.org/10.1007/978-90-481-8894-9_5 101 The CATI survey gathered feedback from key stakeholder groups, including public authorities, data centre operators, cloud
and AI service providers, industry associations, and domain experts. Respondents were asked to assess the expected
effectiveness and impact of a series of policy measures addressing two main problems: limited availability of sustainable
145
data on costs and savings presented in the sections above. These are multiplied by their
respective impact and sub-impact weights. The resulting composite scores provide a
straightforward and replicable measure of performance across scenarios.
The MAUT model employs an additive aggregation structure, assuming preferential
independence across criteria. Each policy option receives a utility score calculated as:
() = ∑ ⋅ ()
=1
where:
1) ()is the normalised score for impact under option ;
2) is the composite weight derived from the impact and sub-impact levels.
The MAUT additive model serves as the foundation for comparison with the more preference-
based methods (PROMETHEE II and ELECTRE III), as described next.
6.1.2 Preference Ranking Organisation Method for Enrichment Evaluation
(Promethee II)
The Preference Ranking Organisation Method for Enrichment Evaluation (PROMETHEE II)102
is a preference-based and partially compensatory method. It differs from the additive models
by considering how strongly one policy option is preferred over another on each criterion,
rather than aggregating scores directly. PROMETHEE II compares every pair of alternatives
across all criteria and calculates a preference degree that expresses the extent to which one
option performs better than another. These degrees are then combined into positive and
negative preference flows to establish a final net ranking of the option. A positive flow
represents how much the option outranks the other while a negative one shows how much it is
outranked. The net flow (positive minus negative) provides a full ranking of the alternatives.
This approach is valuable because it incorporates thresholds to reflect realistic decision-making
behaviour, recognising that small performance differences may be insignificant while larger
one signal a clear preference.
Thresholds
For each criterion k, two thresholds are defined:
1) an indifference threshold (q), below which the difference between two options is
negligible. This was set to 0.25 of the total score range.
2) a preference threshold (p), beyond which one option is clearly preferred. This was set
to 0.50 of the total score range.
computing capacity infrastructure in the EU, and limited availability of European cloud services including for highly critical
use cases. Each measure was rated on a likert seven-point scale from strong decrease to strong increase, complemented by
ranking exercises to identify the most effective packages and instruments. 119 stakeholders (114 economic operators and 5
Public Authorities) assessed expected changes in impact areas under each policy measure using a seven-point scale (from
strong decrease = -3 to strong increase = +3). 102 J. Figueira, S. Greco, M. Ehrgott. Multiple Criteria Decision Analysis State of the Art Surveys. Springer, Berlin (2005).
1045 pp
146
By using these thresholds, the model allowed limited compensation and avoided over-ranking
options that performed only marginally better on criteria. PROMETHEE II is therefore
effective for policy evaluations involving mixed quantitative and qualitative evidence, as it can
reflect nuanced stakeholder preferences while still providing a complete ranking of options.
PROMETHEE II is a preference-based, partially compensatory method relying on pairwise
comparisons between alternatives. For each pair (, ), the method computes a preference
degree ∈ [0,1]for each criterion , based on the difference in their performance:
= (
)
where = () − ()and is a preference function defined by:
1) Indifference threshold : differences below this are considered negligible;
2) Preference threshold : differences above this imply strict preference.
The overall preference index between and is:
Π = ∑
=1
Positive (+) and negative (−) flows are then computed:
+() = 1
− 1 ∑ Π , −()
≠
= 1
− 1 ∑ Π
≠
The net flow (() = + − −) provides a complete ranking of the alternatives.
6.1.3 ELimination Et Choix Traduisant la REalité (Electre III)
The ELimination Et Choix Traduisant la REalité (ELECTRE III)103 method is an outranking
approach. Unlike MAUT, or PROMETHEE II, it does not assume that trade-offs between
criteria are always acceptable. Instead, it tests whether there is sufficient evidence to state that
one policy option outranks another, i.e. performs at least as well overall, without being
significantly worse on any major criterion.
ELECTRE III is a non-compensatory model, which makes it useful when some criteria (e.g.
environmental sustainability) are considered essential and should not be fully offset by gains
elsewhere. It combines two types of information:
1) Indifference threshold (q): small differences are ignored. This was set to 0.25 of the
total score range.
2) Preference threshold (p): a preference threshold (p), larger differences indicate a clear
preference. This was set to 0.50 of the total score range.
103 Uzun, B., Bwiza, R.A., Uzun Ozsahin, D. (2021). ELimination Et Choix Traduisant La REalité (ELECTRE). In: Uzun
Ozsahin, D., Gökçekuş, H., Uzun, B., LaMoreaux, J. (eds) Application of Multi-Criteria Decision Analysis in Environmental
and Civil Engineering. Professional Practice in Earth Sciences. Springer, Cham. https://doi.org/10.1007/978-3-030-64765-0_5
147
3) Veto threshold (v): if the difference exceeds this level, one criterion can block the
overall outranking regardless of others.
Thresholds are set to match chosen the way the PROMETHEE model is configured. ELECTRE
III is particularly beneficial in regulatory and policy contexts, as it can identify incomparability
i.e. cases where neither option is clearly superior, which mirrors complexity of multi decision-
making.
Summarizing, ELECTRE III is a non-compensatory outranking method that assesses
whether one alternative “outranks” another by combining concordance (agreement) and
discordance (veto) indices.
For each pair of alternatives (, ):
1) A concordance index expresses the degree to which criteria support the statement
“ is at least as good as ”:
= ∑ (, )
∈
where is the partial concordance per criterion.
2) A discordance index measures the extent of opposition to that statement, identifying
criteria where substantially outperforms .
3) A credibility index is then computed as:
= × ∏ 1 −
1 −
∈discordant
which quantifies the overall strength of the outranking relation.
The final ranking is obtained through exploitation procedures, balancing credibility thresholds
and identifying dominance, indifference, and incomparability relations
6.1.4 Summary of models used for the multi-criteria analysis
The table below depicts a summary of the models used for this MCDA with their main
characteristics.
Table 46. Summary of models used for the multi-criteria analysis
Feature MAUT PROMETHEE II ELECTRE III
Method type Compensatory,
utility-based
aggregation
Outranking with
pairwise preference
flows
Outranking with credibility
and discordance indices
Compensation Fully compensatory Semi-compensatory Non-compensatory (veto can
block dominance)
Output Full ranking Full ranking Partial ranking
Thresholds None Uses preference
functions
Uses explicit indifference,
preference and veto
thresholds
148
Feature MAUT PROMETHEE II ELECTRE III
Strengths Easy and clear to
communicate
Pairwise comparisons Handles uncertainty, and
incorporates veto
Data used Quantitative scores
and weights
Quantitative, but
requires selecting
preference functions
Mixed data, requires defining
discordance and credibility
indices
6.2. Development and comparisons of policy options (MCDA)
The multi-criteria decision analysis allows to evaluate the relative performance of different
policy options under study. It enables a systematic comparison of alternatives when multiple
and often conflicting objectives must be considered.
The analysis combined quantitative data from two sources:
1) Survey data, in which 119 stakeholders (114 economic operators and 5 Public
Authorities) assessed expected changes in impact areas under each policy measure
using a seven-point scale (from strong decrease = -3 to strong increase = +3). The
sample is described below.
2) Estimated data developed through quantitative modelling, covering administrative
costs, adjustment costs, and cost-saving for each policy measure.
All data is first combined and harmonised to create a single dataset covering all policy options,
policy measures, impact areas, and sub-impact areas.
Weights are applied to reflect the relative importance of each impact and sub-impact. The
reference case is equal weighting. Given the different nature that the two problems deal with,
different weights for problem 1 and problem 2 are also considered.
The impact-level weights define the overall direction of the assessment by indicating which
broad dimensions carry the greatest influence under each scenario. In the equal weighting
scenario, all four impacts – economic/effectiveness, costs/efficiency, social and environmental
- are weighted equally (25%). Under Problem 1 (P1), the weighting shifts towards the
economic/effectiveness dimension (40%), highlighting the priority of accelerating deployment
and enhancing efficiency, while social impacts receive reduced emphasis (5%). For Problem 2
(P2), the weighting moves towards costs and economic/effectiveness (55%) impacts, reflecting
a focus on sovereignty, trust, and resilience.
Table 47. Impacts weighting (Source: Technopolis et al. (2025))
Impact Equal weighting [EW] P1 weights P2 weights
Economic/effectiveness 25% 40% 55%
Costs/efficiency 25% 35% 35%
Social 25% 5% 5%
Environmental 25% 20% 5%
The sub-impact weights specify how the detailed components within each impact contribute to
the overall evaluation. Under the equal weighting scenario, the weights are balanced across
sub-criteria. Under Problem 1, the emphasis shifts towards sub-impacts linked to rapid
deployment, computing capacity, and cost efficiency, aligning with the scenario’s objective of
speeding up infrastructure expansion. Conversely, Problem 2 introduces a strong emphasis on
digital sovereignty, open source capacity, customer choice, and transparency, consistent with
149
its social and governance orientation. Environmental sub-impacts source also adjust,
prioritising reductions in data centre footprints under Problem 1 and digital sovereignty, and
resilience under Problem 2.
Table 48. Sub-impacts weighting (Source: Technopolis et al. (2025))
Sub-Impact Equal
weighting sub -
impact
P1 weights sub
- impact
P2 weights sub
- impact
Speed of construction and deployment
of data centres
14% 30% 0.5%
Computing capacity for AI services and
general purpose computing
14% 20% 0.5%
Adoption of high-performance, low-
carbon, energy-efficient cloud and AI
computing services
14% 25% 0.5%
Cross-border market access for EU-
based cloud and AI computing service
providers
14% 10% 14.5%
Digital sovereignty and resilience in the
public sector including through public
procurement
14% 7% 50%
Open source capacity 14% 5% 14.5%
Ensuring customer choice and the ability
to switch providers across layers of the
AI compute stack
14% 3% 19.5%
Administrative costs 25% 15% 15%
Adjustment costs 25% 15% 15%
cost savings 25% 30% 30%
net costs 25% 40% 40%
Transparency and citizen trust 50% 60% 65%
Citizen/community engagement 50% 40% 35%
Data centres environmental footprint
(energy, water use, greenhouse gas
emissions)
50% 70% 50%
Leveraged investments from public and
private actors for sustainable
infrastructure and services
50% 30% 50%
In this framework, the sub-impact weights shown in the table below are derived from the
product of the higher-level impact weights and the relative importance of each sub-impact
within that impact area.
The calculation works in two steps:
1. Each main impact area, Economic/effectiveness, Costs/efficiency, Social, and
Environmental, is first assigned an overall weight at the impact level (for example,
under Problem 2: Economic/effectiveness = 55%, Costs = 35%, Social = 5%,
Environmental = 5%).
2. Within each impact area, that total weight is distributed across its constituent sub-
impacts according to their relative importance in the scenario. The weight of each sub-
150
impact is therefore the impact-level weight multiplied by its share within that impact
group.
For instance, under Problem 2, the Economic/effectiveness impact (55%) is subdivided across
seven sub-impacts, with digital sovereignty and resilience receiving a larger internal share than
speed of construction. Multiplying this impact weight (55%) by the internal share of digital
sovereignty and resilience (around 35%) yields a sub-impact weight of roughly 12.3% in the
table. The same process applies across all impact areas: the economic/effectiveness weight
(55%) multiplied by the internal weighting for digital sovereignty and resilience in the public
sector including through public procurement (50%) produces a final sub-impact weight of
27.5%.
This hierarchical weighting ensures that the total of all sub-impacts adds up to 100% and that
each sub-impact’s influence on the final composite score reflects both:
1. The relative priority of its parent impact area (e.g. Social vs Effectiveness/economic),
and
2. The specific emphasis placed on that sub-impact within the impact group (e.g. within
Social, transparency outweighs engagement).
Table 49. Sub-impacts weighting used (Source: Technopolis et al. (2025))
Sub-Impact Equal weighting
sub - impact
P1 weights
sub - impact
P2 weights sub
- impact
Speed of construction and deployment
of data centres
3.57% 12.0% 0.3%
Computing capacity for AI services and
general purpose computing
3.57% 8.0% 0.3%
Adoption of high-performance, low-
carbon, energy-efficient cloud and AI
computing services
3.57% 10.0% 0.3%
Cross-border market access for EU-
based cloud and AI computing service
providers
3.57% 4.0% 8.0%
Digital sovereignty and resilience in the
public sector including through public
procurement
3.57% 2.8% 27.5%
Open source capacity 3.57% 2.0% 8.0%
Ensuring customer choice and the
ability to switch providers across layers
of the AI compute stack
3.57% 1.2% 10.7%
Administrative costs 6.25% 5.3% 5.3%
Adjustment costs 6.25% 5.3% 5.3%
cost savings 6.25% 10.5% 10.5%
net costs 6.25% 14.0% 14.0%
Transparency and citizen trust 12.50% 3.0% 3.3%
Citizen/community engagement 12.50% 2.0% 1.8%
Data centres environmental footprint
(energy, water use, greenhouse gas
emissions)
12.50% 14.0% 2.5%
151
Sub-Impact Equal weighting
sub - impact
P1 weights
sub - impact
P2 weights sub
- impact
Leveraged investments from public and
private actors for sustainable
infrastructure and services
12.50% 6.0% 2.5%
6.2.1 Summary of results multi-criteria decision analysis
In both Problem 1 and Problem 2, all policy options perform better than the baseline,
consistently across the three methods tested and stakeholder groups (AI and cloud service
providers, data centre operators, National Public Authorities and SMEs). The assessment
therefore confirms that maintaining the status quo is the least attractive option, and that
intervention is warranted from both an effectiveness and efficiency perspective.
The multi-criteria analysis is based on the original specification of policy options. While some
measures were subsequently refined, the options were retained in their initial form to ensure
comparability across scenarios. The results draw on stakeholder survey data collected in
August-September 2025 and cost inputs updated in early 2026. As such the findings should be
interpreted as indicative, reflective the relative performance of the options.
The table below presents an overview of the comparative performance of the options across
methods and weighting schemes.
Table 50. Summary of option performance across methods and problems (Source: Technopolis et al.
(2025))
Problem Method Main conclusion
P1 MAUT PO1-B and PO1-C are strongest, especially under P1 weights.
P1 PROMETHEE II PO1-B and PO1-C clearly ahead; PO1-A only modest improvement.
P1 ELECTRE III Ranking: PO1-B > PO1-C > PO1-A = baseline
P2 MAUT Under P2 weights, PO2-B and PO2-C are strongest
P2 PROMETHEE II PO2-C best, PO2-A second; PO2-B weak, close to baseline under P2 weights
P2 ELECTRE III PO2-C clearly first; PO2-A and PO2-B at best marginal over baseline
Problem 1
For Problem 1, all three options (PO1-A/B/C) deliver a clear improvement over the baseline in
all three methods, namely, MAUT, PROMETHEE II and ELECTRE III. Under equal
weighting, MAUT shows that all options achieve positive total scores, while the baseline
remains negative. When Problem 1 weights are applied the differences between options become
more pronounced. In this configuration, PO1-B and PO1-C clearly outperform PO1-A, and the
baseline remains the weakest.
Across the three methods, a consistent pattern emerges:
• PO1-B (legislative and financial intervention enforced nationally) and PO1-C (legislative
and financial intervention enforced at EU-level) are the strongest options and both strongly
dominate the baseline.
• PO1-A (collaborative framework) provides only a moderate improvement over the baseline:
positive overall, but not comparable in magnitude to the more interventionist options.
152
The baseline never outranks any option and is always clearly dominated in the outranking
analyses.
At measure level:
• For PO1-A, PM3 (guidelines on sustainable data centres) is the main driver of performance,
as it is the only measure that clearly turns the cost score positive under Problem 1 weights
and contributes solidly to the total score. PM1 (Alliance working group) and PM2
(stakeholder forum) add small but consistent gains on effectiveness and social criteria but
remain modest.
• Within PO1-B, performance is driven by PM4 (project facilitator for the roll-out of data
centres) and PM5 (fast-track area identification). These measures combine improved cost
outcomes with strong effectiveness and therefore explain the option’s high overall score.
• In PO1-C, all measures contribute positively, but PM10 (EU-level identification of fast-
track areas) stands out, especially under Problem 1 weights, by delivering the best cost
outcome in the package alongside strong effectiveness. PM8 (EU R&D funding) and PM9
(EU deployment funding) generate clear environmental and social benefits with a high
effectiveness but retain slightly negative cost scores, reflecting the associated cost burdens.
Stakeholder-specific results show that all groups are better off than in the baseline for all
options, but they differ in their preferred bundle:
• Economic operators favour PO1-B, with PO1-C close behind, driven by better cost and
effectiveness outcomes.
• National public authorities also obtain the highest scores under PO1-B, followed by PO1-
C; PO1-A is positive but less attractive. Improvements for authorities arise mainly from
better cost outcomes relative to the baseline and small social and environmental gains,
rather than from direct effectiveness.
• CSPs and AI service providers see substantial improvements under all options but show a
clear preference for PO1-B, particularly when Problem 1 weights are applied, reflecting the
importance of streamlined permitting and deployment-oriented measures.
• SMEs benefit substantially from all options relative to the baseline, but PO1-B is the most
favourable bundle, as it combines improved costs with higher effectiveness and social
scores for this group.
The evidence for Problem 1 suggests that PO1-B is the most attractive option when priority is
given to deployment and cost efficiency, while PO1-C is favoured when it comes to social and
environmental benefits, at the expense of somewhat higher perceived cost intensity in EU-level
funding measures. PO1-A offers a modest, low-intensity improvement but does not match the
performance of the more ambitious packages.
Problem 2
For Problem 2, all three options (PO2-A/BC) perform clearly better than the baseline across
the three MCDA methods. Under equal weighting in MAUT, total scores for all options indicate
broad improvements in effectiveness, social and environmental outcomes. When Problem-2
weights are applied all options improve further and remain clearly above the baseline, with
PO2-B and PO2-C achieving the highest MAUT scores.
However, once outranking approaches are considered, a stable pattern emerges:
• PO2-C (EU-coordinated procurement and framework) is the strongest option overall. Both
PROMETHEE II and ELECTRE III identify PO2-C as the preferred option, under both
equal and Problem-2 weights.
153
• PO2-A (transparency and visibility) is consistently a secondary but clearly positive option,
always better than the baseline but not reaching the performance of PO2-C.
• PO2-B (voluntary EU framework) is, at best, marginally better than the baseline in
ELECTRE III and is clearly weaker than PO2-C (and often PO2-A) in PROMETHEE II,
particularly when Problem-2 weights are applied.
At measure level:
• For PO2-A, the strongest contributors are PM11 (creating a sovereign cloud and AI
computing services definition), PM12 (guidelines for sovereign services) and PM14
(interoperability flanking measures). These measures are seen as useful and relatively low-
cost, particularly once sovereignty and resilience gain more weight. PM13 (annual
sovereignty conference) provides only limited added value and mainly adds some cost,
resulting in lower total scores.
• In PO2-B, performance is driven by PM15 (sovereignty scheme) and PM17 (cloud
federation), which combine neutral or positive cost impacts with solid effectiveness and
social gains. PM18 (training programme) generates moderate, clearly positive
contributions, while PM16 (voluntary award criteria) is weaker.
• Within PO2-C, PM20 (promotion of open source use) and PM22 (EU-level procurement
and federation) are the main drivers of performance, combining manageable or slightly
positive cost effects with strong effectiveness and social benefits when sovereignty-related
dimensions are prioritised. PM21 (mandatory use of audited services and award criteria)
becomes more attractive under Problem-2 weights, while PM23 (SME support) has low
totals and negative cost scores, limiting its contribution to the option’s overall strength.
Stakeholder-specific results for Problem 2 again indicate clear improvements relative to the
baseline for all groups, but with notable differences in preferences:
• Economic operators see all options as an improvement, but clearly prefer PO2-C, which
combines the strongest cost outcome with robust effectiveness and social gains. PO2-A is
generally second, with PO2-B weaker overall.
• National Public Authorities favour PO2-B under EW and PO2-C under P1 weights. PO2-A
remains positive but leaves a more significant cost burden; PO2-C offers smaller gains and
only marginal cost improvements from their perspective versus PO2-B also under P2
weights.
• Cloud and AI computing service providers see substantial gains under all options, but PO2-
C is the preferred grouping, particularly when Problem-2 weights are applied, with PO2-A
second in EW and PO2-B only marginally improving over PO2-A under P2 weights.
• SMEs benefit substantially from all options compared to the baseline, but PO2-A is the
most favourable option under EW and PO1-C is the preferred one under P1 weights, as it
combines the least negative cost scores with the highest economic and social benefits for
this group.
The results for Problem 2 indicate that PO2-C is the most robust option at system level,
especially when non-compensatory trade-offs and sovereignty-focused priorities are
considered, while PO2-B is most attractive for national public authorities and PO2-A is
particularly beneficial for SMEs. All three options remain markedly superior to the baseline,
confirming that a policy response to Problem 2 is justified.
6.2.2 Multi-criteria analysis for Problem 1
MAUT results for Problem 1 show a clear improvement of all three policy options (PO1-
A/B/C) over the baseline across both weighting schemes. Under equal weighting, all options
154
deliver positive total scores (0.399 for PO1-A, 0.575 for PO1-B, 0.513 for PO1-C), while the
baseline is distinctly negative (-0.359). This reflects the fact that the measures jointly improve
effectiveness, environmental and social outcomes relative to the status quo, while also reducing
the net cost burden. Social impacts are consistently positive for all options and negative for the
baseline, and environmental impacts also improve slightly relative to the baseline under equal
weighting.
When shifting to the Problem 1 weighting (which places more emphasis on effectiveness and
cost impacts, and sustainability), the overall ranking is broadly stable but the differences
between options become more pronounced. PO1-B and PO1-C clearly dominate, with total
scores of 0.688 and 0.545 respectively, compared to 0.312 for PO1-A and -0.387 for the
baseline. The reweighting particularly favours PO1-B, which combines the strongest
performance in terms of effectiveness with the best cost-benefit outcome (costs turning positive
under P1), while still performing better than the baseline on environmental and social
dimensions. This suggests that, under a deployment-focused policy perspective, PO1-B (and to
a slightly lesser extent PO1-C) represent the most attractive bundle of measures among the
options considered.
Table 51. MCDA Aggregate Results MAUT – Problem 1 (Source: Technopolis et al. (2025))
The PROMETHEE II results for Problem 1 clearly indicate that PO1-B and PO1-C are the
strongest options, with PO1-B emerging as the preferred options once Problem 1 priorities are
reflected. Under equal weighting, most options improve on the baseline which shows strongly
negative net flows at both impact and sub-impact level (–0.917 and –0.798). At impact level,
PO1-B strongly outperforms PO1-C with positive impact-level flows (0.500 vs 0.189), while
PO1-A delivers only a modest gain with respect to the baseline (-0.189). At the sub-impact
level, PO1-B still outperforms PO1-C but the difference between the two options decreases
(0.312 vs 0.289), suggesting a marginal advantage when all criteria are treated as equally
important.
When the Problem 1 weighting is applied, the ordering becomes more pronounced. The
baseline remains negative (–0.867 at impacts; –0.671 at sub-impacts), and PO1-A remains
negative at both levels (–0.447 impacts; –0.331 sub-impacts), indicating limited alignment of
this option with deployment- and cost-driven priorities. By contrast, PO1-B and PO1-C both
achieve high impact-level flows (0.583 and 0.147). At the sub-impact level under P1, PO1-B
(0.413) outperforms PO1-C (0.252), which points to PO1-B as the best overall option once
implementation-level effects and the stronger emphasis on effectiveness and cost criteria are
considered.
Table 53. PROMETHEE II - Problem 1 (Source: Technopolis et al. (2025))
Option EW - Impacts EW Sub-impacts P1 - Impacts P1 – Sub-Impacts
Baseline -0.917 -0.798 -0.867 -0.671
PO1-A -0.189 -0.112 -0.447 -0.331
PO1-B 0.500 0.312 0.583 0.413
PO1-C 0.189 0.289 0.147 0.252
Impact Costs Economic Environmental Social Total Costs/efficiency Effectiveness ironmental Social Total
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline -0.538 -0.714 0.242 0.470 -0.037 -0.134 -0.026 -0.008 -0.359 -0.387
PO1-A -0.119 -0.128 0.246 0.421 0.055 -0.024 0.216 0.043 0.399 0.312
PO1-B -0.011 0.143 0.298 0.543 0.074 -0.041 0.214 0.043 0.575 0.688
PO1-C -0.089 0.017 0.307 0.520 0.078 -0.036 0.216 0.044 0.513 0.545
155
For Problem 1, the ELECTRE III outranking analysis shows a very consistent pattern: at impact
level, both under equal weighting and under the P1 weights, the descending order is PO1-B >
PO1-C > PO1-A = baseline, meaning that PO1-B outranks PO1-C, and both outrank the
baseline, while PO1-A does not. At the sub-impact level, in both EW and P1 options the ranking
changes making PO1-C > PO1-B > baseline = PO1-A, with high credibility indices for the
outranking relations between PO1-C and the other options. In other words, once non-
compensatory trade-offs and detailed criteria are fully taken into account, PO1-C is the
strongest option, closely followed by PO1-B, while PO1-A is barely better than or equal to the
baseline and never outranks the two more ambitious options. The difference in these results
between the impact level and the sub-impact results reflects how the structure of the analysis
influences the comparison of the options. When impacts are assessed at a broader level, strong
performance in one area can offset weaker performance in another, i.e. making PO1-B appear
preferrable. However, when this analysis is performed at sub-impact level, the trade-offs are
no longer absorbed within the broader categories, i.e. PO1-C performs more evenly and
therefore emerges as the strongest option. The different ranking thus reflects the differences in
how balanced the policy options are. PO1-B appears to generate higher overall impacts but
with greater divergence across cost-benefit, effectiveness, environmental and social
dimensions, whereas PO1-C delivers more balanced, though less pronounced results across
these criteria. Importantly, PO1-B and PO1-C both outperform PO1-A which does not improve
over the baseline and thus represent improvements over the status quo.
Table 54. ELECTRE III ranking of options – Problem 1 (Source: Technopolis et al. (2025))
Option Impact level (EW) Sub-impact level
(EW)
Impact level (P1) Sub-impact
level (P1)
Overall judgement
PO1-B 1 2 1 2 Very strong
PO1-C 2 1 2 1 Very strong
PO1-A 3 = Baseline 3 = Baseline 3 = Baseline 3 = Baseline Weak improvement
over baseline
Baseline 3 = PO1-A 3 = PO1-A 3 = PO1-A 3 = PO1-A Reference
Stakeholder analysis
For economic operators, all three options represent a marked improvement over the baseline,
but with a clear hierarchy. Under both equal weighting and the Problem 1 weighting, PO1-B
performs best overall (0.623 and 0.742), closely followed by PO1-C, with PO1-A some
distance behind. The main driver is the combination of improved cost outcomes and stronger
effectiveness scores: PO1-B is the only option that yields clearly positive cost results for
operators under EW (0.045), while also outranking the other options under P1 weights in terms
of costs. PO1-B delivers the highest effectiveness score (0.538) under P1 weighing while PO1-
C outperforms it under EW. Environmental scores for operators improve modestly under equal
weighting but turn slightly negative under P1 (reflecting the lower weight and some trade-offs),
whereas social scores are clearly higher than the baseline for all options. The data suggest that
economic operators have better results with national simplification and fast-track mechanisms
(PO1-B) as they are offering the best balance of lower burdens and stronger business
opportunities, with EU-level interventions (PO1-C) being also attractive but slightly less
beneficial in net terms.
For national public authorities, all options also outperform the baseline, but preference patterns
differ and the gains are more modest. Under equal weighting, PO1-B slightly dominates (0.281
156
vs 0.259 for PO1-C and 0.050 for PO1-A), driven by the strongest combination of cost, social
and environmental scores. Under the Problem 1 weighting, PO1-B still has the highest total
score (0.408), but PO1-C moves farther (0.185), while PO1-A remains the weakest performer,
albeit still clearly above the baseline. A notable contrast with operators is that the effectiveness
dimension for public authorities is closer to, or slightly below, the baseline, especially under
P1, suggesting that administrations anticipate less direct effectiveness from the measures.
Instead, their improvement comes mainly from better cost outcomes relative to the baseline
(particularly under PO1-B) and small but positive social and environmental effects. In sum,
authorities tend to favour the national level option (PO1-B), which offers a better overall
balance of impacts from their perspective, even if PO1-C performs best in terms of
environmental impact (under EW) and social impacts, mostly in relation to the measures
focused on funding for strategic and energy-efficient projects.
Table 55. MCDA Economic Operators and Public Authorities MAUT – Problem 1 (Source: Technopolis et
al. (2025))
For cloud and AI computing service providers, all three options improve on the baseline when
impacts are equally weighted, with total scores rising from 0.267 (baseline) to 0.440 (PO1-A),
0.542 (PO1-B) and 0.572 (PO1-C). Since cost impacts are neutral under this Problem (all zero),
the gains come from stronger effectiveness and social performance, with modest environmental
improvements under equal weighting. Under the Problem 1 weighting, which amplifies
effectiveness dimensions, the picture is more nuanced: PO1-A is essentially indistinguishable
from the baseline (0.378 vs 0.381), while PO1-B and PO1-C provide clear net benefits (0.499
and 0.488 respectively), driven by higher effectiveness scores. Environmental scores for
providers become slightly more negative under P1, indicating perceived trade-offs as
deployment objectives are prioritised, but these are more than offset by the effectiveness and
social gains in PO1-B and PO1-C.
For data centre operators, the options deliver very substantial improvements over the baseline
across both weighting schemes. Under equal weighting, the total score rises from –0.501
(baseline) to 0.505 (PO1-A), 0.767 (PO1-B) and 0.654 (PO1-C), with a similar pattern under
P1 (–0.438 to 0.399, 0.848 and 0.670 respectively). The baseline is characterised by strongly
negative cost scores and only modest effectiveness, whereas PO1-B in particular moves higher
(0.767 EW, 0.848 P1) while also delivering the highest effectiveness and social scores. PO1-A
and PO1-C also perform very well, but their cost outcomes are weaker than PO1-B’s.
Environmental performance improves markedly relative to the baseline under equal weighting
Impact Costs Effectiveness Environmental Social Total
Economic Operators
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline -0.529 -0.703 0.260 0.504 -0.038 -0.141 -0.003 -0.004 -0.309 -0.344
PO1-A -0.137 -0.156 0.238 0.410 0.040 -0.036 0.203 0.041 0.344 0.259
PO1-B 0.045 0.216 0.298 0.538 0.058 -0.058 0.223 0.045 0.623 0.742
PO1-C -0.063 0.049 0.313 0.527 0.059 -0.057 0.229 0.046 0.538 0.564
National Public Authorities
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline -0.390 -0.478 0.143 0.374 -0.125 -0.160 0.000 0.003 -0.373 -0.262
PO1-A -0.064 -0.055 0.063 0.096 -0.010 -0.038 0.063 0.013 0.050 0.015
PO1-B -0.023 0.108 0.188 0.333 0.004 -0.056 0.113 0.023 0.281 0.408
PO1-C -0.097 -0.101 0.195 0.304 0.021 -0.047 0.141 0.028 0.259 0.185
157
(especially in PO1-C), though again becomes slightly negative under P1 as the model places
more emphasis on economic deployment. Overall, operators clearly favour the more
interventionist national package in PO1-B, with PO1-C and PO1-A still representing
substantial improvements over the status quo.
For SMEs104, results show that all three options perform substantially better than the baseline,
with PO1-C as the most favourable option under EW and PO1-B under P1 weights. All options
improve over the baseline levels, yielding positive totals of 0.444 (PO1-A), 0.612 (PO1-B) and
0.629 (PO1-C) under EW. The improvement is driven by higher effectiveness scores (up to
0.320 for PO1-C) and clearly positive social and environmental contributions. When the
Problem 1 weighting is applied, the ordering changes: PO1-B performs best (total 0.561),
followed by PO1-C (0.526) and PO1-A (0.391), showing that as economic deployment is
emphasized over other criteria, SMEs tend to favour PO1-B rather than an EU-level approach.
Indeed, under P1, PO1-B improves mainly driver by higher effectiveness (0.569) compared to
the other options and the baseline.
Table 56. MCDA Cloud and AI Computing Service Providers and Data Centre Operators MAUT –
Problem 1 (Source: Technopolis et al. (2025))
6.2.3 Multi-criteria analysis for Problem 2
For Problem 2, all three options outperform the baseline clearly under both equal weighting
and the Problem 2 weighting, but with a distinct ordering. Under equal weighting, total scores
move from a negative baseline (-0.183) to moderately positive values for all options (0.439 for
PO2-A, 0.418 for PO2-B, 0.488 for PO2-C), reflecting broad improvements across
effectiveness, social and environmental dimensions despite remaining negative on costs. When
the Problem 2 weighting is applied, placing greater emphasis on economic and sovereignty-
related aspects, overall performance improves substantially and differences between options
104 SMEs account for cloud and AI computing service providers. Data centers are not included.
Impact Costs Effectiveness Environmental Social Total
Cloud and AI Computing Service Providers
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline 0.000 0.000 0.262 0.508 -0.026 -0.129 0.031 0.002 0.267 0.381
PO1-A 0.000 0.000 0.223 0.386 0.028 -0.046 0.189 0.038 0.440 0.378
PO1-B 0.000 0.000 0.286 0.520 0.050 -0.062 0.205 0.041 0.542 0.499
PO1-C 0.000 0.000 0.306 0.514 0.046 -0.070 0.220 0.044 0.572 0.488
Data Centre Operators
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline -0.534 -0.711 0.253 0.492 -0.083 -0.191 -0.137 -0.029 -0.501 -0.438
PO1-A -0.142 -0.164 0.298 0.509 0.089 0.002 0.260 0.052 0.505 0.399
PO1-B 0.046 0.219 0.342 0.609 0.087 -0.039 0.292 0.059 0.767 0.848
PO1-C -0.063 0.048 0.340 0.576 0.109 -0.008 0.269 0.054 0.654 0.670
SMEs
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline 0.000 0.000 0.233 0.453 -0.039 -0.132 -0.028 -0.011 0.166 0.311
PO1-A 0.000 0.000 0.222 0.386 0.039 -0.032 0.182 0.036 0.444 0.391
PO1-B 0.000 0.000 0.314 0.569 0.062 -0.056 0.237 0.047 0.612 0.561
PO1-C 0.000 0.000 0.320 0.529 0.061 -0.053 0.248 0.050 0.629 0.526
158
become more marked. The baseline remains negative (-0.220), while PO2-A rises to 0.572 and
PO2-B and PO2-C reach 0.709 and 0.928 respectively.
The shift to the P2 weighting particularly benefits PO2-B and PO2-C because they combine
improved cost outcomes (costs turn positive under P2 for both) with strong gains on the
effectiveness dimension, which carries the highest weight in this scenario. Social scores are
consistently positive and higher than the baseline for all three options in both weighting
schemes, and environmental scores, while modest, also improve relative to the status quo. PO2-
A and PO2-B perform well but are clearly dominated PO2-C under P2, indicating that more
ambitious and coordinated intervention on sovereignty and uptake delivers a better balance of
costs and benefits when judged against the Problem 2 policy priorities.
Table 57. MCDA Results MAUT – Problem 2 (Source: Technopolis et al. (2025))
The PROMETHEE II results for Problem 2 clearly point to PO2-C as the best-performing
option, with PO2-A a solid second and PO2-B consistently weakest. Under equal weighting,
all options improve on the baseline (–0.333 at impact level; –0.213 at sub-impacts), but PO2-
C achieves the highest net flows both at the impact level and sub-impact (0.333 and 0.162),
ahead of PO2-A (0.000 and 0.130) and PO2-B (0.000 and –0.079). When the Problem 2
weighting is applied, this ordering becomes even clearer: PO2-C again records the highest net
flows (0.333 at impact level and 0.443 at sub-impacts), PO2-A remains positive at sub-impact
level but more moderate (0.150) and becomes negative at impact level (-0.133), performing
worse than the baseline. Similarly, PO2-B is negative at the impact level and sub-impact (–
0.400, - 0.072 at sub-impacts) and therefore appears poorly aligned with the sovereignty- and
resilience-focused P2 priorities.
Thus, PROMETHEE II indicates that PO2-C offers the strongest and most robust performance
across both impact and sub-impact structures, with PO2-A preferable to the baseline in most
cases, and PO2-B clearly the least attractive option.
Table 59. PROMETHEE II - Problem 2 (Source: Technopolis et al. (2025))
Option EW - Impacts EW Sub-impacts P2 - Impacts P2 – Sub-Impacts
Baseline -0.333 -0.213 0.200 -0.521
PO2-A 0.000 0.130 -0.133 0.150
PO2-B 0.000 -0.079 -0.400 -0.072
PO2-C 0.333 0.162 0.333 0.443
For Problem 2, ELECTRE III at the impact level, under both equal weighting and the P2
weights, the descending ranking is PO2-C > PO2-A = baseline = PO2-B. PO2-C is the only
option that outranks all the other options and the baseline. At the sub-impact level, PO2-C again
dominates in both EW and P2 options the ordering is PO2-C > PO2-A = baseline = PO2-B,
with credibility values close to 1 for the outranking of the other options. This means that, when
non-compensatory logic is applied and deal-breaker criteria are allowed to veto weak options,
PO2-C is the clearly preferred option for Problem 2, while PO2-A and PO2-B offer at best a
marginal improvement over the status quo. Within the second group of options, PO2-C
Costs/efficiency Effectiveness Environmental Social Total
Option EW P1 EW P1 EW P1 EW P1 EW P1
Baseline -0.361 -0.547 0.242 0.345 -0.037 -0.007 -0.026 -0.010 -0.183 -0.220
PO2-A-0.098 -0.103 0.241 0.614 0.048 0.010 0.249 0.051 0.439 0.572
PO2-B-0.009 0.168 0.228 0.490 0.046 0.009 0.199 0.041 0.418 0.709
PO2-C-0.027 0.257 0.192 0.612 0.056 0.011 0.229 0.047 0.488 0.928
159
generates a meaningful improvement relative to the status quo, while PO2-A and PO2-B do not
materially change outcomes compared to the baseline.
Table 60. ELECTRE III ranking of options – Problem 2 (Source: Technopolis et al. (2025))
Option Impact level
(EW)
Sub-impact
level (EW)
Impact level
(P1)
Sub-impact level
(P1)
Overall ELECTRE
III judgement
PO2-C 1 1 1 1 Clearly strongest
PO2-A 2 = Baseline =
PO2-B
2 = Baseline =
PO2-B
2 = Baseline =
PO2-B
2 = Baseline =
PO2-B
At best marginal over
baseline
PO2-B 2 = Baseline =
PO2-A
2 = Baseline =
PO2-A
2 = Baseline =
PO2-A
2 = Baseline =
PO2-A
At best marginal over
baseline
Baseline 2 = PO2-B =
PO2-A
2 = PO2-B =
PO2-A
2 = PO2-B =
PO2-A
2 = PO2-B =
PO2-A
Reference
Stakeholder analysis
For economic operators, all three Problem 2 options deliver a clear improvement over the
baseline, but with different strengths. Under equal weighting, PO2C clearly dominates PO2A
and PO2B (0.552 vs 0.453 vs 0.421), reflecting that transparency measures and coordinated EU
frameworks are seen as offering broad benefits even if costs remain mildly negative. When the
Problem 2 weighting is applied, PO2C again becomes the standout option (0.851 vs 0.634 for
PO2B and 0.611 for PO2A). This is driven by the fact that PO2C combines the strongest cost
outcome for operators (costs shift from negative at baseline to clearly positive (0.036) with
robust effectiveness and social gains. PO2-A still delivers the highest social score for operators
(0.051) under EW and P2, but its weaker cost performance means it ranks second overall once
the higher weight on effectiveness and cost dimensions is considered.
For national public authorities, preferences differ. All options again improve substantially on
the baseline, but PO2-C is the most attractive option when judged against the Problem 2
priorities. This is mostly drive by its cost performance, which outranks the other options and
the baseline. Under equal weighting, PO2-B already has the highest total score (0.215 vs 0.184
for PO2A and 0.136 for PO2C), largely because it nearly eliminates the negative cost gap while
improving effectiveness and social outcomes. Under the P2 weighting, PO2B’s advantage
becomes less pronounced than PO2-C, which improves cost burdens for public authorities with
respect to the baseline. By contrast, even though it outranks the other options across all
dimensions under P2 weights, PO2-A leaves authorities facing notable net costs, which lowers
its overall impact. In short, operators tend to strongly favour the EU-coordinated support
package in PO2-C, whereas public authorities place greater value on PO2-C than PO2-B,
specifically when applying P2 impacts, focused on digital sovereignty and resilience in the
public sector including through public procurement.
Table 61 Economic Operators and Public Authorities MAUT – Problem 2 (Source: Technopolis et al.
(2025))
Impact Costs Effectiveness Environmental Social Total
Economic Operators
Option EW P2 EW P2 EW P2 EW P2 EW P2
Baseline – P2 -0.369 -0.563 0.260 0.375 -0.038 -0.008 -0.003 -0.006 -0.149 -0.201
PO2-A -0.074 -0.044 0.244 0.597 0.036 0.007 0.247 0.051 0.453 0.611
PO2-B -0.009 0.113 0.181 0.474 0.036 0.007 0.195 0.040 0.421 0.634
PO2-C 0.036 0.180 0.238 0.615 0.045 0.009 0.234 0.047 0.552 0.851
160
For CSPs and AI service providers under Problem 2, the MAUT results show that all three
options perform markedly better than the baseline, but PO2-C emerges as the best-performing
configuration overall. The baseline records negative total scores under both equal weighting (–
0.102) and the Problem 2 weighting (–0.205), driven by cost penalties and limited effectiveness
and social effects. Under equal weighting, total scores improve to 0.411 for PO2-A, 0.392 for
PO2-B and 0.509 for PO2-C. When the Problem 2 weighting is applied the results are further
confirmed. PO2-C becomes the clear leader with a total of 0.812, ahead of PO2-B (0.601) and
PO2-A (0.567), while PO-A still performs well in terms of effectiveness and social outcomes.
For SMEs105 results indicate that all three options improve substantially on the baseline, with
PO2-C as the most favourable option under P2 weights and PO1-A under EW. Under equal
weighting, all options outperform the baseline, with 0.508 for PO2-A, 0.499 for PO2-C and
0.394 for PO2-B, mainly by increasing effectiveness, social and environmental scores. When
the Problem 2 weighting is applied, the contrast with the baseline becomes weaker: the baseline
total increases to 0.352, while PO2-C reaches 0.723, compared with 0.653 for PO2-A and 0.518
for PO2-B. In this scenario, PO2-C combines clearly better cost outcomes and environmental
impact, despite PO2-A has a slightly higher effectiveness and social score, indicating that more
ambitious measures offer SMEs the most advantageous balance under Problem 2.
Table 62. MCDA Cloud and AI Computing Service Providers and Data Centre Operators MAUT –
Problem 2 (Source: Technopolis et al. (2025))
7. ENVIRONMENTAL IMPACT ANALYSIS
The analysis of the environmental impact of data centres estimates the operational electricity
consumption and associated CO₂ emissions of projected data centre capacity in the EU over the
105 SMEs account for cloud and AI computing service providers.
National Public Authorities
Option EW P2 EW P2 EW P2 EW P2 EW P2
Baseline – P2 -0.318 -0.475 0.143 0.053 -0.125 -0.025 0.000 0.004 -0.300 -0.443
PO2-A -0.153 -0.180 0.157 0.392 -0.016 -0.003 0.195 0.039 0.184 0.247
PO2-B -0.024 0.075 0.146 0.318 -0.022 -0.004 0.116 0.023 0.215 0.413
PO2-C -0.030 0.111 0.129 0.307 -0.025 -0.005 0.063 0.013 0.136 0.425
Impact Costs Effectiveness Environmental Social Total
Cloud and AI Computing Service Providers
Option EW P2 EW P2 EW P2 EW P2 EW P2
Baseline – P2 -0.369 -0.563 0.262 0.363 -0.026 -0.005 0.031 0.000 -0.102 -0.205
PO2-A -0.074 -0.044 0.229 0.559 0.026 0.005 0.229 0.047 0.411 0.567
PO2-B 0.009 0.113 0.171 0.445 0.029 0.006 0.183 0.037 0.392 0.601
PO2-C 0.036 0.180 0.225 0.582 0.033 0.007 0.215 0.044 0.509 0.812
SMEs
Option EW P2 EW P2 EW P2 EW P2 EW P2
Baseline - P2 0.000 0.000 0.233 0.346 -0.039 -0.008 -0.028 -0.013 0.166 0.325
PO2-A 0.000 0.000 0.237 0.598 0.038 0.008 0.233 0.048 0.508 0.653
PO2-B 0.000 0.000 0.174 0.473 0.034 0.007 0.186 0.038 0.394 0.518
PO2-C 0.030 0.092 0.217 0.580 0.043 0.009 0.210 0.043 0.499 0.723
161
period 2025-2036 under the baseline and the three different policy options (PO1-A, PO1-B and
PO1-C). The calculation proceeded in four steps:
1. Use of installed IT capacity forecast under the baseline and the respective policy options
(MW)
2. Adjustment for utilisation and efficiency
3. Conversion of electricity demand into TWh
4. Application of EU electricity emission factors to derive CO₂ emissions106
7.1. Installed IT capacity
Annual projections of maximum IT power draw (MW) are taken as the starting point for each
scenario (Baseline, PO1-A, PO1-B, PO1-C), reflecting different trajectories of data centre
deployment. Public sector capacity is tracked separately but aggregated into total capacity for
environmental calculations. These projections imply compound annual growth rates of
approximately: 11% (Baseline); 12% (PO1-A); 13% (PO1-C); 15% (PO1-B)
7.2. Utilisation and efficiency assumptions
Installed IT capacity is converted into effective operational load using a utilisation factor of
63%, reflecting the weighted average of data centre utilisation rates reported by Technopolis et
al. (2025). These rates came from the CATI survey for colocation providers in primary markets
(coming out at 50% utilised), with desk research used to calculate the utilisation rate for
colocation providers in secondary and developing markets (45%) and hyperscaler sites (90%
utilised). This yields the “effective IT load” for each year and scenario.
Power Usage Effectiveness (PUE). Colocation PUE was calculated using information from
Data Center Map (coming out an overall average of 1.39) whereas hyperscaler PUE was
calculated using the latest reported hyperscaler figures for their EMEA sites from 2023 (coming
out at an average of 1.1). These figures were validated against market participants during
interview programmes and during both the interim and final workshop. Total facility electricity
demand is derived using scenario-specific PUE trajectories:
• Baseline improves gradually from 1.29 in 2025 to 1.23 in 2036
• PO1 scenarios assume faster efficiency improvements, reaching:1.18 (PO1-A), 1.12
(PO1-B), 1.05 (PO1-C) by 2036
These assumptions reflect accelerated uptake of efficient cooling, infrastructure optimisation,
and best-available technologies under policy intervention. For further details see section 2.3.4.
Total electrical load is therefore calculated as: Total DC load = Effective IT load × PUE
This produces declining electricity intensity per GW over time (TWh/GW), even as total
consumption increases due to capacity growth.
7.3. Conversion to annual electricity demand
Electricity consumption is computed using the standard conversion:
Annual electricity (TWh) = Load (GW) × 8,760 hours × 0.001, equivalently:MW × 0.00876 =
TWh/year.
This yields total annual electricity demand for each scenario, rising from approximately 99
TWh in 2025 to:314 TWh (Baseline), 339–340 TWh (PO1-A / PO1-C), 408 TWh (PO1-B) by
106 All policy scenarios use the same grid decarbonisation trajectories. Differences in emissions therefore arise exclusively
from differences in compute deployment and efficiency.
162
2036. Cumulative electricity demand over 2025–2036 ranges from 2 571 TWh (Baseline) to 3
028 TWh (PO1-B).
7.4. Emissions calculations
Electricity consumption is translated into CO₂ emissions using EU-average electricity emission
factors from the European Environment Agency. The emission factor declines from 0.25 kg
CO₂e/kWh in 2025 to 0.16 kg CO₂e/kWh in 2036, reflecting projected grid decarbonisation107.
Annual emissions are calculated as: CO₂ emissions (Mt) = Electricity demand (TWh) ×
Emission factor (Mt/TWh). The same emission factor trajectory is applied across all scenarios.
Despite improving energy efficiency (declining PUE and falling emissions per GW), total
emissions increase due to strong capacity growth. By 2036 baseline annual emissions are
expected to reach 50 Mt CO₂, PO1-A / PO1-C around 54 Mt CO₂, PO1-B reaches 65 Mt CO₂
At the same time, emissions intensity per GW declines by 41–49% across scenarios, illustrating
that efficiency and grid decarbonisation partially offset, but do not fully neutralise the impact
of capacity expansion. Efficiency improvements substantially reduce emissions per unit of
compute. However, rapid expansion of data centre capacity dominates overall outcomes. Policy
scenarios that accelerate deployment without parallel structural decarbonisation of power
supply materially increase absolute emissions. This highlights the importance of coordinating
compute expansion with grid investment and clean generation.
7.5. Limitations
This assessment provides an order-of-magnitude estimate of operational electricity demand and
associated CO₂ emissions from projected EU data centre capacity. Several limitations should
be noted:
• The analysis covers operational electricity consumption (Scope 2 emissions) only.
Embodied emissions from construction, hardware manufacturing, and network
infrastructure are not included. Life-cycle studies indicate these can be material,
particularly during rapid build-out phases, meaning total climate impacts are likely
understated.
• EU average grid emission factors are applied uniformly across scenarios. In practice,
incremental electricity demand from new data centres may be met by marginal generation
sources that differ from the average mix, possibly leading to higher or lower real-world
emissions. A constant utilisation rate is also assumed across scenarios and years. Actual
utilisation can vary widely over the asset lifecycle, which introduces uncertainty into
absolute electricity estimates.
• Power Usage Effectiveness improvements are modelled as smooth, deterministic paths.
Real-world efficiency gains depend on site-specific factors (climate, cooling technology,
density, heat reuse) and may be slower or more uneven than assumed, particularly for
retrofit facilities.
• The methodology does not distinguish between Member States or regions. Local grid
carbon intensity, cooling requirements, and permitting constraints can substantially affect
environmental outcomes, but are not captured in this EU-wide aggregate approach.
107 EEA’s indicator shows that the GHG intensity of power generation in the EU has been falling for decades and
was about 40% lower in 2024 than ten years before. See here: Greenhouse gas emission intensity of electricity
generation in Europe | Indicators | European Environment Agency (EEA)
163
Taken together, these limitations mean results should be interpreted as indicative ranges rather
than precise forecasts, and primarily as a tool for comparing scenarios on a consistent
methodological basis.
8. SO3 – HOW REALISTIC IS IT TO SET AN OBJECTIVE FOR EU PROVIDERS TO REACH A
MARKET SHARE OF 30% BY 2035?
Specific objective 3 sets the objective that, by 2035, this intervention should increase the market
share of European cloud and AI computing service providers in the European market to 30%.
The figure is not used for further calculation in this assessment, but it is worth reflecting on
whether this is realistic altogether: what would it take for this figure to become reality?
A first element is that cloud adoption varies significantly depending on the user. According to
Eurostat the adoption of cloud by businesses in 2023 is estimated at 45.2%, but while around
78% of the large enterprises use cloud services, only 59% of medium ones and 41.7% of small
organisations have adopted cloud108. The level of adoption of cloud in the public sector was
estimated at 30% in the context of PM15 and PM21. This gives a sense of where non-cloudified
markets are and which ones offer the biggest potential for EU providers to gain new markets
and improve their market share overall.
Looking at the different categories:
• Large customers could display a modest increase in the use of EU suppliers, causing
the EU market share to grow by 3.5% CAGR for this segment. Large organisations,
such as banks, telcos, IT or the manufacturing sector, are global industries that need to
ensure service continuity around the world and will, for an ample part, retain
international providers who can offer such geographic coverage: they are expected.
Growth drivers for EU providers include, for example, multi-cloud strategies, reduction
of vendor lock-in, and the shift towards sovereign solutions, especially for the more
critical workloads. to be among the largest drivers in cloud expenditures in the
following years109. For instance, some sources estimate the cloud expenditure in the
banking and financial sector accounts for 27% of the overall market expenditure in the
EU110. Other large customers, who are less global such as the aeronautic sector are
moving their critical workloads to sovereign cloud services111.
• SMEs could show a steady and more pronounced increase in the use of EU suppliers,
causing the EU market share to grow by 7% CAGR, driven by the fact that a large
percentage of SMEs are not yet on the cloud. With only 42 – 59% of SMEs (depending
on the size) having adopted cloud, SMEs represent a large greenfield opportunity for
EU cloud providers. Key enablers are lower switching costs and price sensitivity
favouring regional providers. To this end, the SME programme (PM23) can act as real
catalysers. Some companies like Odoo, a SaaS company offering a broad range of Open
Source enterprise applications, are specialised in this type of market.
• The public sector could increase their use of EU suppliers much more rapidly, causing
the EU market share to grow by 15% CAGR. This reflects the potential impact of joint
procurement practices as well as procurement of sovereign services. This sector exhibits
108 European Commission, Cloud Computing 109 IDC 110 Dimension market research 111 See Airbus
164
the strongest sovereignty imperative with directly applicable policy levers, such as
public procurement measures, sovereignty concerns, and NIS2 compliance, among
others.
Warning: the respective CAGR, though presented with plausible explanations, have been
picked to broadly match the desired outcome. The figures are presented to show a plausible
evolution, and do not constitute a forecast.
Synergy puts today’s market share of EU providers at 15%, a figure which, for the sake of this
exercise, it is assumed to be the same across all three segments. The evolutions over the three
segments would see the EU share grow as follows:
Table 67. European Cloud and AI computing service providers share of revenues - per customer segment
(%)112
Segment 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
Large enterprises 15.0 15.5 16.1 16.6 17.2 17.8 18.4 19.1 19.8 20.4 21.2
SMEs 15.0 16.1 17.2 18.4 19.7 21.0 22.5 24.1 25.8 27.6 29.5
Public sector 15.0 17.1 19.5 22.2 25.3 28.9 32.9 37.5 42.8 48.8 55.6
Today’s cloud market is assumed to be divided into three segments: large customers (LC)
amounting for 60% of the market113, small and medium enterprises (SME) being around 26%
and the public sector (PS) 14% of the market. Assuming this distribution remains the same, the
share of the market for EU providers could evolve as follows:
Table 67. European Cloud and AI computing service providers share of revenues – aggregated (%)
Segment 2025 2026 2027 2028 2029 2030 2031 2032 2033 2034 2035
Total Share 15.0 15.9 16.8 17.9 19.0 20.2 21.5 23.0 24.5 26.3 28.2
112 Shares indicate the proportion of cloud spend by each customer segment captured by European providers. Projections reflect
assumed growth in sovereign demand, procurement preferences in the public sector, and gradual diversification of enterprise
cloud sourcing. Total EU share is weighted by segment revenues. 113 This is in line with what several sources estimate of the cloud market share for large organisations in Europe. See for
instance GMInsights
165
9. TYPES OF PROCEDURES, PERMITS AND DATA CENTRE DEPLOYMENT TIMELINES PER MEMBER STATE
Table 68. Types of permits and timeline per MS selected (Source: Technopolis et al. (2025))
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Czech
Republic
Prague115 116 117
• Zoning and land allocation
for a data centre in Prague
generally takes 6-12
months
• The city’s zoning plan is
updated quarterly or semi-
annually. Incorporating
changes into the valid
zoning plan typically takes
1.5–2 months after
approval by the city
council
From submission to the relevant
Building Authority (stavební
úřad) for general construction
works to decision and issuance,
it takes 6-12 months
2 to 4 years for medium-sized
projects, but can extend to 5
years or more for large-scale
or high-capacity facilities
• Integrated into the permitting
procedure under the new
Building Act (effective July
2024)
• EIA is required for large
urban/industrial projects;
national screening applies
based on size/impact
• Screening and scoping take 30
days and the public
consultation about 30 days.
Then, the EIA authority issues
a binding opinion on
environmental impacts, which
is incorporated into the joint
building permit (~10 months
total)
24 - 60
114 When it comes to grid connection procedures, a 3 months deadline for receiving information on treatment of the connection request (i.e., the result of the permitting procedure) has been introduced by
2024 amendments to the Directive (EU) 2019/944. It also sets basic rules for third party access to electricity infrastructure in a non-discriminatory and transparent manner, applicable to all grid users, including
DCs. 115 https://geoportalpraha.cz/en/data-and-services/articles-and-projects/uzemni-plan 116 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/czech-republic 117 https://www.czechbusinessguide.com/content-of-book-permitting-construction-permitting-processes/
166
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Estonia
Tallinn118 119 120 121 122
• If a detailed spatial plan is
required <3 years
• If only design
specifications are needed
~30 days
From submission to the Tallinn
Urban Planning Department to
approval it takes 2-6 months
• For standard/small
connection (<2 MW) it
takes 6–12 months
• For large-scale/hyperscale
projects (>10 MW), it can
take 1–3 years or more
• The EIA is integrated into the
permitting process, and its
approval is a prerequisite for
the building permit
• The need for an EIA follows
the EU EIA Directive:
screening for Annex II projects
(urban development/industrial
zones)
• If needed, the Environmental
Board (Ministry of
Environment for large projects)
determines if an EIA is required
(30 days), the scoping and EI
study take 6-12 months. Then
there is a public consultation
(30 days) and the approval
comes 3-6 months after it
• The developer then must
submit again the building
permit application to the local
municipality after EIA
approval (1-3 months in
practice)
• Total duration is 11-24 months
6-24
118 https://www.tallinn.ee/en/ettevotjale/spatial-planning-procedures-and-timelines 119 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/estonia 120 Data collected through interviews 121 https://www.datacentermap.com/estonia/tallinn/datahouse@tallinn/ 122 https://investinestonia.com/estonia-has-the-most-advanced-data-center-in-the-region/
167
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
France
Paris123 124 125 126 127
• If the land is already zoned
for warehouse/data centre
use, allocation takes weeks
to months
• If a zoning change is
needed: 12–18+ months
• Large projects require
additional regional
approval, adding up to 3
months
• From the application
electronically to the Paris City
Hall (Basu office) to approval
takes average 3-6 months
• If the project triggers
environmental thresholds, an
EIA is required by DRIEAT
(Regional Directorate for the
Environment, Planning, and
Housing), bringing up the
duration to 6-18 months
• Standard grid connections
with RTE/Enedis takes 1.5–
3 years
• On utility-ready/pre-
connected sites, time is
under 1.5 years
• For large or complex
projects timeline may
exceed 3 years
• EIA must be done before the
building permit is granted, if
required
• EIA is required if emergency
generators exceed 50 MW
thermal output
• Initial screening takes 1 month,
whereas the EI study
preparation takes 612 months
• Once the public consultation
starts (30 days), the approval
comes 3-6 months after
• After EIA approval, the
developer submits the building
permit application to Basu
office
• Total duration is 13-32 months
24 - 60
Germany
Frankfurt128 129 130 131
• Specific “suitability areas”
(e.g., Sossenheim,
Rödelheim, Griesheim,
Gallus, Ostend,
• From the consultation with the
City Planning Department
(Stadtplanungsamt) and the
Construction Supervisory
• Average waiting time for a
new grid connection in
Frankfurt (and for
Germany)) is up to 7 years
• The EIA is integrated into the
building permit process and
must be completed before the
permit is granted
24 - 84
123123123 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/france 124 https://en.institutparisregion.fr/resources/publications/data-center-growth-and-proposals-for-regulation/ 125 https://www.linkedin.com/posts/stephanieelkhoury_datacentres-energy-connection-activity-7329060095772856320-ubcp 126 https://en.institutparisregion.fr/fileadmin/NewEtudes/000pack4/Etude_3025/Etude-DataCenter-english_2023_VF.pdf 127 https://www.edf.fr/sites/groupe/files/epresspack/9635/PR-Data-center-Vdef-1.pdf 128 https://frankfurt.de/english/service-and-city-hall/service-and-administration/municipal-offices/city-planning-department/construction-advisory-services 129 https://www.germandatacenters.com/en/news-en/detail/figures-of-the-month-data-center-ecosystem-germany-may-2025/ 130 https://www.lw.com/en/insights/energy-infrastructure-insights-data-centres-in-frankfurt-a-city-fit-for-the-future 131 https://www.dlapiper.com/en/insights/publications/real-estate-gazette/real-estate-gazette-infrastructure/power-grid-connections-for-data-centers-in-germany
168
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Fechenheim, Seckbach)
where company-
independent
(cloud/colocation) data
centres can be developed
(outside these areas, new
data centres are not
permitted)
• If land is not zoned or is in
a restricted/exclusion area
(usually the case) a change
or amendment to the
development plan
(Bebauungsplan) is
required
• Amending or creating a
new development plan
involves public
consultation,
environmental review, and
approval by city
authorities, governed by
the German Building Code
(BauGB)
• Timeline is 6-12 months if
there is land available in a
suitability area and 1-2
years if a zoning plan
amendment is required
Office (Bauaufsicht) to
approval, most German states
require the building permit to
be granted within 6 months
• 6–12 months is typical for
well-prepared, straightforward
projects in designated data
centre zones
• More complex or incomplete
applications, or those
requiring additional permits or
plan amendments, can take up
to 12–18 months
• Only if grid connection has
been pre-secured in
suitability areas, it can take
6-12 months
• EIA is triggered if backup
generators exceed 50 MW
thermal capacity
• If required, the steps involved
screening (1-2 months),
scoping and EIA report (6-12
months depending on
complexity), public
consultation (30 days), EIA
approval (by the regional
environmental authority, 3-6
months).
• After EIA approval, the
developer submits the building
permit application to the
Frankfurt Construction
Supervisory Office
(Bauaufsicht)
• The total timeline is 16-38
months
169
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Greece
Athens132 133 134
• 6–12 months is typical for
site selection and
confirming zoning if the
site selected is suitable
(industrial use)
• If zoning change or plan
amendment is needed,
timeline is 12–24 months
• The entire application and
issuance process is electronic
via the TEE e-permit platform
• 2-6 months is typical for the
building permit to be issued,
provided all documentation is
complete and the site is
already zoned for data centre
use
• The process can take 6-12
months or longer, especially if
the site is in a sensitive area
(archaeological or
environmental assessments),
or if documentation is
incomplete or needs revision
• Request must be submitted
to the Independent Power
Transmission Operator
(IPTO/ΑΔΜΗΕ) for high-
voltage connections, or to
the Hellenic Electricity
Distribution Network
Operator
(HEDNO/ΔΕΔΔΗΕ) for
lower voltage
• For most new data centre
projects in Athens, the grid
connection process—
including application,
technical studies,
permitting, construction of
substations or upgrades, and
final commissioning—takes
between 12-36 months
• For hyperscale or high-
capacity data centres (e.g.,
100 MW+), the timeline can
be closer to the upper end
(2–3 years)
• The EIA is a prerequisite for
the building permit. The
building permit application
cannot proceed until the EIA is
approved
• Data centres with ≥2 MW
capacity fall under
environmental regulations and
EIA is likely to be triggered if
capacity is ≥20 MW
• The Ministry of Environment
and Energy determines if an
EIA is required based on
project size, location, and
potential impacts (1-2 months)
• The following steps include
scoping (1-3 months), EIA
report (6-12 months), public
consultation (1-3 months) and
approval (3-6 months post-
consultation)
• Total timeline is 14-32 months
18 - 56
132 https://en.mitos.gov.gr/index.php/%CE%94%CE%94:Issue_of_Building_Permits_(e-adeies) 133 https://inconde.com/building-permits-in-greece/ 134 https://en.mitos.gov.gr/index.php/%CE%94%CE%94:Issue_of_Building_Permits_(e-adeies)
170
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Ireland
Dublin135 136 137 138 139 140 141 142
• If the land is already
zoned, it takes 6–12
months for site selection,
plus 2–4 months for
planning permission
• If a zoning change is
required, the process can
extend to 1–2 years or
more
• Most new data centres in
Dublin are built on pre-
zoned “HT – High
Technology” or industrial
lands
Obtaining a building permit with
the Dublin city Council takes 2–
4 months for planning
permission (longer if appeals or
further information are needed),
with building control certificates
adding another 2–3 months
(often overlapping with
construction preparation)
• Since late 2021, EirGrid
(Ireland’s transmission
system operator) has
imposed a de facto
moratorium on new data
centre grid connections in
the Dublin area (till 2028)
• Outside Dublin, the
connection take’s 1-3 years
(new policy from The
Commission for Regulation
of Utilities, CRU)
• The EIA is a mandatory part of
the planning process and must
be completed before a building
permit (planning permission) is
granted
• A GHG permit required for
generators >20 MW and a full
EIA is typically required for
larger projects
• Screening is performed by the
Dublin City Council, followed
by the preparation of the EI
report (6-12 months), public
consultation (1-3 months) and
final decision (8 weeks
statutory timeline)
• Total timeline is 10-24 months
• 24 (fastest
case, out of
Dublin)
• Indetermined
for Dublin
Italy
• If land already zoned for
data centre use: 6–8
months (with due
diligence and planning
approval)
• Permitting and approvals
(planning + building permit)
takes in average 6–18 months
• If the site is already zoned and
documentation is complete,
• Standard grid connection: 2
to 4 years, due to a single
national grid operator
(Terna) processing all
• The EIA is a prerequisite for
the building permit
• EIA is required for generators
>50 MW thermal output or for
24 - 60
135 https://consult.fingal.ie/en/system/files/materials/22444/26992/BMC%20Vantage%20FCC%20Dev%20Plan.pdf 136 https://www.dublincity.ie/residential/planning/planning-applications/make-planning-application/planning-process-application/planning-application-timelines 137 https://www.dublincity.ie/residential/planning/planning-applications/make-planning-application/planning-process-application/about-planning-process-application 138 https://www.mhc.ie/latest/insights/permitting-irish-data-centres-what-you-need-to-know-1 139 https://www.dublincity.ie/residential/planning/building-control/about-building-control 140 https://www.rte.ie/news/business/2022/0116/1273819-data-centres-eirgrid/ 141 https://www.capacitymedia.com/article/29ot42ikril15nn8s4c2t/news/new-data-centres-in-dublin-face-power-connection-delay 142 https://www.bisnow.com/dublin/news/data-center/regulator-opens-up-possible-data-centre-pathway-amid-fears-dublin-has-been-left-behind-129587
171
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Milan143 144 145 146
• If zoning change or a plan
amendment is needed (12–
18+ months)
• Brownfield site (with
zoning) can happen
in 6–8 months (incentive)
the process can be on the
shorter end (6–8 months)
• If zoning changes or complex
planning are required: 12–18
months
requests, a large backlog
and the need for upgrades
• Brownfield sites, pre-zoned
or fast track projects can get
the connection in 1.5-2
years
• On the other hand,
large/complex projects can
take 3-5 years
a total project capacity >150
MW
• The Municipality of Milan
determines the requirement (1-
3 months) and, from there, the
preparation, public
consultation, approval and
building permit review can add
up to 17-32 months
Netherlands
Amsterdam147 148 149 150 151 152
• If the land is already zoned
and available: 2–6 months
(plus site due diligence)
• If a zoning plan
amendment is required: 6–
12+ months
• Most new data centre
developments are subject
to strict municipal controls
and sustainability
requirements
• If the site is already zoned for
data centre use: 8–14 weeks;
with an environmental permit
(≥15 MW), allow up to 26
weeks
• If a zoning amendment or
deviation is needed: 6–12+
months added before the
permit can be reviewed
• All projects must comply with
the Environmental Act
(Omgevingswet), including
• No new grid connections for
data centres in Amsterdam
are expected before 2030
unless the project is
essential for the city and
meets strict criteria
• Even fully built data centres
may remain idle if they lack
a pre-secured grid
connection.
• Outside Amsterdam e.g.,
Rotterdam, Westland, grid
• The EIA process is integrated
into the permitting framework
and cannot be conducted in
parallel with the building
permit
• An environmental permit is
required at 15 MW and a full
EIA is likely for larger
hyperscale projects (>70 MW)
• The competent authority
(municipality or
Omgevingsdienst
• 12 (fastest
possible case,
out of
Amsterdam)
• Indetermined
for
Amsterdam
143 https://www.assoimmobiliare.it/wp-content/uploads/2024/03/Colliers_Data-Center-Snapshot-2024_Italy__.pdf 144 https://datacentrenews.uk/story/virtus-announces-new-data-centre-plans-for-milan-italy 145 https://www.cliffordchance.com/content/dam/cliffordchance/briefings/2025/05/data-centres-in-italy.pdf 146 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/italy 147 https://www.stibbe.com/publications-and-insights/increasing-control-of-data-centre-locations 148 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/netherlands 149 https://cms-lawnow.com/en/ealerts/2022/11/netherlands-prohibits-creating-hyperscale-data-centres-until-national-guidelines-are-passed 150 https://openresearch.amsterdam/image/2020/4/30/the_governance_of_land_use_amsterdam_2017.pdf 151 https://www.haskoning.com/en/newsroom/blogs/2023/navigating-dutch-data-centre-challenges-and-opportunities 152 https://www.portofamsterdam.com/sites/default/files/2020-06/guide-permit-application.pdf
172
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
• For hyperscale data centres
(>10 hectares/70 MW):
New developments are
effectively banned in
Amsterdam and most of
the Netherlands
energy efficiency and
emissions standards
• Engaging with the city and OD
NZKG early can clarify
requirements and avoid delays
connection is possible
within 1–3 years
Noordzeekanaalgebied, OD
NZKG, for larger projects)
determines if an EIA is
required, by submitting a
screening request through the
Omgevingsloket
(environmental permit portal).
This takes 1-3 months
• The EI report, public
consultation and building
permit review in total take from
10-24 months
Poland
Warsaw153 154 155 156 157 158
• If the site is already zoned:
1–2 months
• If a zoning permit is
needed: 3–9+ months
• If an environmental
decision is required: add
up to 6 months.
• Land acquisition can
proceed in parallel, but
legal title is needed before
the building permit stage
• Obtaining a building permit
takes 65 days to 6 months
• The process is managed by the
district authority (starosta)
• Starts by submitting request
to the network operator
(e.g., PGE Dystrybucja,
Innogy Stoen Operator)
• In general, the process takes
from several months–1 year
• Full grid connection (total
process) is 2–3 years
• In large projects (100
MW+) with major
upgrades/delays: 3+ years
• The EIA is a mandatory
preliminary step that must be
finalized before applying for
the building permit
• The EIA requirement follows
the EU EIA Directive and there
is a screening for industrial
projects based on size/impact
• The screening is submitted to
the Regional Environmental
Protection Authority i.e.,
Mazovian Voivodeship and
24 - 45
153 https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/poland 154 https://www.legal500.com/developments/thought-leadership/a-planning-permit-can-be-obtained-in-poland-if-land-to-be-developed-is-located-within-a-development-zone/ 155 https://dwfgroup.com/en/news-and-insights/insights/2024/9/general-zoning-plan-for-warsaw 156 https://consultant.net.pl/en/consultant-article/how-to-obtain-a-building-permit-in-poland-1 157 https://www.gtlaw.com/en/insights/2024/4/data-centres-in-poland-planning-and-electricity 158 https://natlawreview.com/article/data-centres-poland-planning-and-electricity
173
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
from there, the total process can
take 11-24 months
Romania
Bucharest159 160 161 162 163 164 165 166 167
• Zoning and land allocation
for a data centre in
Bucharest can take a few
months for small plots
(<3,000 sqm) with no PUZ
(Zonal Urban Plan)
required, or 3–6 months if
a valid PUZ is already in
place
• For larger plots requiring a
new PUZ, the process can
take 1–2 years or more but
is currently subject to
major delays due to legal
and administrative
gridlock
• The Romanian Construction
Law stipulates that building
permits must be issued within
30 days. In practice, it takes 6-
12 months
• Building permits are issued by
either the General Mayor of
Bucharest or District Mayors,
depending on project type and
location
• Romania’s energy regulator
(ANRE) has recently
overhauled grid connection
rules. As of June 2025, only
fully documented projects
are placed in the queue, and
developers must provide a
financial guarantee before
receiving a grid connection
permit
• Timeline is 1.5–3 years for
grid connection, from
application to energization.
• The EIA is a prerequisite for
the building permit
• The EIA requirement follows
the EU EIA Directive and there
is a screening for industrial
projects based on size/impact
• From sending the screening to
the Bucharest Environmental
Protection Agency, the total
timeline for an EIA is 11-24
months
18 - 36
Spain
• The municipality
determines the specific
• Granted based on the technical
project submitted by the
developer. 3-12 months (3-4
• Submitting an access and
connection request to Red
Eléctrica de España (REE),
• An Environmental Impact
Assessment (EIA) may be
required (e.g., projects > 15
12 - 48
159 https://www.businessforum.ro/real-estate/20241129/investors-in-data-centers-enter-race-for-land-plots-in-romania-1111 160 https://www.property-forum.eu/news/data-center-investors-enter-race-for-land-plots-in-romania/19253 161 https://www.arl-international.com/sites/default/files/2023-11/Fact%20sheet%20for%20planning%20levels_local_Zonal%20Plan.pdf 162 https://cms-lawnow.com/en/ealerts/2022/08/bucharest-tribunal-rules-against-the-city-s-main-zoning-plan 163 https://buildecon.blog/2025/03/20/bucharests-drop-in-residential-permit-and-completion/ 164 https://www.cristinatudor.ro/en/post/unlocking-romania-s-building-permit-process-your-gateway-to-successful-construction-projects 165 https://www.nglsymbio.com/wp-content/uploads/construction-permits-in-cee-practical-guide-1.pdf 166 https://horizons.legal/wp-content/uploads/2025/02/CONSTRUCTION-PERMITS-IN-CEE-01-25-1.pdf 167 https://www.pv-magazine.com/2025/06/16/romania-updates-grid-connection-rules/
174
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
Madrid168 169 170 171 172
planning designation
(industrial use)
• Amendments/new
planning involves, detailed
planning, municipal
review, and sometimes
higher-level regional
approvals (6-12 months on
top)
• There are no land plans or
zones that automatically
authorise data centres in
Madrid; approvals are
granted on a case-by-case
basis
months in straightforward
cases)
• First Occupancy License
(Licencia de Primera
Ocupación): obtained after
construction, confirming the
building complies with the
building permit. Involves
inspection by municipal
officials (3-6 months).
• Activity and Opening
Licenses (Licencia de
Actividad y Licencia de
Apertura): confirms
regulatory compliance for the
intended use. “Declaración
Responsable” (responsible
declaration), the license is
immediate or nearly
immediate. For activities
requiring a full license
procedure (for significant or
“qualified” activities), the
the national transmission
system operator
• Undergoing feasibility
studies, technical reviews,
and regulatory approvals
• Construction of new
substations or high-voltage
lines if required
• Final commissioning and
energization
• <12 months If the site is
already near a substation
with available capacity, or if
a developer has pre-secured
grid access
• 2–4 years is the benchmark
(especially for data centres
> 100 MW)
MW), depending on the size
and location of the project. This
can add several months to the
process, especially if public
hearings or additional studies
are needed
• The average is 11-24 months,
including the determination
request, EI study, public
consultation and DIA issuance)
• DIA must be issued before the
building permit is granted
168 CMS. (2025). Expert Guide on Real Estate Data Centre Consenting: Spain. Retrieved June 17, 2025, from https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-consenting/spain 169 Aeclu. (n.d.). Cuánto tarda la licencia de primera ocupación en Madrid. Retrieved June 17, 2025, from https://www.aeclu.com/blog/noticias/cuanto-tarda-la-licencia-de-primera-ocupacion-en-madrid/ 170 Madrid Licencias. (n.d.). ¿Cuánto tarda una licencia de apertura y actividad? Retrieved June 17, 2025, from https://www.madridlicencias.com/blog/cuanto-tarda-una-licencia-de-apertura-y-actividad/ 171 Strategic Energy Europe. (n.d.). Spain DC: Centros de datos y renovables. Retrieved June 17, 2025, from https://strategicenergy.eu/spain-dc-centros-de-datos-y-renovables/ 172 Montel News. (2024, March 21). Spain power demand may get 3 GW boost from data centres. Retrieved June 17, 2025, from https://montelnews.com/news/db842f28-b5f8-4c75-b914-cfd2792d219a/spain-
power-demand-may-get-3-gw-boost-from-data-centres
175
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
process typically takes 3-6
months
Sweden
Stockholm173 174 175 176
• Land designated as
“technical facility” use in
the municipal zoning plan
(detaljplan): 0 months
• If not, new/amended
zoning plan is needed
(municipal approval +
public consultation): +12-
24 months
• Building permit (bygglov)
average is 2-3 months (may
take longer for complex
projects)
• After approval, a “start
clearance” is needed before
construction can begin, and a
final clearance before the
facility can be used
• Initiated in parallel or even
before the permitting
process by engaging with
local grid owner (often
Ellevio or Vattenfall)
• Technical and feasibility
studies, land rights and
permits, construction and
commissioning take 2-4+
years
• Data centres themselves
usually do not require a
separate environmental permit
(integrated within overall
permitting)
• if the project involves
significant water use or heat
recovery, additional
environmental permissions
may be needed (in line with the
EU EIA directive)
• The steps are screening and
scoping (60 days), preparation
of EI statement, submission and
completeness check, public
consultation (30 days).
• The permit authority (often the
Land and Environment Court
or Environmental Assessment
Delegation) may hold hearings
or written consultations with all
parties
• 24 (average
case)
• 48+ (worst
possible case)
173 CMS. (2025). Expert Guide on Real Estate Data Centre Consenting: Sweden. Retrieved June 17, 2025, from https://cms.law/en/int/expert-guides/cms-expert-guide-on-real-estate-data-centre-
consenting/sweden 174 City of Stockholm. (n.d.). Building permits in English. https://bygglov.stockholm/in-english/ 175 Node Pole. (n.d.). Node Pole Report (Sweden). https://8866495.fs1.hubspotusercontent-na1.net/hubfs/8866495/Node%20Pole%20Report%20(Sweden)%20-%20FINAL.pdf 176 Rigord, N. (2024). EIA effectiveness in Sweden: a Stockholm case-study [Master's thesis, KTH Royal Institute of Technology]. DiVA Portal. http://www.diva-
portal.org/smash/get/diva2:1898845/FULLTEXT01.pdf
176
Zoning & land allocation Building permit
Utilities & grid
connection114 Environmental permit/s
Average
timeline
(months)
• Timeline is 6-18 months and up
to 24 months for complex cases
• The building permit cannot be
granted until the environmental
permit (including the EIA) is in
place
ANNEX 5: COMPETITIVENESS CHECK
1. OVERVIEW OF IMPACTS ON COMPETITIVENESS
Dimensions of
Competitiveness
Impact of the
initiative
(++ / + / 0 / - / -- / n.a.)
References to sub-sections of
the main report or annexes
Cost and price
competitiveness + Section 6.1.5, Annex 4
International
competitiveness + Section 6.1.5, Section 6.1.6
Capacity to innovate ++ Section 6.1.1, Section 6.1.4
SME competitiveness + Section 6.1.1
SYNTHETIC ASSESSMENT
Cost and price competitiveness
CADA will impact both costs and price competitiveness. Building data centres is a capital-
intensive and complex activity that requires multiple permits and investments in its whole
lifecycle (design, construction and operation). The facilitation of access to land and
streamlined grid connections along with fast-track permitting procedures is expected to
reduce total cost of ownership for data centre operators (PO1-B). By shortening approval
times, simplifying and harmonising requirements the cost of the immobilised capital as
well as transaction costs will be lowered. The introduction of a national facilitator for data
centre projects will allow to save administrative time, very important for SMEs. The result
of the preferred package is expected to increase the number of providers, enhancing
competition and reducing market concentration.
Under the preferred package (PM8 and PM9), research and innovation will fund novel
technologies (e.g. advanced cooling, AI-based energy optimization, etc..) to achieve much
more efficient data centres in terms of use of energy. This innovation is expected to have
positive spillovers across the European cloud and AI value chain. Businesses benefitting
from the development of such innovations can achieve a lower total cost of ownership
(TCO), and transfer part of these gains to customers through more competitive pricing for
access to data centre capacity. This is particularly important for EU companies as they
operate at smaller economies of scale than their international counterparts.
With respect to cloud and AI computing services, PO2-C under the preferred package
implies additional costs for cloud and AI computing service providers to be audited with a
view of achieving a sovereignty qualification which allows for providing services to public
administrations in highly critical use cases. This may result in an increase of prices for
sovereign cloud and AI computing services, as the cost of adjusting (e.g. due to the creation
of new legal constructs) may be passed on to the customers in the form of a price premium.
This will be the case until the market has stabilised, and the providers have recuperated the
investment undertaken (short-term). The risk sovereignty mechanism is expected to be an
important cost aspect for SMEs due to the effort needed to obtain and maintain the audit.
Over time, however, a harmonised and recognised mechanism valid across the EU will
reduce transaction costs and increase trust and market access.
178
Joint procurement and federation are likely to have a two-fold cost impact. First, they will
create economies of scale on the demand side, reducing unit prices for procured services.
Second, they will generate predictable demand signals, which will decrease the risk
premiums of the suppliers. Joint procurement and federation can reshape the market by
creating a large pool of aggregated demand open to multiple providers, notably allowing
smaller providers to compete for specialised services. This will result in an overall
improved situation of the public sector using cloud and AI computing services, mitigating
the risks stemming from dependence on a small number of providers. An additional
element is transparency: through the access to large contracts, smaller cloud and AI
computing service providers can retain customer acquisition costs (CAC) and amortise
fixed development expenditures. This will result, in the medium term, in a more stable cost
structure and greater price competitiveness.
The creation of a cloud and AI toolbox is expected to promote further competition by
allowing European providers to better integrate their service offerings. Currently, there is
a large fragmentation across national markets and cloud, and AI computing services are
not comparable which create uncertainty. This is expected to have a positive impact in
SMEs.
The directly affected sectors are data centre operators and the providers of cloud and AI
computing services. While both sectors will need to undergo some adaptation in the way
they work, this is not expected to not cause a disruption and will mostly result in benefits,
such as savings in construction time, aggregated offers, and more security and resilience.
Indirectly affected sectors are on one end of the value chain the public sector and the
customers of the cloud and AI computing services market and on the other, the
semiconductor industry, energy systems and utilities.
Cost-price impacts of the preferred package (PO1-B + PM8 + PM9 + PO2-C) are expected
to be different depending on the problem driver targeted. Faster permitting and access to
land as well as clear procurement criteria will cause effects in the short term (0-2 years).
In the medium term (2-5 years), the effects will arise from the go-to-market of the research
and innovation funds, the putting in place of the federation mechanisms, which will allow
an intensification of competition, a decrease of prices and efficiency gains. Long-term
impacts (>5 years) will come from the combination of selected locations for computing
infrastructure and the access to energy and other natural resources, notably by gaining
location advantage.
The preferred package shall notably improve competitiveness. It implies a learning curve
to implement some of the policy measures proposed (e.g. energy efficiency technologies,
research and innovation, federation, interoperability) and network effects (e.g. open
source). However, one of the major risks is on the implementation of permitting and access
to grid and how they are managed (e.g. sequentially, in parallel) as these have an effect on
the structural costs and durable price competitiveness.
International competitiveness
CADA strengthens the competitive position of data centre operators active in the EU and
of European cloud and AI computing service providers.
In the case of data centre operators, it will reduce structural cost disadvantages stemming
from the fragmented permitting regimes existing nowadays (see Annex 4, section 8). Data
179
centre operators in the EU will be able to expand their capacity faster at lower cost.
Simplification of administrative procedures will reduce entry barriers, notably for mid-
sized firms, favouring the scalability of European players (PM4, PM5). Both EU and non-
EU players, however, will benefit from vast economies of scale.
The deployment of computing infrastructure across the EU, and notably the deployment
of sovereign cloud and AI computing services, will allow the EU to decrease its
dependencies on data infrastructure located outside of the EU and its dependence on non-
European providers of cloud and AI computing services, while increasing the economic
security of the EU in the field of cloud and AI. The creation of a common definition of
sovereignty, with clear criteria, along with the mandatory sovereignty risk assessment
(PM21), will boost the trust and credibility of sovereign services, allowing their providers
to differentiate themselves in markets where security, sovereignty and compliance are key
items. Providers of sovereign services will hence be able to compete on trust, transparency
and robustness rather than on price alone.
The establishment of research and development funding, addressing technological
asymmetries with global incumbents will allow to develop next-generation technologies
for energy efficiency, AI and other aspects (PM8, PM9). Currently, non-EU providers
often lead with proprietary solutions. The use, reuse and promotion of open source in
public administrations, will accelerate innovation allowing EU companies to compete in
terms of performance, energy efficiency, and novel offers (PM20).
All measures under this initiative are conceived to be in line with the EU’s international
trade obligations. The initiative looks to incentivise local investment and aggregate
demand by anchor customers, ensuring that competitive advantage is grounded in trust and
operational excellence rather than on market power.
Finally, the initiative seeks to enable European providers to gain market share at EU and
international level, by equipping them with the necessary mechanisms to enhance the scale
and scope of their service offering. In the mid to long term, the levers proposed in this
initiative (e.g. fast-permitting, national facilitator, funding mechanisms, joint procurement,
federation) to solve very clear challenges of today (e.g. sustainability, skills, leverage of
open source, aggregation of demand, sovereignty) will allow European providers to expand
their market presence, in the EU and beyond.
Capacity to innovate
CADA goes directly at the heart of improving the capacity to innovate (PM8, PM9) and
the improvement of skills (PM23). The funding mechanisms (PM8, PM9), complemented
with the cloud and AI adoption scheme targeted towards SMEs and the skills certification
(PM23) will boost the capacity to innovate at product, service and process level. The
spillover effects of research funds, where a myriad of EU stakeholders can participate (e.g.
academia, SMEs, large enterprises), allow also for the initiative to anchor intellectual
capital and innovation capabilities within the EU.
Other aspects such as the promotion of open source in the public sector enlarge the amount
of resources available for future innovative products and services (PM20). Open source
allows to balance open innovation with adequate intellectual property protection. EU
public sector bodies are hence able to reuse and extend these solutions under clear legal
certainty, lowering entry costs for smaller innovators.
180
Finally, the use of well-defined sovereign criteria (PM21) are expected to create a demand-
pull effect in the market, where companies will invest in R&D in order to develop
technologies that will equip them to meet these criteria in a more efficient way.
SME competitiveness
The initiative directly affects SMEs, mostly in a positive manner. By removing
bureaucratic friction and accelerating permitting times, boosting innovation capacity and
mechanisms to adopt or migrate to cloud and AI and allowing to share resources and
services, the initiative levels the playing field between smaller and larger providers by
reducing fixed regulatory costs that otherwise would deter SMEs from participating in the
operation of data centres and the provision of cloud and AI computing services. The
measures allow SMEs to compete on merits and specialization, not marketing budgets. The
audit scheme, however, generates a burden for SMEs, both to obtain it and to renew it on
a yearly basis. But this can have a double effect. First, since it will be valid throughout the
EU, it may open new markets to providers that now are out of reach due to other
requirements (e.g. national certification schemes or labels), notably in the context of public
procurement. Second, being audited under clear sovereign criteria may be used by SME
providers to capture part of the private sector market dealing with highly critical data (e.g.
defence) and to overcome the buyers’ bias towards solutions offered by larger providers.
The minimisation of the burden on SMEs is one of the goals of this intervention, as well
as simplification, proportionality, harmonisation, supportive ecosystems and open access.
All these considerations stay at the core of the initiative, by for instance, transforming
compliance into an innovation catalyst, by providing compensatory enablers (e.g. grants,
national support measures).
For more information see Annex 6 SME check.
181
ANNEX 6: SME CHECK
OVERVIEW OF IMPACTS ON SMES
Relevance for SMEs
Based on SME filter and the (first) ISG discussion, this initiative is relevant for SMEs177
(1) IDENTIFICATION OF AFFECTED BUSINESSES AND ASSESSMENT OF RELEVANCE
Are SMEs directly affected? In which sectors?
SMEs are directly affected, though in different ways depending on their role in the data
centre, cloud and AI value chain. Firstly, on the limited and geographically concentrated
availability of computing capacity (P1), the affected SMEs are those companies building
and operating data centres, which encompass data centre operators, cloud service
providers that build and operate their own data centres, construction and real estate
companies. Building data centres is a capital-intensive activity and hence the number of
SMEs affected is limited. Secondly, on the dependence on cloud and AI computing
services supplied by non-European providers (P2) there are two strands of SMEs
stakeholders affected by the current initiative. First, cloud and AI computing service
providers, that is, those organizations that develop and offer their services to business
and governmental customers and on the other, SMEs that use cloud and AI computing
services.
Estimated number of directly affected SMEs
SMEs directly affected by the initiative, i.e. involved in building or operating data
centres are approximately 150 across the EU178. Furthermore, the Technopolis et al.
(2025) mapped179 existing cloud service providers in the EU (284) and the results yielded
that around 60% of the identified providers are SMEs or startups headquartered in the
EU (175) whereas including non-EU headquartered, the number amounts to 200180. The
same mapping exercise was carried out for AI service providers but the results to
quantify the affected organizations are less conclusive.
Beyond cloud and AI computing service providers, the policy measures under
consideration, notably the cloud and AI adoption scheme for SMEs, also affect ICT SME
companies181 (65 000) that wish to use cloud and AI computing services. More broadly,
177 https://ec.europa.eu/docsroom/documents/63274. 178 Technopolis et al. (2025), “Study: cloud and AI” 179 The mapping carried out by Technopolis et al. (2025), “Study: Cloud and AI” included a manual search cross checked
and complemented by existing market reports, entities participating in the Alliance for industrial data, edge and cloud,
and MS reports (e.g. qualified entities by SecNumCloud). 180 In contrast, the study identified 59 large cloud service providers headquartered in the EU. 181 According to Eurostat, there are closed to 1.3 bn ICT companies, divided as follows: 1.2 bn are micro, 53,000 are of
small size and 12,000 are of midsize. For this domain, only small and midsized companies have been taken into
consideration.
182
SMEs in general (1.6 bn)182 will also be affected by the broader availability of cloud and
AI computing services, and through the cloud and AI adoption scheme.
Estimated number of employees in directly affected SMEs
The directly affected SMEs, i.e. those building or operating data centres and those who
would increase their user of cloud and AI computing services, are likely employed
between 50 and 250 people each, as they are typically small/medium-sized, highly
specialised technology firms. With roughly 300 such SMEs across the EU, the total
number of employees directly affected can be estimated at around 75.000 people EU-
wide.
Are SMEs indirectly affected? In which sectors? What is the estimated number of
indirectly affected SMEs and employees?
The value chain consists of three types of SMEs: data centre operators, cloud and AI
computing service providers and cloud and AI computing service users. Service
providers deploy their services on a data centre and offer the services to the user.
Depending on the policy measure, service providers are either directly targeted or
indirectly.
In the context of the first problem, the limited and geographically concentrated
availability of computing capacity, the policy measures directly affect data centres
operators and indirectly affect cloud and AI computing service provider. A broader
availability of data centres will (indirectly) affect positively cloud and AI computing
service providers as they will be less subject to a shortage of infrastructure where to
deploy their applications and services or where they can train or inference their models.
This is particularly the case for European SMEs leading in AI, which face significant
challenges in obtaining the resources, funding, and market opportunities they need, often
finding themselves pushed into early commercial dependencies with infrastructure
providers. The number of cloud and AI computing service providers aligns with the
values indicated above.
In the context of the second problem, the dependence on cloud and AI computing
services supplied by non-European providers, the policy measures directly affect SMEs
that provide cloud and AI computing services and indirectly affect those SMEs that use
these services. Here, the extent of SMEs indirectly benefiting from an increase cloud
and AI update is virtually all of them as for example such services are increasingly
necessary to declare taxes, produce electronic bills or manage the day-to-day operations
of a business, even for single persons’ businesses. The same logic applies for the subset
of SMEs that need sovereign cloud and AI computing services with a higher level of
protection.
182 SME Performance Review - Internal Market, Industry, Entrepreneurship and SMEs estimates the existence of 26 bn
SMEs, out of which more than 24.5 bn are considered micro.
183
(2) CONSULTATION OF SME STAKEHOLDERS
How has the input from the SME community been taken into consideration?
Input from the SME community has been taken into consideration through the public
consultations (survey and call for evidence), targeted consultations by the study team,
stakeholder workshops, and continuous dialogue with industry associations, and
innovators active in AI and cloud technologies. The following SME respondents have
participated to the:
• Public consultation: 17 SMEs (or 25% of total business replies)
• Call for Evidence: 25 SMEs (or 45% of total business replies)
• Targeted consultations by the study team
• Stakeholder workshops
• Bilateral dialogues
Are SMEs’ views different from those of large businesses?
Yes. SMEs’ views differ significantly from those of large businesses, mainly due to their
resource constraints, market position, and regulatory capacity. While large companies
generally support stronger cloud and AI ecosystems, they often focus on scaling
infrastructure, interoperability, and competitiveness at a global level. In contrast, SMEs
emphasize accessibility, affordability, and simplicity as they need easier access to
compute resources, clearer guidance on compliance, and reduced administrative and
financial burdens.
(3) ASSESSMENT OF IMPACTS ON SMES183
What are the estimated direct costs for SMEs of the preferred policy option?
Qualitative assessment
PO1-A improves SME access to information through clearer guidelines, reducing costs
from information gaps and helping SMEs voice concerns collectively. PO1-B simplifies
permitting, cutting legal and administrative burdens and lowering entry barriers to the
data centre value chain. PO1-C has similar effects but may feel more distant due to EU-
level approval; however, added public R&D and deployment funding strengthens SME
participation in advanced projects without heavy capital needs. Potential direct
administrative costs under the preferred policy option among these, i.e. PO1-B with PM8
and PM9, would be related to the time required to apply for EU or national funding
support.
PO2-A provides clarity on sovereign cloud and AI computing services without extra
administrative load, while interoperability and the EU sovereignty conference boost
SME visibility. PO2-B lowers verification costs via validity throughout the EU and
supports SME cloud and AI adoption, with procurement criteria and vendor-neutral
training improving SME competitiveness. PO2-C delivers the strongest benefits:
although mandatory audit adds some cost, EU-wide validity expands market reach,
open-source access reduces barriers, funding programmes favour SMEs. Overall, PO2-
C and the SME cloud and AI adoption scheme enhance competitiveness, reduce upfront
183 The costs and benefits data in this annex are consistent with the data in annex 3. The preferred option includes the
mitigating measures listed in section 4.
184
costs, and support wider SME digital transformation. Similarly, as above, quantifiable
administrative costs for SMEs stem from the time required to apply to receive funding
under PM23.
Quantitative assessment
Under PM23, within PO2-C, SMEs are expected to incur in modest administrative costs
related to their need to apply to receive the funding. The effort to prepare a project
proposal is estimated to be 2 persons for 10 days (20 staff days), with total costs
estimated at EUR 55 373 928 (NPV, 10-years), considering over 60,000 SMEs applying
to receive funding.
What are the estimated direct benefits/cost savings for SMEs of the preferred
policy option184?
Qualitative assessment
For PO1-B, SMEs would see direct benefits mainly through reduced legal, consultancy,
and administrative costs. Streamlined permitting makes procedures easier to navigate,
which cuts the time and external support SMEs typically need to comply with complex
rules. Faster approvals also shorten project timelines, lowering costs and making market
entry cheaper. With public funding under PM8 and PM9, SMEs save directly by
reducing the upfront capital normally required for R&D, innovation, and deployment
projects. Higher funding rates for SMEs lower financial risk and make participation in
advanced initiatives more affordable. Funding also helps SMEs join larger consortia
more easily, lowering the financial barriers to entering markets that typically favour
large players.
Under PO2-C, SMEs gain the largest net cost savings. Validity throughout the EU of
audits means an SME audited in one Member State avoids repeating costly verification
processes elsewhere. Support measures for adopting cloud and AI computing services
lower training, integration, and compliance costs. Together, these reduce technology
investment, administrative overhead, and business development expenses.
Quantitative assessment
Under PM23, within PO2-C, SMEs are expected to benefit directly from EUR 328 412
809. This policy measure puts forward a targeted scheme to provide financial support to
SMEs for adopting cloud and AI computing services to increase their productivity and
competitiveness. Most of the yearly budget considered for this measure, i.e. EUR
38.500.000, is dedicated to supporting the design and planning phase of cloud and AI-
based transformation projects for SMEs. The grants are fixed amounts that the SMEs
can spend in consultancy services to design digital transformation projects based on
cloud and AI technology. The objective is reaching 2% of the small (10 to 50 employees)
and midsize (50 to 250 employees) SMEs over the 10-year period. Every year, SMEs
that have designed the most innovative cloud and AI-based transformation project plans
will receive additional support to fund their implementation.
What are the indirect impacts of this initiative on SMEs?
184 The direct benefits for SMEs can also be cost savings.
185
Most of the policy measures contribute to the development of European datacentre
infrastructure and cloud computing capacity, a richer European cloud and AI computing
service offering and a more dynamic technological and industrial ecosystem,
contributing to a higher adoption of cloud. This will improve the efficiency in software
and IT projects, increasing productivity of IT staff, ensuring the continuity of business
processes (less downtime) and facilitating migration processes, for SMEs and large
organizations alike. This impacts not only the efficiency of the IT staff but also the
flexibility and capability of the European private sector to adapt in an agile way to
evolving market needs and conditions, and to the dynamic technology landscape,
increasing its competitiveness. This agility is particularly important for SMEs, who need
to continuously adapt their technologies and business models in order to not lose market
revenues and remain relevant.
This could have been measured in terms of effort saved in IT tasks (FTEs) but it would
have been difficult to allocate the specific contribution for each specific measure, adding
potential risk of double counting their effect.
SMEs outside the cloud sector also gain from a more trusted and interoperable EU
cloud/AI ecosystem through safer access to digital tools, easier cross-border operations,
enhanced innovation, and participation in data-driven markets. While quantification in
some cases appears difficult, the qualitative impact of the initiative could be summarised
as follows:
1. Access to Advanced Digital Services
Enhanced trust and interoperability in the EU cloud and AI market gives SMEs in other
sectors access to advanced digital services. Businesses can safely and efficiently adopt
cloud-based tools for e-commerce, logistics, HR, or data analytics, even if they are not
part of the cloud industry itself.
2. Cross-Border Expansion
Standardised regulations and secure, cross-border data flows make it easier for SMEs to
expand across Member States. Companies can scale products and services without
facing legal or technical barriers, simplifying international operations within the Single
Market.
3. Innovation through Collaboration
SMEs can integrate AI, IoT, or data-driven services into their offerings, driving
modernization in traditional industries. Trusted cloud infrastructure lowers risks and
investment barriers, enabling experimentation and digital transformation.
4. Cost Efficiency and Resilience
Reliable EU cloud services improve cost efficiency and resilience185. Reduced
dependence on non-European providers decreases operational risks while competitive
185 The adoption of AI and cloud technologies by SMEs can significantly boost economic performance. Some university
studies indicate that AI adoption can increase revenue by up to 91% for SMEs, with operational costs potentially reduced
by up to 30% and time savings exceeding 20 hours per month arXiv.
186
pricing and high availability could benefit sectors like manufacturing, healthcare, and
fintech186.
5. Participation in Sectoral Data Spaces
SMEs can participate in sectoral data spaces that provide shared datasets for areas like
e.g. health, transport, energy, and finance187. Access to these resources opens
opportunities for new products, services, and partnerships, fostering innovation and
growth across the Single Market.
(4) MINIMISING NEGATIVE IMPACTS ON SMES
Are SMEs disproportionately affected compared to large companies? No
If yes, are there any specific subgroups of SMEs more exposed than others? N/A
Have mitigating measures been included in the preferred option/proposal? Yes,
financial support for SMEs to adopt cloud and AI (PM23)
CONTRIBUTION TO THE 35% BURDEN REDUCTION TARGET FOR SMES
Are there any administrative cost savings relevant for the 35% burden reduction
target for SMEs? SMEs are expected to benefit directly from PM23.
186 Trusted cloud and AI adoption empowers SMEs across sectors: in manufacturing, AI improves operations and
competitiveness (ScienceDirect); in healthcare, cloud solutions enhance patient experiences and administrative
efficiency (ACE Cloud Hosting); and in fintech, AI tools support revenue forecasting, risk modelling, and financial
decision-making (ResearchGate). 187 In healthcare, AI enhances diagnostics and patient care, improving efficiency and outcomes; in transport, AI optimizes
logistics and route planning, reducing costs and emissions; in energy, AI facilitates smart grid management and energy
efficiency, supporting sustainability goals; and in finance, AI-driven tools assist in risk assessment and financial
planning, enhancing decision-making capabilities. These advancements are supported by EU initiatives such as the
Digital Europe Programme and Horizon Europe, which aim to bolster AI and digital infrastructure across the Union.
ANNEX 7: EXTERNAL COHERENCE WITH RELEVANT EU LEGISLATION AND POLICY INITIATIVES
For each Policy Option, the table below analyses its external coherence.
PO External coherence
P O
1 -A
PM1 complements the existing Alliance for Industrial Data, Edge and Cloud. PM2 creates a permanent forum that currently does not exist. PM3 complements
existing best practices that focus on operating energy-efficient data centres with guidelines on deployment.
P O
1 -B
PM4 complements the upcoming Regulation on accelerating and streamlining environmental assessments, which focuses on administrative simplification
for environmental assessments, by adding a facilitator to accompany all permitting stages (incl. beyond environmental assessments) for a given data centre
project. PM5 complements envisaged provisions for industrial acceleration, notably in the upcoming Industrial Accelerator Act, where areas and measures
for industrial acceleration focus on manufacturing industries and do not account for the specific needs (e.g. connectivity) and benefits (e.g. their ability to
provide flexibility services to the grid) of data centres, which are a services industry and thus not in scope of the Industrial Accelerator Act. PM6 steers
national funding towards the deployment of innovative and sustainable data centres, while PM7 complements and ultimately leverages the existing Digital
Decade monitoring cycle. The scale of the compute capacity gap and its impact on EU competitiveness require such dedicated support measures for data
centre deployment, which cannot be achieved by horizontal initiatives that omit the specific needs of data centres.
P O
1 -C
PM8 complements existing framework programmes, which do not sufficiently cover data centre innovation and deployment. PM9 creates the possibility,
without prejudice to the outcome of the negotiations on the next MFF proposal, for strategic data centre deployment projects to be supported by Union
programmes, funds and financial instruments, in accordance with the objectives set out in the regulations establishing those funds and programmes. PM10
is an EU-level version of PM5.
P O
1 -A
,
B , C
PO1 B and C would leverage the future rating scheme for data centres under the Energy Efficiency Directive as a way of identifying particularly sustainable
data centres. Moreover, they would build on other initiatives which address data centre input factors, such as the future Grids Package (energy) or the future
Digital Networks Act (connectivity).
188
PO External coherence
P O
2 -A
PM11 would provide the first-ever definition of a sovereign cloud and AI computing service, going beyond technical cybersecurity and thus complementing
the Cybersecurity Act (CSA). PM12 adds guidelines for providers and PM13 is a new way of enhancing visibility of sovereign offerings. PM14 is designed
to complement the Data Act’s provisions on interoperability standards by enacting flanking measures that ensure the usability of the tools created by the
Data Act, notably by advancing standardisation work.
P O
2 -B
PM15 complements the cloud cybersecurity certification under the CSA and uses it as a way of assessing whether a service is sufficiently cybersecure.
However, the focus of the sovereignty scheme is fundamentally different as it covers sovereignty in a way that is not addressed by any other initiative,
including the AI Act, which is a product safety legislation. PM16 complements the horizontal procurement acquis with the necessary sectoral approach that
accounts for the nuances of purchasing cloud services, i.e. services with a global value chain, and empowers public purchasers to reward specific service
characteristics which the horizontal acquis cannot provide for. By focusing on cloud and AI computing services, it is clearly delineated from the AI Act,
which sets safety requirements for AI systems and general-purpose AI models. PM17 creates the tools for joint procurement, so far non-existent in this
sector. The same is true for PM18 which is an entirely new addition to the EU’s policy toolbox.
P O
2 -C
PM20 is a new policy tool, with no legal precedent. PM21 makes the use of the sovereignty risk assessment under PM15 mandatory, leveraging the NIS2
Directive to identify sectors of high criticality. Beyond giving a framework for cybersecurity certification, the CSA deals with supply chain risks with a
specific focus on high-risk vendors. However, it does not address public procurement. PM21 complements the CSA in that respect. This is enhanced by
PM19, which mandates the use of specific award criteria for all public procurements of cloud and AI computing services, thus complementing the
Procurement Directives. PM22 leverages the tools of PM17 and establishes the first-ever EU framework for joint procurement in this sector. PM23 leverages
and complements the European Competitiveness Fund, while PM24 presents an entirely new policy tool.
P O
2 -
A ,B
,C
All PMs under PO2 would build on the Data Act’s provisions on the right to switch and use different cloud services in parallel, which is a prerequisite for
users to diversify or change the cloud services they use.
189
Several existing and planned pieces of legislation and policy initiatives relate to the specific objectives pursued under this initiative. However, each of
them shows gaps with respect to reaching these objectives, which this initiative would fill. The table below illustrates this in detail and elaborates on the
relationship of these other pieces of legislation or policy initiatives with the future Cloud and AI Development Act.
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
SO1 -
Increase
computing
capacity
deployed in
the EU
through
innovative
and
sustainable
technologies
The Horizon Europe programme provides funding for
research and innovation in the EU, including in digital
technologies. Cluster 4 (Digital, Industry and Space)
covers enabling technologies like cloud as well as
advanced computing.
Horizon Europe lacks scale to meaningfully advance innovation in computing
capacity. Between 2021 and 2027, the total budget of Horizon Europe
amounted to EUR 95.5 bn. Estimates indicate that in 2023 only, the part of
AWS’s R&D spending on cloud computing alone was ca. EUR 18 bn188. This
initiative aims to ensure that innovation in computing capacity becomes a
stand-alone objective.
EU Framework Programmes also suffer from a gap between R&D and
commercialisation. For example, the SUNFISH project189 devoted 4.5 m of
Horizon 2020 budget to developing a solution for secure information sharing
in federated heterogenous private clouds. However, no European cloud
provider was able to capitalise on the project outcomes, which were not widely
taken up. This initiative aims to tackle not just the development but also the
uptake of innovative technologies through dedicated deployment measures and
incentives for the deployment of innovative data centres.
The IPCEI for Next Generation Cloud Infrastructure
and Services involves 19 companies from seven
The focus of this IPCEI is on research, development and innovation. Beyond
that, it only covers the first industrial deployment of innovative solutions
188 How much does Amazon invest in R&D? Here's an informed guess. 189 SecUre iNFormation SHaring in federated heterogeneous private clouds | SUNFISH | Projekt | Fact Sheet | H2020 | CORDIS | European Commission.
190
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
Member States: France, Germany, Hungary, Italy, the
Netherlands, Poland, and Spain. These seven Member
States provide up to EUR 1.2 bn in public funding to
developing data processing capabilities, and software
and data sharing tools that enable federated, energy-
efficient and trustworthy cloud and edge distributed data
processing technologies and related services.
(particularly in edge computing). Most notably, the IPCEI focuses on beyond-
the-state-of-the-art solutions. Consequently, the IPCEI does not target capacity
deployment in a way that would contribute to closing the current compute
capacity gap. Moreover, compared with innovation efforts by leading
international cloud providers, the funds available to European companies under
this IPCEI are much lower. However, the research and development efforts
supported under this initiative will build on and integrate achievements under
the IPCEI. Moreover, deployment incentives will positively affect the broader
deployment of solutions developed under the IPCEI.
The AI Factories and future AI Gigafactories
already/will provide high performance computing
capacity to developers and deployers of AI models. The
initiatives leverage the EuroHPC Joint Undertaking.
These initiatives mainly serve to drive advancements in AI applications across
different sectors, going beyond the capabilities of general-purpose and AI-
optimised data centres. They focus on large concentrations of computing power
needed for the training of AI models. They do not address the deployment of
more decentralised computing capacity in data centres, necessary for the large-
scale uptake of AI throughout the EU and for enabling lower latency. The
deployment of AI Factories and Gigafactories may benefit from measures
under this initiative, such as identified areas for accelerated deployment.
The Energy Efficiency Directive contains measures on
improving the energy performance of data centres. It
encourages waste heat re-use and requires data centres
above 500 kW to report annually on their energy usage
and on other sustainability metrics. This reporting will
form the basis for a future rating scheme, which will
be enacted as part of the upcoming Data Centre Package.
The Energy Efficiency Directive is the key piece of legislation for guiding the
data centre industry towards greater energy efficiency through measures such
as reporting and the future rating scheme. However, beyond transparency
measures and reputational considerations, the Directive does not set incentives
for data centre operators to improve their sustainability performance and does
not contain measures for accelerating the roll-out of sustainable data centres
across the EU or increasing related investment (although the possible
191
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
This will also kick off the work on a separate proposal
for data centre minimum performance standards,
building on the sustainability metrics established under
the Energy Efficiency Directive.
establishment of minimum performance standards, stemming from the
Directive, could set the minimum specifications of sustainability for future data
centres in Europe). The initiative will contain measures to incentivise the roll-
out of energy-efficient data centres. To identify which data centres are
sustainable, this initiative will refer to the rating scheme developed under the
Energy Efficiency Directive.
The Regulation laying down ecodesign requirements
for servers and data storage products sets
requirements for sustainable products.
Data centre hardware is subject to minimum performance and information
rules; however, the Regulation does not contain measures to directly
incentivise the deployment of innovative and sustainable technologies.
The Digital Decade Policy Programme (DDPP)
sets the target of 75% of EU businesses adopting cloud
services and of 10.000 edge nodes being rolled out by
2030 and measures progress in all Member States and at
EU level.
Beyond the edge node target, the DDPP does not contain targets for the
deployment of compute capacity or data centres in the EU. And beyond setting
out targets and monitoring progress (2257 edge nodes in 2024), it does not
contain concrete support measures for such deployment. This initiative will
complement the DDPP in this regard: The data centre capacity target would be
integrated into the DDPP, and by leveraging the existing yearly monitoring
exercise, this initiative would create synergies with the DDPP.
The EU Code of Conduct on Data Centre Energy
Efficiency is a reference document to assist data centre
operators in identifying and implementing
measures to improve the energy efficiency of their data
centres. This Best Practice document contains a full list
The Code of Conduct does not concern measures to deploy a data centre,
instead focusing fully on practices for operating it.
192
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
of the identified and recognised data centre energy
efficiency best practices within the Code of Conduct.
SO2 -
Ensure
attractive
conditions
for the
deployment
of
sustainable
and
innovative
computing
capacity
European Grids Package, put forward by the
Commission [on 10 December 2025] aims to modernise
and enhance the EU’s electricity grids. It
contains targeted legislative changes to the TEN-E
Regulation and related regulations and directives, with a
view to improving pan-European grid planning. It will
consider permitting for grid expansion and cost-sharing
in a cross-border context, as well as providing guidance
on measures to accelerate grid connections within the
current legislative framework.
As discussed throughout this assessment, energy is a key input factor for data
centres and sufficient grid capacity is a prerequisite for their deployment in the
EU. The European Grids Package aims to ensure grids will be in place and
ready to uptake future loads in a horizontal manner, as the issue at stake (lack
of grids capacity) concerns more stakeholders. The initiative assessed here
focuses on data centres as an ultimate client of grid capacity in the EU. It aims
to complement the Grids package by ensuring data centres location considers
grid availability, information is exchanged sufficiently in advance to feed into
grid planning and hence ensure timely connection of data centres.
The guidelines on grid interconnection lay down ways in which Member States
can accelerate grid connections, including a more efficient structure of their
grid connection queue, for example by considering a project’s readiness and
grid-friendly uses as opposed to a pure first-come-first-served approach. It also
requires strong involvement of industry, including data centre sector in the
grids planning. This initiative will leverage these considerations, for example
by clarifying that public support for strategic projects shall only concern
projects which are able to provide needed flexibility for the grid, as required
by relevant technical conditions, for the sake of their efficient grid integration.
193
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
The future Digital Networks Act (DNA) seeks to
incentivise the building of the digital networks of the
future and improve digital connectivity for end users.
The DNA will improve digital connectivity for end users and thus be beneficial
to the deployment of data centres in the EU, for which high-performance
connectivity is a prerequisite. This initiative will leverage the DNA’s
advancements for connectivity and will thus stay focused on the deployment of
the data centre (where data processing occurs), not the prior or parallel build-
out of the necessary connectivity infrastructure. The two initiatives are thus
complementary.
The DNA will also address the convergence of network infrastructure. For
example, it will address scenarios where a cloud service provider operates a
network and has so far not been subject to obligations under the Electronic
Communications Code. The intention is to clarify the modalities of
interconnection between operators of networks and other market participants.
This relates to a specific market situation and is not impacted by possible
measures under this initiative.
The Gigabit Infrastructure Act sets rules to accelerate
the rollout of Gigabit networks installations.
While data centre operators and cloud service providers will benefit from an
improved rollout of very high-capacity networks, the Gigabit Infrastructure Act
does not cover the constructions of data centres. This initiative would thus
build on the advancements in accelerated network deployment under the
Gigabit Infrastructure Act.
The upcoming Industrial Accelerator Act will set rules
for foreign investment contributions and create clusters
of industrial activity for the manufacturing sector
On permitting, the IAA sets out the principle of ‘one project, one procedure’
and establishes a single digital procedure to cover the entire permit-granting
process. It also puts forward other measures to accelerate permitting (time
limits for project approvals, single point of contact, overriding public interest
194
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
etc.). This initiative will leverage the newly created approach to simplified and
digitised permitting by cross-referencing the IAA.
The IAA will require Member States to designate at least one manufacturing
industrial acceleration area/cluster where further business facilitation measures
apply (prioritised access to materials, access to EU finance, regulatory
sandboxes etc.). However, these areas won’t account for the specific needs of
data centres, which are a services sector.While data centres consume a lot of
energy, a characteristic they share with other more traditional industries, they
present several fundamental differences. First, unlike industries that may adjust
energy use based on production schedules, the energy demand of data centres
are more constant, as their primary purpose is to ensure continuous service and
data availability. Moreover, data centres can assist the grid through flexibility
services. Second, their quality-of-service obligation imply that data centres are
equipped with powerful energy generation capacities that can be leveraged by
grid operators for flexibility purposes. Third, unlike other energy-intensive
industries, the physical location of a data centre is dictated by the presence of
robust connectivity: low latency is critical for supporting the real-time data
processing and application needs of other industries, such as financial
transactions or AI applications. That is why the cloud and AI Development Act
will complement general industrial acceleration measures with measures that
are tailored to the accelerated deployment of data centres.
195
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
The future Regulation on accelerating and
streamlining environmental assessments will simplify
and speed up environmental screenings and assessments.
Data centres are not subject to mandatory environmental impact assessments
under the existing Environmental Impact Assessment Directive. However,
Member States can determine themselves that an individual data centre project
should undergo such an assessment, causing them to be de facto mandatory in
many Member States. The new Regulation will establish a toolbox with
additional provisions (overriding public interest, tacit approval and dispute
settlement) that can be leveraged in sectoral legislation regarding strategic
sectors or categories. For all environmental assessments, the new Regulation
establishes streamlined provisions, including maximum timelines for the
duration of the screening process and for decisions by competent authorities.
Procedurally, the proposal sets up a single point of contact for Environmental
Impact Assessment and digitises and streamlines assessment procedures (e.g.
through the re-use of documents across different stages). The future Regulation
on accelerating and streamlining environmental assessments focuses on
environmental assessments, not on other steps of the permitting process such
as zoning and land allocation and building permits.
The cloud and AI Development Act will leverage these simplifications and
complement them, notably by establishing a facilitator to accompany data
centre operators throughout their overall permitting journey, not only the
possible environmental assessment. The cloud and AI Development Act will
also leverage the toolbox established by the new Regulation, by referencing it
within the Act to activate the additional provisions for strategic data centre
projects.
196
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
The Net-Zero Industry Act (NZIA) sets out to boost
the EU’s manufacturing capacity for net-zero
technologies, e.g. through simplified permitting and
areas for faster deployment.
Data centres do not qualify as net zero technologies. The acceleration measures
under this initiative are thus not in conflict with NZIA. Data centre operators
may however benefit from NZIA as downstream customers or users of net zero
technologies such as batteries.
The Renewable Energy Directive (RED III) simplifies
and accelerates permitting procedures for renewable
energy and storage projects, including
by establishing renewables acceleration areas.
Data centres may benefit from the increased availability of renewable energy
and storage in the EU but are not themselves covered by the Directive. This
initiative is thus fully complementary to RED III and can leverage it: Proximity
to renewables acceleration areas may be a relevant factor in designating sites
for faster data centre deployment.
SO3 -
Decrease the
overall
reliance on
non-
European
cloud and AI
computing
services
The Data Act,which enters into application in
September 2025, enables cloud switching by removing
key sources of contractual, commercial, and technical
vendor lock-in. As part of this, it sets rules for fair cloud
contracts, creates an EU repository for harmonized
standards and open interoperability specifications, and
lays the basis for standardization requests.
By enabling switching and removing key sources of vendor lock-in, the Data
Act seeks to ensure that cloud service providers in the EU compete on quality,
innovation, and price. It seeks to enable cloud users to freely choose the
provider that best meets their needs and combine offers of different providers
in a multi-cloud approach. However, the Data Act does not contain elements to
shape up a more competitive offer of EU cloud services or encourage the entry
into the market of a more diverse set of cloud service providers. The Data Act
opens the path towards a possible reduction of dependencies on non-EU
providers but does not build the road towards a more sovereign and trusted EU
cloud computing sector. The cloud switching and interoperability provisions,
however, make it possible for users to embrace European cloud computing
services more strongly. The Data Act is thus an enabler for the cloud and AI
Development Act.
197
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
The Digital Markets Act covers cloud services as a
core platform service, meaning that cloud service
providers designated as gatekeepers would have
to comply with a set of obligations, for example related
to enabling interoperability.
So far, no cloud and AI computing service provider has been designated as a
gatekeeper for their services (Amazon designated for the marketplace and
amazon advertising; Microsoft is designated for LinkedIn and for Windows PC
Operating System). On 18 November 2025, the Commission opened three
market investigations on cloud computing services under the DMA: Two
market investigations will assess whether Amazon and Microsoft should be
designated as gatekeepers for their cloud computing services, Amazon Web
Services and Microsoft Azure, under the DMA, in other words whether they
act as important gateways between businesses and consumers, despite not
meeting the DMA gatekeeper thresholds for size, user number and market
position. The third market investigation will assess if the DMA can effectively
tackle practices that may limit competitiveness and fairness in the cloud
computing sector in the EU.
Irrespective of the gatekeeper designation, the DMA does not contain measures
that would actively promote the uptake of sovereign cloud services. The DMA
aims at remedying specific market behaviours of certain providers that are
gatekeepers and thus intervenes at a different level than the cloud and AI
Development Act, which focuses on the uptake and use of the services
provided.
SO4 -
Contribute
to the
protection of
fundamental
The Public Procurement Directives cover tenders
above certain thresholds and prescribe transparency,
equal treatment, open competition and sound
management for procurement procedures. The
upcoming revision of these Directives intends to align
Public authorities in the EU rely heavily on non-EU cloud and AI computing
service providers. The problem drivers justifying the policy measures of the
cloud and AI Development Act regarding the public procurement of sovereign
cloud and AI computing services in sectors of high criticality are very specific
to the technologies and applicable legal frameworks concerned. For example,
198
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
interests of
society by
enhancing
the
resilience of
supply of
cloud and AI
computing
services, in
particular in
the public
sector
them more closely with strategic goals of the EU,
including resilience considerations and the announced
EU preference in public procurement within strategic
sectors.
a cloud computing service value chain and the underlying infrastructure used
to deploy the service can be distributed across different geographies and subject
to corresponding jurisdictions, including those with laws with an
extraterritorial effect. Such problem drivers therefore require a nuanced and
targeted sectoral approach, which is not covered by the existing Public
Procurement Directives and would be difficult to account for sufficiently
through an overarching EU preference approach. This initiative will therefore
provide a sector-specific approach to sovereignty – the many layers of which
cannot be addressed in the horizontal acquis that lays down general principles
for the design of procurement procedures – which will be complemented by
award criteria that are also tailored to the specificities of cloud and AI
computing services.
The Cybersecurity Act, currently under revision, lays
down a voluntary certification framework for ICT
processes, services and products – with cloud services
considered as a subset of ICT services. It gives the
legislative framework for adopting a European
cybersecurity certification scheme for cloud services by
means of an Implementing Regulation.
ENISA has been working on developing an EU-wide cybersecurity
certification scheme for cloud services (EUCS), which has not yet been
adopted. Certification under the CSA can address technical
cybersecurity criteria but is not suited for addressing sovereignty concerns that
go beyond these technical elements. This initiative will thus complement the
CSA’s cloud cybersecurity focus with sovereignty considerations. A finalised
EUCS could be leveraged in the audit scheme for sovereign cloud and AI
computing services as a way of ensuring that a service subject to a sovereignty
audit meets the highest cybersecurity standards.
The Directive on Security of Network and
Information Systems (NIS2) sets out cybersecurity risk
management and incident reporting obligations to a set
The NIS2 Directive improves the cybersecurity risk management of cloud
service providers and data centres in the EU, resulting in greater trust.
However, it does not contain measures geared at the uptake and use of such
199
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
of essential and important entities across critical sectors,
including cloud services and infrastructure.
services and remains fully focused on technical cybersecurity as opposed to
broader sovereignty considerations. This initiative will therefore complement
the NIS2 Directive.
The Digital Operational Resilience Act (DORA) aims
to strengthen the digital resilience of financial entities.
It ensures that banks, insurance companies, and
investment firms can withstand, respond to, and recover
from ICT disruptions, such as cyberattacks or system
failures.
While DORA shapes compliance obligations for cloud service providers and
indirectly covers AI computing service providers if they provide services to
financial entities covered by DORA or if their role is significant enough in
terms of operational resilience, it has a sectoral scope and is financial-sector
specific. Under DORA, cloud service providers must implement ICT risk
management, conduct regular testing and incident response to comply with the
requirements for critical third-party service providers. The covered financial
institutions, which could be public in nature, must carry out due diligence on
the cloud providers they work with.
The Artificial Intelligence Act (AI Act) sets out rules
for AI systems and general-purpose AI models and not
cloud and AI computing services. It follows a risk-based
approach and provides rules to ensure that AI systems
used in the EU are safe, for example by introducing
safety and security requirements, such as accuracy,
robustness and cybersecurity for data and data
governance.
The AI Act harmonises rules for AI systems and general-purpose AI models to
be placed on the EU market, improving the functioning of the internal market
and promoting the uptake of human-centric and trustworthy AI along the value
chain. As a product safety legislation, the AI Act ensures a high level of
protection of health, safety and fundamental rights, It does not cover aspects of
sovereignty regarding AI computing services, which the cloud and AI
Development Act will address. The two initiatives are thus complementary.
The EU-US Privacy Framework (DPF) is a
mechanism established to facilitate the transfer of
personal data from the EU to the US, while ensuring
While the DPF addresses transatlantic data transfers, it does not remove
sovereignty concerns about dependence on US hyperscalers and on the
subsequent exposure to surveillance laws, such as the US CLOUD Act,
200
Specific
objective
Legislation or policy initiative Gap with regards to specific objective / relationship with the cloud and AI
Development Act (“this initiative”)
adequate levels of data protection. It replaces various
legal frameworks that were invalidated by the
EU courts for insufficient protection of personal
data (e.g., the Privacy Shield). On 10 July 2023, the
Commission issued an adequacy decision recognizing
that US organizations certified under the framework
offer an adequate level of protection so that data
transfers could take place without additional legal
barriers. On 3 September 2025, the General Court of the
EU dismissed an annulment action, confirming that the
framework is valid.
obliging US providers to hand over data to US authorities even if stored in the
EEA. Moreover, the notion of sovereignty goes beyond data transfers and also
relates to operational autonomy, an aspect that goes beyond the DPF. This
initiative thus complements the DPF.
ANNEX 8: KEY TECHNICAL CONCEPTS
A data centre provides the walls, cooling, and connectivity needed to run the IT equipment
that enables computing, which are typically processors mounted on servers, storage units
sometimes called memory, and networking equipment. Data centre capacity is measured
in megawatts (MW), indicating the total electrical power made available to the IT
equipment190. The capacity of the equipment is therefore limited by the electrical power
available in the data centre. Throughout this report, the term ‘computing capacity’ is thus
used interchangeably with data centre capacity.
The International Standardization Organization (ISO) defines cloud computing as a
“paradigm for enabling network access to a scalable and elastic pool of shareable physical
or virtual resources with self-service provisioning and administration on demand”. These
resources are located in data centres. Operating a data centre and providing a cloud service
are two distinct segments of a same supply chain. Data centres can be owned by cloud
service providers (CSPs) themselves (vertical integration). Otherwise, data centres
operators are often specialised independent businesses, which rent out space for customers,
including CSPs, to place their equipment in. The data centre can have a single client or
multiple, which are referred to as co-location data centre providers. Companies which
provide cloud services on a very large scale are commonly referred to as hyperscalers, a
category understood to include Microsoft, Amazon and Google. For very fast response
times, i.e. low latency, data must be stored and processed closer to the user, in edge data
centres or edge nodes, which can be seen as mini-data centres (an edge facility can be as
small as the size of a fridge, while the largest data centres can be as big as several football
fields).
While data centres powering traditional cloud services and AI computing services look
similar, they present important differences. First, at the microprocessor level: while data
centres specialised in AI computing services use a mix of Graphic-Processing-Units
(GPUs) and Central-Processing-Units (CPUs), others are exclusively powered by CPUs.
Second, GPU-powered data centres, sometimes called ‘AI data centres’, are much more
energy-intensive and require a different physical layout implying a design choice from the
outset, making refits difficult. Throughout this assessment, the term ‘data centre’ is used
to refer to both types of data centres in order to be future-proof and end-use agnostic.
190 This is a widespread industry practice. However, it is imperfect as it does not reveal the type, quantity and capacity
of the chips hosted in the data centre and does consider the efficiency of this equipment.
202
Figure 1. Key concepts along the cloud and AI value chain
When training an AI model, longer response times/higher latency are not as problematic
while a large concentration of computing power is paramount. The size and complexity of
an AI model is often described by the number of parameters it is built on, a figure that can
reach a trillion. Training a model, even on the worlds’ largest computing facilities, can take
several months of uninterrupted calculation. That is why training is usually executed in
High-Performance Computing (HPC). HPC is provided over specialised data centres and
differ from cloud computing in the sense that, typically, the pool of resources is not shared
and/or not scalable and/or not elastic. Conversely, the fine-tuning, inference and service
integration of AI models rely on cloud or even edge computing. Fine-tuning an AI model
consists in customizing pre-trained models using domain-specific data. Inference then uses
this fine-tuned model to generate predictions or outputs on new, unseen data by applying
the learned parameters without further training. Edge computing makes it possible to run
AI applications close to the data source, allowing for very low latency, for example in AI-
assisted real-time imaging and surgery, in autonomous navigation and real-time decision-
making in autonomous driving. The different stages of an AI model, reaching from training
to application, including with low response times, require a continuum of computing
capacity spanning HPC, data centres, and edge nodes.
Cloud services are often broken down in three layers: infrastructure-as-a-service (IaaS)
consists in supplying basic computing services, essentially storage and raw processing;
platform-as-a-service (PaaS) consists in offering a platform over which services can
deployed; and software-as-a-service (SaaS) is a software application delivered over the
cloud such as an office automation suite or a video streaming service. Cloud service
providers can offer services that span across all layers (e.g. Microsoft offers Office 365
running over its own Azure infrastructure), or not (e.g. Netflix operates over Amazon’s
infrastructure).Cloud computing and AI are deeply interconnected, each fuelling the
other’s potential. Cloud computing provide the scalable infrastructure and computational
power AI needs to process data, train models, and deploy applications efficiently. In turn,
AI enhances cloud services through automation, predictive analytics, and smarter resource
management, making them more adaptive and user-friendly. AI computing services
provide computing infrastructure that allows users to run trained AI model (inference)
203
without hosting the AI system on their own hardware. Moreover, AI functionalities are
increasingly integrated into cloud services, for example into SaaS offerings, to enhance
their functionalities.
This assessment systematically uses the term ‘cloud and AI computing services’
understood as offering computing resources for the running (inference) of AI systems. AI
systems, pursuant to the AI Act, are not covered by the sovereignty scheme described in
this Impact Assessment.
Cloud services may also include other data processing-related services, e.g. data
analytics191.
Figure 6. Interplay between cloud and AI
191 The Digital Decade Policy Programme, for example, measures uptake for AI, cloud and big data together: 'Path to the
Digital Decade': the EU's plan to achieve a digital Europe by 2030 - Consilium.
204
ANNEX 9. BASELINE SCENARIO (POLICY OPTION 0)
The baseline scenario represents the reference case against which policy options are
assessed. It provides a realistic and neutral projection of how the EU cloud and AI
computing services market, and the related data centre infrastructure, are expected to
evolve in the absence of additional policy intervention beyond current measures and
commitments. The scenario integrates the effects of:
• Existing EU legislation, especially for infrastructure development (EED and RED III),
that is Problem 1. In the case of cloud services (Problem 2), the main applicable
legislation is the Data Act, which entered into force in September 2025, whose effects
are not yet observable. Similarly, the impact of EU public funding cannot yet be fully
assessed, as the current Multiannual Financial Framework (MFF) is still subject to
evaluation. In addition, no dynamic outlook on future funding allocations can be
provided, given that negotiation on the next MFF have not yet started;
• Market-driven technological progress and private investment trends;
• Macroeconomic forecasts;
• Global developments affecting demand for computing capacity, cloud services, and AI
workloads.
It covers the period 2025-2036 and the EU-27, compares to global market evolution and
provides a projection of likely developments in the absence of further policy action, thus
allowing for the quantitative and qualitative comparison of the policy options’ expected
impacts. It describes how each identified problem and their underlying drivers would
develop to 2036 without new policy action, drawing on available quantitative evidence and
market projections.
1. Limited and geographically concentrated availability of computing capacity
For problem 1, The EU’s total installed compute capacity, measured in data-centre IT load,
stands at approximately 12.4 GW (and 13.9 GW if public sector capacity is also
considered), representing around 20%of global capacity. Growth is projected to reach 42.5
GW by 2036, yet in the central scenario demand data centre capacity is expected to increase
to 61.5 GW over the same period, creating a structural capacity gap of 19 GW relative to
projected needs, as shown in the figure below.
Figure 7. Forecasted data centre capacity and demand in EU-27 (Source: Technopolis et al. (2025))
205
In this central scenario data centre demand growth is based on the following assumptions:
• Enterprise IT budgets across the EU-27 continue to expand at a steady and predictable
pace, reflecting a stable macroeconomic environment and a measured approach to
digital transformation.
• IT spending as a share of enterprise turnover increases gradually, reaching 3.7% by
2032. From 2032 onwards, this ratio stabilises, indicating a maturing digital spending
profile with spending only increasing to 3.8% by 2036. Within this broader trend,
investment in data centre infrastructure grows in line with the historical CAGR of
14% through to 2030, reflecting sustained demand for compute and storage
infrastructure. This growth supports both the retrofitting of legacy environments and
the deployment of new, AI-capable facilities, particularly in markets where cloud
adoption and enterprise workloads continue to scale. From 2032 onwards, the growth
rate of data centre spending decelerates, but outlays continue to increase, sustained
by lifecycle refreshes, AI-capable retrofits, compliance and resilience upgrades, and
steady colocation demand
• Workloads evolve steadily, with increased deployment of AI, machine learning, and
data-driven services contributing to higher computational requirements.
• Digital policies boost private sector demand. As governments roll out e-government
services, data-sharing initiatives, and sector-specific AI guidelines, private enterprises
respond by adopting initiatives which steadily drive increases in compute and storage
requirements.
Data centre supply growth is based on the following assumptions:
• Capacity growth is around 14% CAGR through 2025–2030, then eases over 2030–
2036.
• Expansion remains concentrated in FLAP-D hubs, with steady investment in secondary
markets (Nordics, Spain, Italy) and slower uptake in developing markets (around
6.49% CAGR). Capacity expansion follows market trends concentrated in existing
hubs, where grid constraints, land availability, and permitting delays limit new
deployment. Investment remains market-led; new entrants face high upfront capital
costs (around EUR 8-10 million/MW).
• Aggregate growth is around 17.7% over 2025–2030 (consistent with the survey’s
~14% CAGR lens) as committed pipelines, AI-readiness upgrades, and new cloud
regions come through. From 2031–2036, growth moderates to around 7.1%, reflecting
tighter site selection driven by power and land availability, longer permitting lead
times, and higher financing costs. This trend risks constraining AI model training and
cloud workloads within the EU, particularly for SMEs and public-sector users.
• Electrical grid limitations remain a key issue in certain primary markets such as Ireland
and the Netherlands. However, regulators and operators in several markets have begun
investing in grid modernisation and demand management. Without targeted policy
intervention long lead times for grid connection remain across Europe (>30 months).
• The Energy Efficiency Directive requires operators of large data centres to disclose
annual energy and water use, improving transparency and benchmarking. Efficiency
improvements are moderate, but do not offset the growth in total electricity demand,
which is expected to rise from 105 TWh to over 200TWh in 2036. Energy consumption
by data centres could reach around 340 TWh annually in 2036, growing at a compound
annual growth rate of 11.4%. If electricity production remained stable at the levels
206
reported by the IEA in 2022 then data centres would consume 12% of EU27’s
generated electricity.
• Energy efficiency in data centres improves gradually; the average Power Usage
Effectiveness (PUE) declines from 1.29 in 2025 to about 1.23 by 2036.The carbon
intensity of this demand declines as the electricity mix decarbonises under the Fit for
55 package. Incremental advances in cooling, waste-heat recovery, and on-site
renewable generation occur as a result of existing regulatory incentives under EED and
RED III.
• Surveys have identified permitting delays (up to 3-5 years in some jurisdictions), grid-
connection queues, and land-use restrictions as principal d availabili new capacity.
Without policy harmonisation, these bottlenecks persist, with moderate alleviation
from national administrative reforms. Therefore, data centre expansion continues but
below potential. Under continued existing conditions, roughly 25-30 % of announced
projects experience significant delay or downsizing
• Governments aim to balance digital infrastructure expansion with environmental
priorities by introducing power efficiency targets, and sustainability requirements.
National measures, such as tax incentives for green data centres in the Nordics or
Italy’s simplified permitting for renewable-powered facilities, continue at present scale
• Policy evolution remains gradual and inconsistent, creating a fragmented regulatory
landscape across Europe. Efforts to modernise infrastructure and streamline permitting
are underway, but progress is uneven, often slowed by regulatory complexity and
localised decision-making.
Under policy option 0, Europe’s compute supply increases in absolute terms but lags
behind global competitors. North America and Asia-Pacific expand faster, increasing their
share of global compute. The EU maintains its global market share of around 20-23%.
While the region remains attractive for investment due to regulatory stability and skilled
labour, its relative cost position remains less favourable than that of North America or
some Asian economies. Considering the sectors and region’s expected growth and
technological developments in AI, quantum computing and 6G, international providers are
expected to maintain a strong EU presence under existing trade and competition rules.
Goldman Sachs forecast that total global capacity will reach 122GW by 2030, growing at
a CAGR of 16%. The forecast produced by the study indicates that EU27 capacity will
grow at 18% over the same period, meaning EU27’s share of data centre capacity is
expected to grow from 20% (2025) to 23% (2030), as shown in the figure below.
Figure 8. EU-27 share of forecast compared to the global capacity (Source: Technopolis et al. (2025))
207
Estimates for the United States suggest that by 2028 data centres will account for between
7-12% of total consumption.192 The IEA in turn estimate that data centre electricity
consumption in the United States will grow to 240 TWh annually by 2030 (from 180 TWh
in 2024). This would equate to 6% of the United States’ 2024 electricity consumption. The
IEA also estimated China’s data centre electricity consumption to grow to 175TWh (from
100 TWh in 2024). This would equate to 2% of China’s 2024 electricity consumption.193
2. Dependence on cloud and AI computing services supplied by non-European
providers
For problem 2, the baseline scenario describes how the market share of European providers
evolves without any EU intervention beyond initiatives and regulations already adopted,
such as the Data Act.
The core aim of the Data Act is to reduce friction when switching from cloud provider, but
its full effects will take time to materialise.
Since October 2025, a set of standard contractual clauses recommended by the
Commission will make contractually easier for cloud users to switch from provider, as they
are progressively adopted. Also, by January 2027, switching from one cloud provider to
another should be free of charge. Finally, aspects related to the technical switching will
take more time to fully materialise. A soon to-be-published Commission study has
identified only a handful of industry driven specifications that the Commission will
propose to become mandatory through the foreseen mechanisms of the Data Act. This will
require the adoption of several implementing acts following which industry will have a 12-
month transition period to ensure compatibility of relevant data processing services with
the specifications. Further standardisation will have to undergo a formal request from the
Commission to the European Standardisation Organisations, a process prone to take time.
The effects of the Data Act for facilitating the emergence of an integrated offering by
European providers are at this stage therefore limited and indirect. Finally, it must be noted
that switching is a customer-driven operation and the effects on European providers and
the growth of their customer base rely on the willingness of customers to make use of their
new right to switch
The baseline presented for the policy scenario 0 for problem 2 is dynamic and focuses
mainly on IaaS and PaaS using data from Synergy194 on the market share (15%) and
Statista for the market revenues (EUR 125.22 bn)195. Due to the heterogeneity of the SaaS
market and the unclear categorization of what is included in the SaaS category, these are
excluded from the baseline scenario. AI computing services can be considered as an
extension of cloud services for the specific purpose of running an AI model. The two
categories overlap significantly, and AI computing services are currently not separately
192 Berkley Lab (2024). 2024 United States Data Center Energy Usage Report. Available at:
https://escholarship.org/uc/item/32d6m0d1 193 IEA (2025). Energy and AI. Available at: https://www.iea.org/reports/energy-and-ai 194 Synergy Research. Data from July 2025. 195 Statista
208
captured in statistics. That is why the below considerations are based entirely on figures
and developments related to the above-mentioned market for cloud computing services.
The base year considered for the baseline scenario is the latest full year for which there is
stable data available, namely, 2024. In this base year, Synergy shows that EU-
headquartered providers account for around 15% of the EU cloud infrastructure market,
with the remainder provided by non-EU firms, in particular large US and other global
providers. The base-year distribution and levels are taken as factual anchors. The dynamic
baseline is built around the following assumptions:
• The European cloud market grows at CAGR 17% until 2030, as pointed out by
Synergy and from 2031 to 2036, it is assumed to grow at a slower CAGR of 12%.
• Within the EU market, the market share served by EU providers, follows a flat
scenario until 2036, similar to the almost flat trend that can be observed in the last
two years. This is the status quo scenario.
• Two sensitivity variants are also considered:
o A declining scenario, falling from the current market share of 15% to 10%.
This is the pessimistic scenario.
o A slightly optimistic scenario, rising from the current 15% to 17%.
The time horizon for the baseline spans until 2036, which is the time used for the modelling
of the different policy measures, and a reasonable period in which structural changes in the
dynamics of the cloud market can be expected.
In terms of market volume, the assumption is that, while the market grows, the EU
providers’ market share grows in parallel, but it slightly loses weight in the total market,
leading to a small decline in European providers ‘overall share of the cloud market
revenues.
Figure 9. Market share of European cloud providers in the European market (Source: Synergy
Research)
Within the evolving European cloud market, the central dynamic baseline assumes that
EU-headquartered cloud service providers are unable to fully keep pace with their non-EU
competitors. Due to a growing demand in the EU, the revenues of the European players
keep steady, but their overall position barely holds.
209
These projections are grounded in recent market evidence (see Figure 9) and structural
factors. Over the last years, European cloud providers’ local market share has declined
from roughly 29% in 2017 to about 15% and has since stagnated at that level, despite
continued strong growth of their absolute revenues, as it can be seen in the figure above.
At the same time, the three leading hyperscalers invest at a scale (around EUR 10 bn per
quarter in European CAPEX according to Synergy research196), which results in very high
barriers to significantly increase the market share for smaller European players. This
suggests that in the absence of major policy interventions or market shocks, the relative
position of EU providers is likely to remain broadly unchanged in the mid-term. The only
cloud – specific policy currently applicable in the EU, the Data Act, entered into force in
September 2025. As explained previously, its effects are not yet observable, rendering
difficult to provide a more dynamic outlook than the one shown here.
The 15% used as the baseline reflects steady but constrained growth driven by potential
gradual demand shifts generated by rising sovereignty expectations and niche market
needs, while still facing structural issues (e.g. lock-in to hyperscalers, bundling advantages
and integration needs). The 10% pessimistic scenario assumes that there is a widening AI
infrastructure investment gap between US and EU cloud and AI service providers
combined with regulatory initiatives failing to deliver meaningful constraints on lock-in,
dampening switching and interoperability, along with price pressure from bundled offers
and potential providers exits triggering a collapse in customer confidence towards cloud.
The more optimistic scenario, 17%, assumes improved interoperability and portability and
more trusted EU-level assurance, raising conversion rates and accelerating adoption.
All three scenarios converge within a narrow seven–percentage-point band because the
core competitive disadvantages faced by European cloud providers are deeply rooted. Even
in the most optimistic case, European providers only recover to about 17% market share.
This is largely due to Europe’s fragmented regulatory environment and the difficulty of
scaling cloud and AI services, that can match the breadth and integration of those offered
by the hyperscalers. Conversely, the 10% pessimistic scenario includes the demand of
cloud and AI services by highly regulated sectors that are genuinely sovereign. These needs
ensure a stable baseline level of demand of cloud services.
Other considerations taken into account for the assumptions include structural market
factors, demand-side dynamics and competition environment, as described in section 2,
notably 2.2.
Scenario Market share
of EU
providers in
2036
Cumulative revenues of
EU providers (2025 –
2036) (EUR bn)
Cumulative revenues
of non-EU providers
(2025 – 2036) (EUR
bn)
Baseline (pessimistic scenario) 10% 438 33.212
Baseline (flat-share scenario) 15% 564 31.956
Baseline (optimistic scenario) 17% 611 31.483
Considered together, the three presented trajectories present a dynamic baseline scenario
(Policy Option 0) towards which the EU cloud market is likely to continue evolving if the
196 https://www.mobileeurope.co.uk/european-cloud-providers-tread-water-in-growing-market/
210
EU will not adopt any additional policy measures. Under this scenario, the global and
European cloud markets would modestly grow but without a strong quantitative shift in
the structure. This would lead the EU to remain structurally dependent of non-EU players
covering most customers’ needs. Furthermore, EU-homegrown providers’ scale and
investment in capacity and creation of new services would remain limited, constraining
their ability to compete, especially with hyperscalers that present a more integrated and
wider offer. In the absence of an EU action, European providers won’t meaningfully
improve their positioning in the market and European cloud users will continue to rely
heavily on non-European providers.
Policy options under PO2 are measured both in quantitative and qualitative terms.
Quantitatively, policy options are evaluated by how much they shift the baseline trajectory
in terms of higher market share of European providers and reduced dependency, especially
important in highly critical use cases. Qualitatively, the baseline underlines that without an
EU intervention the EU market landscape would remain characterised by a high external
dependency, limited European scale and persistent asymmetries.
These projections serve as the counterfactual reference for quantifying the incremental
effects of policy options in terms of infrastructure deployment, environmental impact, and
digital sovereignty. The evolution of these challenges under the baseline scenario
illustrates why EU-level action may be warranted. The objectives of the intervention, as
defined in the problem tree, relate directly to the outcomes described above:
Table 69. Link between Policy Option 0 and the intervention
Problem
Driver Problem
Policy Option 0
(baseline trend) Link to Intervention Objectives
PD1: Lack of
scale and scope
of European
cloud and AI
computing
service
providers
PD2:
Bottlenecks to
expand data
centre capacity
in the EU
P1: Limited
and
geographically
concentrated
availability of
computing
capacity in the
EU
• Supply grows but is
outpaced by demand for
compute
• EU share of global compute
declines.
• Average project lead time ≥
3 years; grid and permitting
delays persist.
• Enterprise AI use rises to 40
%, below Digital Decade
target (75 % cloud, 75 % AI
use in enterprises).
“SO1: Increase computing capacity in
the EU through innovative and
sustainable technologies”: an
increase in capacity is foreseen with a
9% CAGR. However, the industry’s
demand for energy will continue to
grow. Without strategic energy
planning and a focus on sustainable
infrastructures, data centre expansion
will challenge existing data centre
hubs and regions with high strain on
natural resources, at the risk of
crowding out electrification
objectives in other sectors and
generating increasing public
opposition.
“SO2: Ensure attractive conditions
for the deployment of sustainable and
innovative computing capacity”: This
goal will not be met consistently
throughout the EU, with geographical
imbalances and potential price
increases across regions.
PD3: Limited
public sector
uptake of cloud
and AI
P2:
Dependence
on cloud and
AI computing
• Market share of non-EU
cloud providers remains 75-
80%.
SO3: Decrease the overall reliance on
non-European cloud and AI
computing services: Without
incentives for European providers to
211
Problem
Driver Problem
Policy Option 0
(baseline trend) Link to Intervention Objectives
computing
services
supplied by
European
providers
PD4: Absence
of clarity
around the
concept of
sovereign
cloud and AI
computing
services
services
supplied by
non-European
providers
• Cloud use in public services
remain
• Fragmented definitions
persist; sovereignty and
limited interoperability.
grow in their integrated offers as well
as in the demand by main drivers of
the economy such as the public sector,
the dependency from non-EU
providers would increase or remain
the same in a positive scenario.
SO4: Contribute to the protection of
fundamental interests of society by
enhancing the resilience of supply of
cloud and AI computing services, in
particular in the public sector:
Sovereign-washing would be likely to
happen without a clear definition of
sovereign cloud and AI computing
services and the associated
enforcement mechanisms.
ANNEX 10. CADA INTERVENTION LOGIC
ANNEX 11. OPERATIONAL MONITORING AND EVALUATION
SYSTEM
The monitoring and evaluation framework has been designed as an integrated system that
follows each level of the intervention logic, from outputs to outcomes and long-term
impact, based on a coherent set of indicators and baseline values with targets, where
available. The monitoring activities will be coordinated at EU level, with the European
Commission (EC, DG CONNECT) acting as the coordinator, with the support of relevant
EU agencies and Member State authorities for possible data provision. The framework is
defined to assess the indicators throughout the lifetime of the intervention and remain
coherent with the five-year evaluation horizon.
Below are individual sections that describe how outputs, outcomes and impacts will be
tracked over time, based on a coherent set of indicators for each level of the intervention
logic (see Annex 10), accompanied with baseline values and targets. Each table presents
data sources, collection method, frequency of collection and the responsible entity for each
data input.
The underlying assumption of this approach is that digital policies produce effects through
various, interrelated channels, i.e. from investment, regulation, energy systems,
infrastructure, market dynamics and user behaviour, and over different time horizons. As
such, the monitoring and evaluation process is to be intended not only as a tool to
understand if the policy objectives have been met but also as a learning mechanism to
better inform future policy implementation.
1. Output monitoring: deliverables and direct effects
At the output level, monitoring focuses on the direct and immediate results of the
intervention, i.e. how the tools produced by the Cloud and AI Development Act have an
immediate effect on processes and procurement dynamics. These outputs can be tracked
using administrative data and information reported by operators and Member States,
typically quarterly and at the end of the first year/18 months after implementation. Their
goal is to ensure that the intervention is proceeding according to plan, and offer early
warning on possible bottlenecks, which may need corrective measures during the
implementation phase.
Indicator Baseli
ne Target
Data
source
Collection
method Frequency
Responsi
ble entity
No. of national
strategies for data
centre and AI
infrastructure
deployment
<10
(out of
27 MS)
>20
(out of 27
MS)
National
authorities/
Ministries’
websites
Desk
research
Annual EC with
MS
validation
(if needed)
No. of identified
DC fast-track
areas
63197 >5 per MS in
the first year
of
National
authorities/
Ministries’
Desk
research
Annual EC with
MS
validation
197 The French government has published a list of sites judged favourable for data-centre implantation to help attract
infrastructure investment and compete at the European level. See here: 25112025__Guide Datacenters.pdf; The German
government is developing a national strategy to promote the operation and settlement of data centres: National Data
Centre Strategy - Federal Ministry for Digital and State Modernisation; Similar, Italy has developed a strategy for
attracting investment in data centres. See: ATTRAZIONE_DEI_DATA_CENTER_2025.pdf. Portugal is finalising its
strategy: Governo anuncia estratégia nacional de centros de dados para alimentar procura por Inteligência Artificial -
214
Indicator Baseli
ne Target
Data
source
Collection
method Frequency
Responsi
ble entity
implementati
on
Public funding
instruments
operational for
energy-efficient
technologies/
strategic projects
NA Definition of
calls with
clear
eligibility
rules and
selection
criteria
Internal NA Annual/
every two
years
EC
Monitoring
framework for
monitoring data
centre capacity
NA Defined
methodolog,
reporting
templates
Existing
methodolo
gy and
industry
datasets
Survey +
desk
research +
interviews
Once, to
implement
PM7
EC
Cloud and AI
development
plans
NA >20
(out of 27
MS)
National
authorities/
Ministries’
websites
Desk
research
Annual EC with
MS
validation
(if needed)
Compliance rate
with the
sovereignty
scheme by
contracting
authorities
NA Risk
assessments
and guidance
consistently
applied in
major public
tenders
TED data,
national
authorities
data
Desk
research +
interviews
with MS
representati
ves
Annual, 1
year/ 18
months after
implementati
on
EC with
MS
validation
Number of
sovereign services
within the
repository
NA Repository
data
NA Annual EC
Joint
tenders/procurem
ent mechanisms
launched
None Functioning
joint calls /
tenders
TED data,
national
authorities
data
Desk
research +
interviews
with MS
representati
ves
Annual, 1
year/ 18
months after
implementati
on
EC with
MS
validation
EuroCloud
platform
establishment
NA Functional
platform
within 1-2
years
NA NA Annual, 1
year/ 18
months after
implementati
on
EC
No. of public
sector open
source solutions
released
NA Continuous
increase
OSS
catalogue
Desk
research
Annual EC
2. Outcome monitoring: mid-term change
Outcome monitoring assesses the medium-term impact of the intervention on market
dynamics, user behaviours and overall ecosystem. Related indicators can be measures
Expresso; Netherlands has developed a roadmap for the growth of data centres in the country (2019): Opmaak 1;
Finland has recently developed a national roadmap for data centres. See: National Roadmap for Data Centres :
Rapporteur's Report; Spain also developed a national strategy for IA, with measures on the deployment of data centres.
See: Estrategia Inteligencia Artificial 2024
215
through market data, structured surveys, procurement records etc. Data collection could
correspond to the evaluation phase of the intervention. The monitoring framework is
designed to identify key data and ensure that this is collected from a wide range of
stakeholders affected by the policy intervention.
Indicator Baseline Target Data source Collection
method Frequency
Responsible
entity
Installed
computing
capacity in
the EU (IT
load) and
across MS
(outside
existing
hubs)
12 GW Additional
operational
IT load and
AI
infrastructure
by 2030
Industry
market
studies and
datasets
(EUDCA,
DataCentre
Map)
Desk research
+ survey (see
methodology
above)
Annual and
interim
snapshot, 5
years after
implementat
ion
EC with MS
validation
Share of EU
global
computing
capacity
20% ≥ 25% Industry
datasets,
market
research
Desk research Same as
above
EC
No. of MS
with
simplified
frameworks
(permitting,
zoning,
energy
access)
<5 ≥ 20 by 2030 National
authorities
Desk research
+ interviews
Biennial EC
Average
permitting
time for new
DC projects
Average
time: 32
months
<18 months
by 2030
and/or
reduction
across MS
National
authorities,
data centre
operators,
real estate
analysts
Desk/market
research +
interviews +
surveys
Annual EC with MS
validation
No. of pilot
projects for
innovative/
energy-
efficient DC
technologies
NA Continuous
increase
Project
registries and
evaluation
Desk research
+ Survey +
interviews
Annual EC
Annual
public and
private
investment
in EU DC
NA Increase to
close the
capacity gap
Market
analyses
Desk research
+ market
research
Annual EC
Improved
energy and
water
efficiency of
deployed
capacity
PUE,
WUE,
utilisatio
n levels
today
(see
Annex 4)
Convergence
towards best-
available
performance;
Share of
supported
sites using
specified
cooling/powe
r
technologies
TSOs/ DSOs
data,
providers
data
Desk research
+ Survey +
interviews
Annual EC with MS
validation
216
Indicator Baseline Target Data source Collection
method Frequency
Responsible
entity
Share of
renewable
energy used
by data
centres
NA Convergence
towards best-
available
performance
TSOs/ DSOs
data,
providers
data
Desk research
+ Survey +
interviews
Annual EC with MS
validation
No. of
audited
sovereign
cloud and
AI services
0 Continuous
increase
Repository Administrativ
e data
Annual EC
Market
share of EU
providers
15% Increase in
market share
Market data Desk research Annual EC
Annual
value of EU
procuremen
t of
sovereign
cloud and
AI
computing
services
0 ≥ 100% TED data,
national
authorities
data, audited
service
repository
MS reporting
obligations +
Desk research
+ interviews
with MS
representative
s
Annual EC with MS
validation
3. Impact monitoring: long-term structural effects
Impacts concern the long-term changes, i.e. materialised over longer horizons (often
beyond 5 years) in the EU’s economic, technological and strategic position, following this
policy intervention. Since these effects are expected to emerge progressively, they cannot
be assessed through monitoring alone but should be the outcome of a more thorough
analytical exercises which is able to disentangle the effects of the intervention from other
policy choices or natural market developments.
Indicator Baseline Target Data
source
Collection
method Frequency
Responsible
entity
EU computing
capacity
covering EU
demand
2025
Gap: 3
GW
Future
residual
gap < 0.5
GW
Operator
disclosures;
ENTSO-E
grid
connection
data;
industry
market
studies
Desk
research +
survey (see
methodology
above)
Annual (Year-
5 interim
evaluation
synthesis)
EC with MS
validation
Lower carbon
intensity of
digital
infrastructure
Estimated
kgCO₂e
per
compute
unit (see
Annex 4)
Decreasing
kgCO₂e per
compute
unit
Operator
energy
reporting;
national
grid
emission
factors; life-
cycle
assessment
parameters
Desk/
market
research +
Survey +
interviews
Annual (full
assessment at
Year 5)
EC with MS
validation
217
Indicator Baseline Target Data
source
Collection
method Frequency
Responsible
entity
Market
contestability
and higher
global
competitiveness
of EU industry
Low More
bidders,
lower lock-
in signals,
higher
market
shares for
EU
providers
Public
procurement
databases
(TED),
portability
metrics
from
providers
Tender
analytics +
interviews
with
providers
and users +
survey (e.g.
SME target)
At least 5 years
after
implementation
EC/others
Improved
strategic
autonomy and
reduced
dependencies
Baseline
study
Structural
reduction in
dependency
Market
sizing;
customs
statistics;
provider
data
Desk
research
Annual + Year
5
EC/others +
MS and
relevant
agencies
4. Evaluation arrangements and timing
The evaluation will be structured to understand how the observed impact and changes can
be attributed to this specific policy intervention. The study conducted as part of the
monitoring of capacity deployment during the first year of implementation will be used to
validate indicators, data sources and targets. A mid-term evaluation can be conducted after
the third year of implementation of the intervention, to better understand the early
outcomes and effectiveness of the intervention (using data collected above). The full
evaluation is then expected five years after the implementation of the intervention to assess
the outcomes and identify early impacts.
To prevent parallel reporting, avoid duplication and benefit from synergies with existing
initiatives, the programme should reuse:
• Eurostat business statistics
• National regulatory authorities and BEREC reporting
• Procurement electronic databases, e.g. TED
• Energy systems data from TSOs and DSOs, national energy agencies
• Existing and future programmes, such as the Digital Decade Policy Programme, RRF
reporting templates
A preliminary set of evaluation questions can be found below. This will be updates and
enriched following the implementation of the initiative.
Evaluation
criterion Assessment objective and possible evaluation questions
Effectiveness
Goal: assess the extent to which the objectives of the initiative have been
achieved and how benefits have accrued to different stakeholders.
To what extent did the intervention increase EU installed computing capacity
and create the conditions for easier data centre deployment? How did it foster
the development and deployment of innovative and sustainable data centres
and a better use of energy sources? To what extent did it increase clarity
218
Evaluation
criterion Assessment objective and possible evaluation questions
around the concept of sovereign cloud and AI computing services? How did it
improve the market share of homegrown EU cloud and AI computing service
operators? What is their market share? To what extent did it increase federated
resources across the public sector and joint procurement for cloud and AI
computing services? To what extent did it increase the use of open source
solutions?
Efficiency
Goal: assess the extent to which the initiative has been cost-effective, analysing
the relationship between expected and actual benefits and costs.
Have benefits and cost savings been achieved at proportionate costs for
different stakeholders?
Relevance
Goal: assess the extent to which the objectives of the initiative still reflect
current and future needs.
To what extent the initiative still addresses relevant needs? How is it still
aligned with EU priorities?
Coherence
Goal: assess the initiative’s internal and external coherence, i.e. if the different
elements of the intervention worked together to reach the set goal and if it
worked well or overlapped with other initiatives, both at EU level and national
level.
To what extent is the initiative consistent with existing and future energy,
digital, competition, environmental, security rules at EU level and national
level?
EU Added
value
Goal: assess the extent to which the initiative brought EU added value
compared to what could have been achieved by Member States alone.
To what extend did EU-level action prevent fragmentation of DC rules? How
did it improve cross-border service delivery and competitiveness of EU
providers?
219
ANNEX 12. COSTS OF MIGRATING AND PORTING APPLICATIONS
Cloud porting consists in moving a service from one cloud provider to another cloud
provider
Cloud migrating consists in moving a service from on-premises to cloud
Today, by and large critical use cases [of the public sector] are not in the cloud. Most of
these use cases are instead run ‘on-premises’, that is, on local infrastructure which are not
scalable (and cannot therefore be called ‘cloud’). Policy measure 21 creates an obligation
for the public sector to carry at least one sovereignty risk assessment to identify which
public sector use cases within a Member State require the use of which sovereignty level
as described under PM11. The details of the measure are described in the main text (section
5.2.2) and in annex 4 (section 3.21). None of the proposed policy measures obliges public
authorities to port an existing cloud service from one provider to another, or to migrate on-
premises service to the cloud or to port. These decisions are made on a case-by-case basis
by public authorities alone, considering the authority's specific needs, existing systems,
and procurement cycle.
Examining costs is however relevant since the results of Member States' sovereignty risk
assessments are expected to induce cloud porting decisions198. As well, the expected
increase in trust resulting from the existence of a sovereignty framework is expected to
increase or anticipate cloud migration decisions. These costs are not immediate
consequences of the intervention's entry into force, but rather potential future expenses that
may arise during the normal course of business.
12.1 Considerations about migrating and porting
Applications can be broadly qualified in different levels of cloud-readiness:
• Legacy ‘on-premises’ applications that are not candidates for cloudification, for
example because they are to be phased out, because their use is too small, or
because the technology is too old to migrate them. As an example, an important
number of banking applications use still today services written in COBOL, a
language popular in the decade of the 60s.
• Cloud-aware applications: A cloud aware application is designed or adapted to
leverage cloud computing features but nevertheless present challenges that span
security, architecture, evolvability, and scalability.
• Cloud native applications, specifically designed to benefit from the capabilities of
the cloud, such as scalability.
In the public sector, most applications fall within the two first categories: when
cloudified/cloudifiable, they are ‘aware’ but not ‘native’, which requires important
198It must be noted that porting can also be triggered naturally as a consequence of regular procurement decisions after
the expiration of cloud service contracts.
220
adaptations (e.g. a re-architecting or a refactoring of the code) when seeking to port them
to another cloud service.
There is no one-size-fits-all when migrating (legacy-to-cloud) or porting/switching
services (cloud-to-sovereign cloud), as it largely depends on many aspects: the technology
used to develop the application, the size, its complexity, or third-party dependencies. This
process is not automated and is most often a one-off effort that can hardly be repeated,
even if economies of scale can exist. Literature explains the process, both from a technical
point of view and a business point of view199 but there are seldom reports available on the
costs associated to migrating or porting/switching, given that this is mostly an ad hoc, one-
off activity with many variables interplaying.
At the time of migrating or porting / switching, some key cost elements that should always
need to be considered are:
• The size, technology, complexity, and third-party dependencies of the system as-is
• The specificities of the to-be environment.
This will decide how the application will be ported / migrated. In the industry, the 7Rs
of migration200 provide information on what is the most adequate approach for the
problem at hand. In legacy-to-cloud and cloud-to-sovereign cloud migration, these are
the approaches most commonly followed in the industry:
o Rehost or lift-and-shift, which usually require minimal code changes. This
is usually the fastest option to put a service into production, but unless
already cloud-native, it incurs into a huge technical debt and post-migration
optimization costs. Overall, while the upfront cost of the migration is small,
the long-term benefits will be limited, and a large proportion of costs will
be spent in optimizing the application to reduce the spendings.
o Replatform, where the application is adapted to managed databases or to
other technologies such as containers. This is often the ‘by-default’ option
that offers the stronger Return on Investment (ROI) over time. Also, this is
a much more beneficial option in the mid-term than rehosting an application
and can lead to significant operational savings.
o Refactor, which requires major or full code rewrites and a new architecture
design of the application (e.g. from an n-tier to micro-services or
containers), as well as new continuous integration continuous deployment
(CI/CD) pipelines, among other considerations. This is a rather costly
project and risk-prone, and it is only implemented if the application is
expected to deliver large long-term benefits and operational efficiency that
outweigh the costs.
199 Orue-Echevarría Arrieta, L. (2016). From software as a good to software as a service (SAAS): a methodology to
define the transformation towards the SAAS business model. Universitat Abat Oliba.
https://hdl.handle.net/10803/398024 200 The 7Rs of migration include the following: Rehost (lift-and-shift): Move the application to cloud with minimal
changes, Relocate: Move VMs without OS or code changes, Re-platform: Moderate changes to leverage managed
services, Refactor (rearchitect): Major or Full rewrite for cloud-native architecture., Repurchase (drop-and-shop):
Replace with a SaaS equivalent, Retire: Decommission application, Retain (revisit): Keep on-premises for the time being.
Adopted from IBM Strategy
221
The cost drivers differ depending on whether it is a porting from a cloud service to another
(cloud-to-sovereign cloud) or a migration of a legacy application to the cloud. (legacy-to-
cloud) The dominant cost driver of porting an application from one cloud service to
another is the labour related to the service re-mapping, re-testing and parallel run in
dual infrastructures, and a potential re-adaptation of the application away from
proprietary services or older versions of the technologies. Thanks to the entering into
force of the Data Act, egress fees are in the process of disappearing or have already
disappeared. Conversely, when moving a legacy application to the cloud
(‘cloudification’), the dominant cost is the technical team’s labour – architects,
software engineers, DevOps - to resolve the technical debt, undocumented
dependencies, and architectural incompatibilities with cloud-native patterns, along
with the decision to what cloud service to migrate to, both from a technical and a
business point of view, and the impact in the organizational processes. The dual run
during the cloudification while critical, does not consume any additional cloud resources,
as the legacy application keeps on running on-premises.
Due to the singularity of each migration and porting project and the lack of data from
independent resources, the current analysis addresses two distinct migration scenarios in
qualitative terms:
(1) migrating an existing on-premise workload to the cloud (legacy-to-cloud); and
(2) porting an existing workload from one cloud provider to another (cloud-to-
sovereign cloud).
The second scenario is presented in the subsequent sections in quantitative terms. This
scope decision reflects a limitation in the available literature: cost data for legacy-to-cloud
migrations remains highly variable and context-dependent, making reliable generalisation
difficult. Cloud-to-sovereign cloud migrations, by contrast, offer a more stable basis for
estimation, as the source environment is often already well documented, the application
containerised, and it is operating under known cost parameters, considerations that have
been used for the quantitative modelling presented in the next sections.
While both scenarios follow the same broad phases, the effort intensity of the individual
activities within each phase of a migration project – that is, how much labour would be
dedicated to that specific activity - differs considerably depending on whether the project
is a legacy-to-cloud or a cloud-to-sovereign cloud migration. The table below provides a
qualitative comparison of the effort intensity per main activity across both scenarios.
Table 65. Comparative table of the intensity effort (qualitative) needed to migrate applications legacy-
to-cloud and cloud-to-sovereign cloud (Source: European Commission)
Activity Legacy-to-cloud effort
intensity
Cloud-to-sovereign cloud
effort intensity
Architecture design Medium / High Low
Refactoring and code migration High Medium
Data migration High High
Service re-mapping N/A High
Testing and validation High Medium
Parallel operation (dual run) Low Medium
Tooling and licenses Medium Low
222
Activity Legacy-to-cloud effort
intensity
Cloud-to-sovereign cloud
effort intensity
Training in new technologies Medium / High Low
Some considerations on the above effort intensities:
• Architecture is medium / high in legacy-to-cloud: mapping undocumented
dependencies of on-premise software is genuinely complex, often requiring reverse
engineering techniques. The output is however well defined; it is a target
architecture and a wave plan (migration plan). In cloud-to-sovereign cloud the
effort intensity in architecture is lower, because usually the application deployment
and cloud architecture are already often well documented and reverse engineering
techniques, if needed, are much less effort consuming. In this case, the main
challenge faced by the architects is to achieve a well and clear structured service-
to-service mapping rather than a service discovery and identification from scratch.
• Refactoring and code migration is high in legacy-to-cloud. Resolving technical
debt, refactoring the code, changing the programming language and technologies,
the application of cloud-native or cloud-aware patterns (e.g. load balancers),
containerising monoliths or n-tiers, and creating CI/CD pipelines from scratch,
represent the hardest engineering work in the entire migration and the one most
prone to scope expansion when hidden dependencies surface in the middle of the
project. In cloud-to-sovereign cloud this effort is considered to be medium, because
the application is already cloud-native or cloud-aware. Here, the work is mainly
adaptation, not transformation.
• Service Re-mapping is high in cloud-to-sovereign cloud and not yet applicable in
legacy-to-cloud. Translating proprietary managed services across platforms
demands deep simultaneous knowledge of both providers, and incompatibilities in
data models or execution environments could trigger some re-engineering that was
invisible at planning time. This is why it is critical to carry out this mapping in
earlier phases of the migration.
In any case, the migration or porting should be balanced against the costs of not doing
anything:
• Maintaining legacy applications carries a real and often underestimated cost.
Beyond the direct expense of keeping ageing infrastructure operational, including
on-premise data centres that frequently fall short of modern standards,
organisations must continuously maintain all dependent software components,
manage growing security and vulnerability exposure, and remain with a shrinking
pool of engineers who retain the relevant skills as technologies become obsolete.
Legacy systems also impose additional overhead cost: the older the application, the
more difficult and expensive it becomes to modify, extend, or integrate,
progressively limiting an organisation's ability to respond to changing
requirements.
• Remaining with a non-sovereign provider poses a dependency risk not without
cost. Vendor lock-in forecloses the price competition that periodic provider
switching would otherwise enable. More significantly, a non-sovereign service
223
exposes the organisation to geopolitical risk: a provider subject to third-country
jurisdiction may face legally compelled data access requests or pressure to degrade
service quality, neither of which the customer can effectively resist or even
anticipate. These risks carry a cost, whether measured in compliance exposure,
quality of service or the political cost of dependency, that cannot be easily
quantified.
The final decision follows a total cost of ownership approach and, in addition to the above,
would also have to consider aspects such as the acceptance of the technical debt, the level
of dependencies from third-party libraries or proprietary source code, the business case,
the return of investment or the risk appetite. As explained above, this decision needs to be
taken ad-hoc for each application to be ported or migrated. The prioritization of which
applications should be ported is a business-driven decision, often based on which
application would report a quicker and bigger return on investment (ROI).
12.2 Cost modelling for public administrations to port services (cloud-to-
sovereign cloud)
A cost analysis has been carried out to estimate the expenditure involved in porting an
application from one cloud service offered by a cloud service provider to another that is
functionally equivalent but either provided by another provider or being the same provider,
under another type of technical and legal construct.
Published literature on this subject is scarce, largely owing to the highly context-specific
nature of cloud migrations: most available sources are online, deal in broad ranges, and
provide little explanation of the underlying cost drivers, or the methodology used to derive
their estimates the phases considered for the migration, or the type and size of the
application ported.
Despite these limitations, and in the absence of more rigorous or detailed benchmarks,
these online sources remain the only externally available reference point against which the
results of this analysis can be validated. They have therefore been used as cross-validation
references, with appropriate caution, alongside resources consulted in confidence from
existing projects. Given the inherent uncertainty in this type of estimation, a min-max
sensitivity analysis has also been conducted to illustrate the plausible range within which
porting costs may fall depending on the application characteristics and migration approach.
The starting point for this assessment is that sovereignty risk assessment will identify
public sector IT systems that will need to be ported to sovereign cloud services of at least
sovereignty level 2. To follow a conservative approach, it is assumed that derogations will
be exceptional. Moreover, it is assumed that the porting will be from one cloud service to
one cloud service (1:1 cardinality). Migrating to multiple cloud services at the same time
would add an important layer of complexity and risk to the project201. Most EU public
201 Alonso, J., Orue-Echevarria, L., Casola, V. et al. Understanding the challenges and novel architectural models of
multi-cloud native applications – a systematic literature review. J Cloud Comp 12, 6 (2023).
https://doi.org/10.1186/s13677-022-00367-6
224
administrations are still in the process of adopting cloud services202 under a “cloud first
approach”203, and while deploying workloads following a multi-cloud204 strategy is a
plausible one, this is still a marginal practice.
While it has been reported by Member States in various events (e.g. Member States Cloud
Coordination Group (MSCCG) as part of the European Alliance for industrial data, edge
and cloud) that the migration from on-premise applications to cloud-aware and cloud-
native is growing, there is few available public data that clearly quantifies this growth.
France reported in 2025 EUR 84 m of public procurement for cloud - in State
administration, healthcare, local administrations, budget operators, public establishments
or similar, authority or institution - with an annual growth of 62%. The report stresses that
these values stem mostly from small projects and a handful of large ones, positioning
France in a cloud adoption trend205. Spain and Germany are investing in their own
governmental clouds, NubeSara (EUR 84 m206) and Deutscheverwaltungslcoud
respectively but there is no clearly available disaggregated information on the number of
workloads on the cloud. The UK, through its G-Cloud offering reports an expenditure of
£400 m in hosting services in fiscal year 2024/2025207.
202 See France as an example, in their document Bilan de la doctrine cloud au centre where it is stated: “La dynamique
actuelle d'arrivée de nombreux petits projets montre que nous sommes toujours dans une phase d'adoption” (p.6). This
can be extended to various MS, which as part of their RRF plans are in the process of migrating their legacy applications
to cloud-aware and cloud-native architectures. 203 See for instance Spain, with its approach “hybrid cloud first” here and here 204 There are several ways to define multi-cloud. Alonso, Orue-Echevarria et al (2023) concluded in their Systematic
Literature Review (SLR) which analysed 88 sources from an initial set of 1 000 papers that multi-cloud applications can
be characterized as follows: (a) Replicated multi-cloud applications are deployed across multiple clouds sequentially
rather than simultaneously. They migrate between providers for economic or operational reasons, such as for cost
optimisation, backup needs, emergency failover, or contract expiry. They are purpose-built to run on, and transition
between, different cloud environments. The primary concern is seamless switching between providers, making data
portability and interoperability central design requirements. (b) Distributed multi-cloud applications run subcomponents
concurrently across resources from multiple providers, drawing on several independent cloud services in combination.
This simultaneous multi-provider model delivers tangible benefits including high availability, fault tolerance, and cost
efficiency. It is typically adopted when no single provider offers the full functionality an application requires, allowing
organisations to selectively consume the services best suited to their needs within given cost constraints. However, this
approach introduces significant complexity: security governance becomes harder to enforce, operational management
grows more demanding, and the distributed architecture raises design-phase challenges such as the partitioning of
application logic and data across provider boundaries. (c) Hybrid multi-cloud applications encompass both of the above
patterns: combining the sequential provider usage characteristic of replicated deployments with the concurrent, multi-
provider model of distributed architectures, offering the broadest flexibility but also the greatest management overhead. 205 Bilan de la doctrine cloud au centre 206 España digital 2026 with a breakdown of the services at Boletín Oficial del Estado, mainly targeting computation
power, network and storage. 207 See G-Cloud tab, select “central government”, fiscal year “2025 – 2025”
225
12.3 Phases of a porting operation
Next, the different phases are described, adapted from the few existing resources in the
literature208 209 210 211 212.
Phase 1 – Feasibility assessment (one-off adjustment costs): The first step before
tackling the porting is to have a clear understanding of the problem that the application is
solving, perform a technical feasibility analysis to understand well the complexity, the size,
dependencies as well as the technology, architecture, licenses, compliance requirements
and quantity of data. In addition, it is important to have a clear understanding of what are
the features and technologies accepted by the target service as this will determine the
decision on which porting model will be followed.
The following activities have been quantified under this phase taking as input the
information provided above and consultation with experts and practical experience in the
field:
• Discovery tooling and licensing
• Infrastructure inventory
• Dependency mapping
• Total Cost of Ownership (TCO) analysis
Among the most important challenges to be faced in this step is to underestimate the
dependencies across modules (end-to-end calls), the lack of documentation which may
result in an incomplete inventory and data classification gaps, underestimating the
existence of Personal Identifiable Information (PII) or regulated data in unexpected
components requiring at a later stage, for instance a DPIA.
The main outcomes of this phase is a porting inventory, a dependency graph, risk register,
and a porting readiness assessment report.
Phase 2 – Strategy and planning (one-off adjustment costs): The second step of the
porting is the development of the porting strategy and the planning itself. At this stage,
once the characteristics of the application and the target platform are known thanks to the
first step, it is a matter of deciding, based on the 7Rs of porting, what is the most adequate
approach for the problem at hand: rehost, re-platform or refactor, as the most widely used.
208 Andreas Menychtas, Kleopatra Konstanteli, Juncal Alonso, Leire Orue-Echevarria, Jesus Gorronogoitia, et al..
Software modernization and cloudification using the ARTIST migration methodology and framework. Scalable
Computing : Practice and Experience, 2014, 15 (2), pp.131-152. ⟨10.12694/scpe.v15i2.938⟩. 209 J. Alonso, L. Orue-Echevarria, M. Escalante, J. Gorroñogoitia and D. Presenza, "Cloud modernization assessment
framework: Analyzing the impact of a potential migration to Cloud," 2013 IEEE 7th International Symposium on the
Maintenance and Evolution of Service-Oriented and Cloud-Based Systems, Eindhoven, Netherlands, 2013, pp. 64-73,
doi: 10.1109/MESOCA.2013.6632736. 210 Orue-Echevarría Arrieta, L. (2016). From software as a good to software as a service (SAAS): a methodology to
define the transformation towards the SAAS business model. Universitat Abat Oliba.
https://hdl.handle.net/10803/398024 211 Pahl, C., Xiong, H., Walshe, R. (2013). A Comparison of On-Premise to Cloud Migration Approaches. In: Lau, KK.,
Lamersdorf, W., Pimentel, E. (eds) Service-Oriented and Cloud Computing. ESOCC 2013. Lecture Notes in Computer
Science, vol 8135. Springer, Berlin, Heidelberg. https://doi.org/10.1007/978-3-642-40651-5_18 212 P. Jamshidi, A. Ahmad and C. Pahl, "Cloud Migration Research: A Systematic Review," in IEEE Transactions on
Cloud Computing, vol. 1, no. 2, pp. 142-157, July-December 2013, doi: 10.1109/TCC.2013.10.
226
Each option has different consequences in the effort needed to port, the timeline and the
risks.
Once one of the above strategies have been decided upon – this can be unique for the
complete application or alternatively each part of the application may undergo a different
strategy - , the next activities include grouping the workloads into migration waves213,
addressing the identified dependencies, mapping the source services with the target
services, re-architecting where needed, define the data porting strategy (e.g. bulk, transfer,
incremental, …) as well as setting up the porting programme team. As part of this activity,
it is also important to keep a rollback mechanism defined in case the porting does not
succeed.
These are the activities evaluated in this phase:
• Cloud architecture design
• Porting plan
• Provider (target platform) feature evaluation and service re-mapping
• Project management setup
The main challenges in this phase stem from the gap in the team skills and the lack of one-
to-one match across the functionalities of the various cloud services. That is, while cloud
services may theoretically be similar or may have an equivalent functionality, there is not
a one – to – one match. This matching must be often manually performed therefore by the
team members start carrying out the porting.
The outcomes of this phase are the porting strategy document, the plan, the target
architecture design and topology and the RACI214 matrix.
Phase 3 – Target environment set up (one-off adjustment costs): This step includes
the provision and configuration of the infrastructure (e.g. the Virtual Machines),
implementing the basic security policy measures (e.g. identity provider), observability
environment and logs, the CI/CD pipelines and the Infrastructure-as-Code (IaC) practices
on the target environment.
Under this phase, these are the activities considered:
• VM / compute provisioning
• Network configuration (Virtual Private Cloud, DNS, Firewall)
• Storage setup
• Identity and Access Management, access and monitoring setup
• IaC scripting and DevOps pipelines
213 A migration wave breaks down migration projects into more manageable batches for synchronized and concurrent
migration. This is an iterative approach that allows organizations to minimize business disruption and risk, as well as
incorporate the lessons learned from the previous wave into the next one. 214 Responsible, Accountable, Consulted, Informed
227
The main obstacles in this phase have to do with the potential differences in the Identity
and Access Management (IAM) roles and permissions, overlapping IPs, the setup of tools
such as security scanning tools or a SIEM (Security Information and Event Management).
The outcomes of this phase are the network topology, the IAM baseline and the DevOps
pipeline.
Phase 4 – Porting (one-off adjustment costs): This phase entails the porting of both the
business logic of the application as well as the data, while maintaining the integrity, full
auditability and minimal downtime of the application. The porting of data may be rather
challenging, in terms of duration, strategy, and risks and they require a verification of the
integrity of the data at destination (e.g. row counts, checksum, etc) as this is a very error-
prone activity. Migrating the business logic may involve refactoring part of the code, re-
adapting to new technologies, moving and / or rebuilding the images for the containers,
adapting the Kubernetes manifests, redeploying them and reconfiguring certain parts of the
application such as the managed services (e.g. load balancers). This step also includes some
smoke tests (initial tests) to demonstrate that the application is functional and responsive.
Following, the activities considered:
• Database porting
• VM/server image replication
• Application code adaptation (refactoring) and configuration for porting
• Porting of the tooling licenses
• Parallel infrastructure dual - run
The main problems usually faced during this phase is the data format compatibility, the
maintenance of the integrity of the data and the avoidance of data loss during the porting.
In terms of application, the integrability and reconfiguration of the different components
migrated onto the target platform may hinder the deployment of the application.
The deliverables after this phase is the data and application migrated onto the cloud, and
the application deployment view.
Phase 5 – Testing and validation (one-off adjustment costs): This step includes the
functional and non-functional testing, including performance, load, stress and resilience
testing. Security, vulnerability and penetration testing are also part of this. At this stage,
there is a disaster recovery plan in place and clear metrics to ensure business continuity.
These are the activities taken into account under this phase:
• Functional testing
• Performance and load testing
• Integration / API testing
• Security and vulnerability scanning
• Rollback plan validation
The main obstacles in this phase entail usually the test coverage and the responsiveness of
the application in extreme conditions. Another key aspect is that the targeted application
228
must mirror in functional terms the migrated one so there has to be a parallel run of both
applications for a certain period of time in order to define errors before it goes into
production.
The outcome of this phase is the test results from the different tests carried out.
Phase 6 - Cutover and Switching DNS (one-off adjustment costs): This is the most
critical phase as it involves the switching from the ‘old’ environment to the target
environment. When going on production, the DNS records are switched, the application
health needs to be ensured, and the code is frozen. The definition of a rollback mechanism
is also part of this activity.
The tasks included under this phase are:
• DNS / traffic cutover
• Final data synchronization and delta transfer
• Go live
• Downtime window management
• Incident response and hotfixes
The challenges of these activities relate to the DNS switching and how stateful sessions
are being handled, that is active user sessions could be lost due to the cutover. The outcome
of this phase is the application in production.
Phase 7 – Post – porting (one-off adjustment costs): After a successful cutover, the
source environment will still be running but on standby to identify potential performance
problems in production. This parallel run is usually for a certain amount of weeks, for
example, 2-4 week, even if not the whole time. Moreover, this phase covers the
decommissioning of resources, the archiving of the source environment configuration for
audit purposes and the update of the architecture documents. Subscriptions or licenses shall
also be annulled.
Under this phase the following activities are considered:
• Cost optimization, performance and right-sizing (FinOps)
• Monitoring and alerting setup
• Old infrastructure decommissioning
• Staff training and knowledge transfer
The main challenges include the decommissioning of the source before the rollback
window expires risking permanent data loss as well as the engineers who led the porting
moving on before the documentation is complete. The outcome of this phase is an
application that is operational.
Phase 8 – Operation of the cloud service (recurring adjustment); This phase includes
the operation of the service, including observability, performance optimization, FinOps,
continuous vulnerability assessment, and maintenance of a functional application (e.g.
correction of bugs but also new features being deployed).
229
12.4 Aspects considered for the analysis of costs
Since not all use cases have the same complexity, for this impact analysis three types of
applications are assumed: small, mid-sized, large, as described below. The criterion are
industry standard values and were cross referenced with confidential sources:
Table 66. Application taxonomy used for this impact analysis
Criterion Small application Mid-size
application
Large application
Code base <10k lines, n-tier
(cloud-aware), small
number of
microservices
(cloud-native)
10k – 100k lines, 10
– 20 microservices
>100k lines, > 20
microservices or
smaller in number
but complex
microservices
Data volume < 100 GB 100 GB – 10 TB 10 TB – PB
Third party
dependencies
Few (1-3) Moderate (5 – 15) High (>20 or
tightly coupled)
Users < 500 users / < 50
concurrent
500 – 10,000 users /
50 – 500 concurrent
> 10,000 users / >
500 concurrent
Team size Small Medium Large
Typical
Estimated
Timeline
1 – 3 months 3 - 9 months > 9 months
Example
application
Internal tool / Small
application
Departmental Heavy stakeholder
based, such as
healthcare
applications
As noted above, reliable data on the cost of porting or migrating an application between
cloud providers is scarce. This scarcity is structural rather than incidental: each porting
project is inherently unique, shaped by how the application was originally architected, the
technologies involved, the degree of vendor-specific coupling, and the accumulated
technical debt. No two porting project present the same cost profile, which makes
meaningful generalisation difficult and explains why the literature has not converged on
robust benchmarks. The estimates presented in this section represent a best-effort attempt
to triangulate and structurally aggregate the limited sources available reporting costs215.
215 Cloud migration cost calculator provides a calculator to calculate where the user can insert the type of migration he
is targeting (public to public, public to private and to which of the large three hyperscalers, the criticality of the
application, the workload size, the data migration size, the migration strategy – rehost, replatform, refactor and other
aspects such as observability, landing zone, security hardening, DevOps pipelines, and team training). Cloud Migration
Cost Analysis: Lift-and-Shift vs. Refactoring; Software Modernization Costs provides empirical benchmarks from 200+
real-world cloud-to-cloud migration projects out which 35 have reported cost data. This source reports single-point
median figures per strategy at enterprise scale, without any distinction per application size Cloudaware: Cloud Migration
230
These sources follow different methodologies – application size count mapping, server of
VM count mapping, porting strategy per application type and empirical benchmarks from
real world cloud-to-sovereign cloud porting projects and show certain bias as they are
companies that offer porting services. Under these considerations, the structured
aggregated ranges of costs for porting applications, irrespectively of rehosting, re-
platforming or refactoring due to the lack of disaggregated data, can be considered as
follows:
- Small application: EUR 20 000 – EUR 300 000
- Mid-sized application: EUR 100 000 – EUR 500 000
- Large application: > EUR 500 000
Hence, they should be read as informed approximations rather than precise figures.
However, these have been cross-checked with confidential information on past and
existing re-platform porting projects in the public sector. The baseline used are as follows:
porting a small workload is estimated at a value of EUR 20 000 - 50 000, a mid-size
application as of around EUR 200 000 while the porting of large application is estimated
to cost around EUR 500 000. In terms of effort, confidential sources report that the porting
of a large application has been estimated in 1 000 activity days, which has been taken as
benchmark. In this case, this application will not only be moved to another cloud service
but will also undergo an adaptation of some of the technologies used such as changing the
proprietary web server and data bases for an open source alternative. This is also the
plausible scenario selected for the modelling for this impact assessment, given that it is the
go-to scenario in the majority of the cases.
There is a rationale behind the distinction of sizes of applications. The cost gap between
migrating small and large applications is not only just about volume but largely about
dependency and architecture complexity. A factor often underestimated in small
applications is the refactoring: the code may be small but poorly documented, with a high
technical debt. In mid-sized and large applications, the challenge primarily remains in the
integration points (e.g. APIs, databases), which are the main cost drivers and dominate the
complexity. Tightly coupled applications, or a bad designed microservices architecture
along with a lack of or a bad organizational change management process can outrun the
technical costs estimate.
The underlying assumptions for the quantification of the costs are detailed next.
• The modelling accounts for the porting of an application deployed on a VM (IaaS
to IaaS), but not from PaaS to PaaS or from SaaS to SaaS. While most of the
activities may be similar, less literature resources exist in this regard, rendering the
estimations highly complicated and uncertain.
• The modelling measures the porting costs of moving from a public cloud to a
sovereign cloud with sovereignty level 2 – 4. The baseline used for the public
Costs in 2026: Complete Guide to Savings, Benefits & Enterprise Planning uses server or virtual machine count rather
than application count as the primary sizing unit, which reflects an infrastructure-first (IaaS) perspective well-suited to
compute-migration analysis.; Cloud Migration Cost Analysis: Rehost vs. Refactor provides a representative numerical
case drawn from their own company, Opsio, which implements migration projects.
231
cloud is AWS eu-central (Frankfurt) and used the prices provided by the calculator
available online at the provider’s site (more details on the instance types selected
and their associated costs below), with a +5% price premium. Sensitivity
analysis is +12%, -10%216.
• The values for the costs are calculated considering a manual porting process
following the traditional software engineering process, as, even if limited, there
exist some literature sources in this regard. However, part of this porting could be
performed using AI automatic code generation, Agentic AI 217 218 219 or low-code-
no-code platforms220 that could have an impact on the costs. However, given the
novelty of this technique and its incipient use there is, for the time being, scarce
literature sources on the costs and on its results in terms of benefits and quality of
the source code generated221.
• The number of affected public essential entities under NIS2 Annex 1 equals to
6 400. However, not all essential entities make a similar use of cloud services. To
respond to this situation, the number of essential public entities has been stratified,
as depicted next:
o High-intensive users of cloud services: under this category fall national
administrations that make an important use of cloud services either because
they have a large proportion of applications on the cloud (both cloud-aware
and cloud-native) or the number of services that are deployed on the cloud
are low but make a high intense use of cloud services. An example of this
could be the national tax services. Based on empirical observations on how
the EU public sector is structured, the proportion of this type of users is
considered as 20%.
o Medium intensive users of cloud services, which include systems from
national ministries that are less intensive in terms of cloud usage and some
regional authorities / federal authorities that cover for instance healthcare,
or tax services (e.g. Autonomous regions of Spain). Based on empirical
216 This range is explained in section 2.3.4 of the main document. 217 Agentic AI is an autonomous system that leverages from specialized AI agents to analyse, refactor, transform, and
validate legacy codebases with minimal human intervention. Traditional code generation tools often only suggested
snippets of code or the structure of the application, whereas agentic AI promises to manage independently the entire end-
to-end migration process, including the decomposition of a monolithic application to migrating to the new language and
platform maintaining the full functionality 218 See IBM, Reimagining the application migration and modernization value-chain: The agentic way 219 Bandi, A., Kongari, B., Naguru, R., Pasnoor, S., & Vilipala, S. V. (2025). The Rise of Agentic AI: A Review of
Definitions, Frameworks, Architectures, Applications, Evaluation Metrics, and Challenges. Future Internet, 17(9), 404.
https://doi.org/10.3390/fi17090404 220 Low code platforms is a software development environment that allows for the development of applications using
visual, drag-and-drop interfaces instead of extensive manual coding, leveraging from model-driven design (MDD) and
pre-built components. No code platforms follow the same approach as low code platforms but in this case, there is no
need to write any line of code. While often used together, low code allows for the inclusion of custom code, whereas no
code is more restrictive and focused on non-technical users. Source: IBM 221 Wang, H., Gong, J., Zhang, H., Xu, J., & Wang, Z. (2025). Ai agentic programming: A survey of techniques,
challenges, and opportunities. arXiv preprint arXiv:2508.11126.
232
observations on how the EU public sector is structured, the proportion of
this type of users is considered as 40%.
o Low-intensive users of cloud services, upon which national ministries
with transactional digital public services are included and regional public
administrations. It is to be clarified that while the entities that manage water
are considered essential services, they are mostly at municipal level, which
help reduce the scope. Based on empirical observations on how the EU
public sector is structured, the proportion of this type of users is considered
as 40%.
• The usage of cloud services has been stratified as per the values above.
• Based on discussions with three distinct sets of public stakeholders that represent
~200 public authorities, the essential entities are considered to have each 1 200 IT
services, where 30% of the services are already cloudified and would be effect of
a cloud-to-cloud porting, 30% are on-premises and will be cloudified, and 40%
should remain on premise (e.g. legacy systems that are being phased out, very small
IT services).
• Based on the discussions mentioned above, the distribution of sovereignty needs
resulting from the risk assessment across the public sector are Level 1: 70%, Level
2: 20%, Level 3: 9%, Level 4: 1%.
• The effort to migrate a legacy-to-cloud application is estimated to be 1.5x the effort
estimated to port a service from cloud-to-cloud, taking as baseline the comparison
shown in previous sections.
• The ratio of applications in terms of size and types of cloud intensity users is
stratified as shown next. The proportion of small, mid-size and large applications
in the case of high-intensive users of cloud services has been obtained from the
experience of confidential sources. For medium and low intensive users of cloud
services, the values are estimated, taking as anchor value the potential proportion
of large applications that there could be in each type of entity.
Table 67. Distribution of types of users of cloud services in public essential entities operating under
NIS2 Annex I (Source: European Commission)
Small
App.
Mid-size
App.
Large
App.
Total
High intensive users of cloud services 50% 45% 5% 100%
Medium intensive users of cloud services 60% 37% 3% 100%
Low intensive users of cloud services 70% 29% 1% 100%
233
• Labour costs are considered as average costs of EUR 75 / h, irrespective of their
profile (cloud architect, developer, DevOps engineer, project manager or external
consultant). These costs are higher than the ones considered in the rest of the impact
assessment, given the large number of profiles participating in a porting of an
application.
• Out of the three approaches presented beforehand, the ‘re-platform’ option has
been selected, as being the most plausible one, since it combines porting the
application as-is, creating cloud-agnostic components that can facilitate a future
porting to another service provider and other common activities such as some
adaptations some of the used technologies or a small refactoring of the code.
• For the calculations on the pricing of VMs and storage, comparing sovereign
solutions level 2 and sovereign solutions level 3 / 4, internal knowledge has been
considered, complemented with desk research data that report, through use cases
and empirical analysis, similar prices for computation resources (VM, storage,
network)222 223.
• Switching charges are considered to be null, as an effect of the Data Act.
• For each of the application sizes, the following number of VMs and sizes are
considered. Pricing is based on AWS EC2 on-demand rates in eu-central
(Frankfurt) in USD (1 USD rounded to 0.9 for EUR for the conversion), and
dedicated instances (Prices of March 2026). The rationale for the decision of the
instance types is provided next.
o T-series such as t3 are burstable instances, which provide a baseline level
of CPU and can burst above when needed. They are recommended for
workloads with variable CPU usage such as small web servers or
microservices that mostly idle but have occasional spikes.
o M5-series are often used as the default instance type for applications that
need consistent CPU performance and that do not fit into compute or
memory-optimized type of applications. M5 instances are balanced CPU
and memory instances, designed for midsize databases and back-end
applications. M-instance types are at the crossroads of T (general purpose),
C (compute-optimized) and R (memory-optimized) instance types in terms
of compute and memory allocation.
Both T and M instance types are considered General Compute.
Table 68. Types of VMs considered for the cost estimations for small, mid-size and large applications
(Source: European Commission based on AWS Calculator, March 2026)
222 See for example this reported experiences https://european.cloud/2025/06/a-basic-look-at-pricing-of-european-cloud-
vendors/ , https://www.dataminded.com/resources/cloud-independence-testing-a-european-cloud-provider-against-the-
giants, https://www.fromeuropewithlove.eu/en/blog/european-cloud-providers-vs-aws-comparison-2026 223 BCG
234
Parameter Small Mid-Size Large Baseline metric
Instance Type t3.large m5.xlarge m5.8xlarge AWS EC2, Dedicated
instance, EBS gp3, eu-
central (Frankfurt)
VM Count 1–3 3–5 5-10 VMs per app
vCPU 2 4 32 cores
RAM 4 16 64 GB
Network
bandwidth
Up to 5 Up to 5 Up to 25 Gbps
Price 0.12 0.39 3.09 EUR/h
As anticipated above, across all, labour costs represent the largest component irrespective
of it being legacy-to-cloud or cloud-to-sovereign cloud, and in all phases. The calculation
of the labour effort stem from the few available sources and estimations provided in
confidence by experienced personnel of the public sector who have tackled complex
porting projects.
Table 69. Estimation of effort incurred in a porting, cloud-to-sovereign cloud, broken down in phases
(central values)
Porting phases Sovereign Cloud Small
application
(hours)
Mid-size
application
(hours)
Large
application
(hours)
Phase 1 – Feasibility assessment 40 290 470
Phase 2 – Strategy and planning 75 470 1,060
Phase 3 – Target environment set up 28 205 540
Phase 4 – Porting 173 1,496 3,860
Phase 5 – Testing and validation 64 640 1,200
Phase 6 - Cutover and Switching 16 128 410
Phase 7 – Post – porting and decommission 12 108 460
Total one-off effort (hours) 408 3,337 8,000
As mentioned above, the effort for legacy-to-cloud migration is 1.5x the effort shown
above.
The min-max values considered, and their explanation are as follows:
Table 70. Estimation of effort incurred in a migration, min-max values
Migration
phases
Sovereign
Cloud
Small
application
(hours) [min
value, max
value]
Mid-size
application
(hours) [min
value, max
value]
Large
application
(hours) [min
value, max
value]
Rationale
Phase 1 –
Feasibility
assessment
[26, 54] [180, 400] [300, 640] Given the
importance of all the
activities under this
235
Migration
phases
Sovereign
Cloud
Small
application
(hours) [min
value, max
value]
Mid-size
application
(hours) [min
value, max
value]
Large
application
(hours) [min
value, max
value]
Rationale
phase, experience
demonstrates that a
good understanding
of the application
(assets,
dependencies,
licenses) are equal
and their effort
varied.
Phase 2 –
Strategy and
planning
[50, 100] [332, 608] [760, 1360] While all values are
increased, the main
driver for this
change is the cloud
architecture design.
This is a critical
activity that, if not
performed properly,
can result in over-
costs in the
migration and
operation phase, as it
can cause over costs.
Phase 3 –
Target
environment
set up
[18, 38] [138, 272] [380, 700] VM provisioning
and configuration is
the main effort
intense activity in
this phase.
Configuring the
security policies or
the computation
resources in an
incorrect way can
result in cost
inefficiencies. A
common error in this
phase is to over-
provision capacity to
be able to respond to
potential load spikes.
Phase 4 –
Migration
[110, 236] [860, 2 132] [2 200, 5 520] The main drivers in
this phase are the
migration of the data
and the adaptation of
the application. The
success of these
activities largely
depends on the
previous activities
236
Migration
phases
Sovereign
Cloud
Small
application
(hours) [min
value, max
value]
Mid-size
application
(hours) [min
value, max
value]
Large
application
(hours) [min
value, max
value]
Rationale
such as the
dependency
assessment or the
cloud architecture
design and service
mapping. Moreover,
the skills of the team
and the wave plan
have also a large
effect in the effort
needed for the
migration of the
code.
Phase 5 –
Testing and
validation
[42,87] [440, 840] [700, 1 700] All testing activities
considered
(functional,
integration and
performance) are
driving the min –
max values in this
phase. The decision
of the coverage of
the tests plays an
important role. A
low coverage testing
will result in more
bugs during
production.
Correcting errors
during production
are much more
costly and can have
further implications
if an efficient CI/CD
pipeline has not been
well implemented.
Phase 6 -
Cutover and
Switching
[10, 22] [90, 166] [280, 540] The variation mainly
stems from the final
data migration and
the process of going
into production.
Phase 7 – Post
– migration
and
decommission
[8, 16] [68, 148] [280, 640] Under this phase, the
most important
activities that drive
the increment are
setting up the
observability
mechanisms, the
237
Migration
phases
Sovereign
Cloud
Small
application
(hours) [min
value, max
value]
Mid-size
application
(hours) [min
value, max
value]
Large
application
(hours) [min
value, max
value]
Rationale
decommissioning of
the infrastructure
and verifying that
everything was
correct, and the
training of the staff
that will be operating
the application.
Total one-off
effort (hours)
[264 - 553] [2 108 – 4 566] [4 900– 11 100]
The remaining costs of a migration / porting span infrastructure (VM), tooling licenses,
and the cost of parallel running both environments during transition. The table below
shows the values considered for the min-max approach.
Table 71. Estimation of VMs in the different migration phases (min, max, central values)
Phase Small Mid-sized Large
P3 - Target
Environment
Setup
VM provisioning:
ranges between 3 and
6 days [only working
hours in 3 weeks]
VM provisioning:
ranges between 6 and
9 days [only working
hours in 3 weeks]
VM provisioning:
ranges between ranges
between 9 and 12 days
[only working hours in
3 weeks]
P4 -
Migration
VM / server image
replication: ranges
between 3 and 6 days
[only working hours
in 3 weeks]
Parallel
infrastructure (dual-
run): ranges between
8 and 16 days [only
working hours in 8
weeks]
VM / server image
replication: ranges
between 6 and 9 days
[only working hours
in 3 weeks]
Parallel
infrastructure
(dual-run): ranges
between 16 and 24
days [only working
hours in 8 weeks]
VM / server image
replication: ranges
between 9 and 12 days
[only working hours in
3 weeks]
Parallel
infrastructure (dual-
run): ranges between
16 and 24 days [only
working hours in 8
weeks]
238
Phase Small Mid-sized Large
P6 - Cutover
and
Switching
Go live activity:
given the criticality of
this activity, no
ranges are considered.
Duration 2 weeks,
24/7
Go live activity:
given the criticality
of this activity, no
ranges are
considered. Duration
2 weeks, 24/7
Go live activity: given
the criticality of this
activity, no ranges are
considered. Duration 2
weeks, 24/7
P7 - Post-
Migration
Optimization: given
the criticality of this
activity, no ranges are
considered. Duration
2 weeks, 24/7
Optimization: given
the criticality of this
activity, no ranges
are considered.
Duration 4 weeks,
24/7
Optimization: given
the criticality of this
activity, no ranges are
considered. Duration 4
weeks, 24/7
One-off adjustment costs for the porting and migration of cloud services (legacy-to-
cloud, cloud-to-cloud) incurred by public essential entities listed under NIS2 Annex I
Based on the consideration above, namely:
• 6 400 entities, classified in high-intensive users, medium intensive users, and low
intensive users of cloud services, distributed as explain above.
• 1 200 IT services are considered, where 30% of the services are already cloudified
and would be subject to a cloud-to-cloud migration, 30% are on-premises and will
be cloudified, and 40% should remain on premise (e.g. legacy systems that are
being phased out, very small IT services).
• Three types of IT services depending on the size and complexity, small, mid-sized
and large, distributed as explained previously.
• Based on the discussions mentioned above, the distribution of sovereignty needs
resulting from the risk assessment across the public sector are Level 1: 70%, Level
2: 20%, Level 3: 9%, Level 4: 1%.
• The effort to migrate legacy-to-cloud is 1.5x bigger than porting cloud-to-cloud.
The total estimated costs of porting cloud – to – cloud for different sovereignty levels (1-
4, 2-4) is shown in the table below. For sensitivity purposes, a baseline for the price markup
of sovereign services over current cloud offerings has been set in +5% and a range from +
12% to -10% is considered224.
224 Rationale for the range is described in section 2.3.4. of the main document.
239
Table . Cloud-to-sovereign cloud porting costs (m EUR) of the IT systems of NIS 2 public essential
entities for different price markups and sovereign assurance levels (Sovereignty Assurance Level
(SAL) 1-4, SAL 2-4)
IT system SAL Max
(+12%)
Base
(+5%)
Min
(-10%)
SAL 1-4 288,423 287,629 285,158
SAL 2-4 86,527 86,289 85,547
As shown in the tables above, the price markup of sovereign solutions does not influence
significantly the cost of porting or migration as most of this cost comes from the human
effort to run that activity.The central EUR 86.3 bn figures should be regarded as a
maximalist estimate of the potential cost of porting applications to sovereign cloud services
on levels 2-4 across around 6 400 public entities. This estimate assumes that 30% of the
cloudified solutions of these entities, all those requiring sovereignty assurance levels 2 to
4, would be subject to porting because of the intervention. Moreover, this estimate
corresponds to a “big bang” implementation scenario, whereby all 6 400 entities would
undertake the porting activities in the first year. In practice, this scenario is unlikely to
reflect the actual phases of implementation, as the porting process would typically be
expected to occur progressively over a longer period, possibly extending up to five years.
Finally, this cost estimate cannot be fully attributed to the policy intervention itself. A
significant share of this porting would likely arise anyway as a result of natural
procurement decisions taken by public authorities over time, including routine renewal,
replacement or migration of cloud services. Treating this full amount as an incremental
cost of the intervention therefore overstates its direct impact.
Accordingly, under PO2-C, a more proportionate estimate has been considered. While
remaining conservative, as it continues to include 6 400 entities for potential porting costs,
it considers only a subset of critical applications for which porting to cloud services with
sovereign levels 3-4 could be directly accelerated by the policy intervention. This would
correspond to an anticipated range of 1% of applications in the minimum scenario (that
considers porting of all SAL4-demanding applications) up to 6% in the maximum scenario
(including also the porting of a good part of SAL3-demanding services), ported over 3
years. On this basis, the estimated cost under this option would amount to approximately
EUR 2.5bn in the minimum scenario and EUR 14.8 bn in the maximum scenario. These
figures are calculated as the sum of discounted values over a three-year period, following
a natural progressive migration approach, starting in 2030 and using 2025 as the baseline
year.
Under PO2-B, the measures are not mandatory and would thus be expected to apply to a
smaller number of public authorities. For this option, the relevant number of entities is
assumed to be around 1 600 public authorities, i.e. 25% of the ones considered under PO2-
C. As a result, the associated porting costs are assumed to decrease proportionally to 25%
of the PO2-C estimates, amounting to around EUR 620 m in the minimum scenario and
EUR 3.7 bn in the maximum scenario.
Recurring operating costs
240
Operating a cloud service requires a continuous monitoring and optimization. Under this
impact assessment and in line with industrial practices, operating an IT system on the cloud
involves also the release of new features.
Some of the main activities under the operation phase include release engineering,
infrastructure provisioning, and observability, with each one generating their own
expenses, even if they all are highly interdependent.
Release pipelines offer the earliest opportunity for cost control. Every execution consumes
ephemeral compute and produces artifact storage, which means that improvements to build
reliability and release validation compound in quick manner. Organizations that invest in
release discipline by reducing failure rates, tightening validation gates, and minimizing
rollbacks, consistently achieve lower pipeline costs than those running at higher
deployment velocity without equivalent quality controls. The returns are direct and
measurable.
Infrastructure provisioning, typically the largest cost category, responds well to deliberate
management. Virtual machines can be right sized to match actual utilization patterns,
eliminating the idle compute overhead that accumulates when capacity is provisioned
conservatively. Monitoring agents while often overlooked, are similarly important to
rationalization. They allow to reduce the overhead per-host that otherwise would scale
independently of the application workload. regardless of application load. Organizations
that treat infrastructure sizing as an ongoing practice rather than a one time decision tend
to maintain materially lower baseline costs.
The estimated effort in hours for operating a cloud service, considering also the addition
of new features and continuous optimization, per type of application is shown in the next
table. This estimation is based on experiences from industry experts.
Table . Estimated effort to operate and maintain a cloud service (hours)
Migration phases
Sovereign Cloud Small app Mid-size app Large app
Phase 8 - Operation 968 4.840 6.600
12.5 Limitations of the analysis
The cost model presented in this analysis carries several important limitations that must be
acknowledged when interpreting its outputs. First, labour rates are modelled as a static
band (EUR 75/hr) that does not account for overtime, or additional consultant charges.
Second, VM configurations are drawn from tier averages rather than empirically measured
utilization profiles, which is the proper way of estimating the sizes, introducing upward
bias notably in the infrastructure costs assumed for Phase 3 and Phase 4 through implicit
over-provisioning, reflected in the VMs chosen. Third, all infrastructure pricing uses AWS
public cloud on-demand rates as a baseline. On-demand instances are on the costly end.
Fourth, the two-month parallel infrastructure (dual-run) window in Phase 4 is fixed to
several days a week for a longer period of time. This is however one of the riskiest phases
that would need to be estimated proportionately and accurately on a case-by-case basis.
However, while fixed to several days, these have been estimated considering industry’s
241
best practices. Finally, the cost ranges are based on benchmarks that present total costs of
migration, aggregating labour, infrastructure, training, and tooling licensing. Again, this is
largely dependent on the as-is situation and the to-be situation.
242
ANNEX 13. COMPARATIVE ANALYSIS OF SELECTED CLOUD
SERVICES
European cloud providers offer far narrower service catalogues than US hyperscalers:
AWS alone spans over 200 distinct services. Yet this gap is less consequential than it
appears. The latest Eurostat data for 2025 shows that the services EU enterprises actually
rely on most are email, office software, file storage, database hosting, and compute
power225, which are primarily IaaS and SaaS. Compute power, hosting, file storage and
database are the basis for core IaaS workloads: virtual machines, managed databases,
storage, and nowadays also PaaS, where managed Kubernetes (orchestration) are robustly
available across European providers. Email and Office software are SaaS. Given the
heterogeneity of SaaS, which is much larger than in IaaS and PaaS, and where each
application on the web could be called, even if in simplistic terms, software as a service,
renders the comparison between US and EU providers, albeit a few exceptions, much more
difficult.
However, the real gaps are narrower and more specific that it can be initially thought.
These include mainly native AI/ML platforms, serverless, and integrated analytics
pipelines as they remain areas where European providers have not yet reached hyperscaler
maturity but are working towards it. For the standard private and public sector
workloads constituting nowadays the majority of cloud adoption needs, the European
providers represent already a fully viable, equivalent in performance, functionalities
and quality.
The tables below show that major European cloud providers, for the purpose of this
analysis, IONOS, OVHcloud, and STACKIT, already offer services broadly equivalent to
those of AWS, Microsoft Azure, and Google Cloud Platform across the core infrastructure
categories that are most widely used: compute, storage, network, and managed container
orchestration. The comparison has been extended to office automation suites, one of the
most widely adopted SaaS categories in the public sector, where credible European and
open-source alternatives are also available.
OVHcloud's226 portfolio spans across compute, storage, networking, container
orchestration, databases, analytics, and AI/ML services. IONOS227 mainly covers storage,
compute, managed container orchestration, databases, and observability. STACKIT's228
catalogue includes compute, databases, observability and logging, and security services.
Taken together, these providers cover the functional surface that the majority of the private
and public sector workloads actually require.
The comparisons shown below demonstrate that EU providers already offer the most used
services and with a similar level of functionality. The first table aims to provide a high-
level overview of the main cloud services used in the Union and their coverage by EU
providers compared to non-EU incumbents. In the public sector, interviews have shown
225 See Eurostat 226 OVH 227 Ionos 228 Stackit
243
that most public administrations use an average of twenty services, instead of the whole
range offered by hyperscalers.
Table 74. High level overview of cloud services and their coverage by EU providers
Service Type Layer Coverage by
EU providers
Virtual Machines / Compute IaaS Full
Object Storage (S3 compatibility) IaaS Full
Block Storage IaaS Full
Managed Load Balancer IaaS/Network Full
VPC / Private Networking IaaS/Network Full
CDN IaaS/Network Partial
Managed containers / containers
(Container as a Service - CaaS)
PaaS Full
Managed PostgreSQL/MySQL PaaS Full
Managed NoSQL / MongoDB PaaS Full
Managed Kafka / Streaming PaaS Full
Serverless Functions (FaaS) PaaS Partial
Serverless Containers PaaS Partial
AI/ML Training Platforms PaaS Partial
Managed AI Inference PaaS Partial
ERP SaaS Full
Office automation / Collaboration /
Messaging
SaaS Full
The second set of tables provide a much more granular analysis comparing per feature five
of the widest adopted services: compute, storage, network, managed containers and office
automation tools.
A sample of EU providers have been selected to demonstrate that, also when analysing per
functionality, European providers’ offerings are equivalent to those provided by non-EU
incumbents.
This analysis is based on desktop research, more specifically, through a manual review of
each provider's official user guides, technical documentation, and online manuals. Where
a service appeared to fulfil the same functionality as its counterpart, but the documentation
did not allow for an unambiguous confirmation, the mapping has been marked as partial
rather than full. This conservative approach reflects an inherent limitation of the
methodology: the analysis is theoretical in nature, derived from published specifications
rather than from hands-on testing or validated benchmarking.
Table 75. Comparative table compute (selected sample of providers)
Functionality AWS EC2 Google
Compute
Azure
Compute
OVH
Compute
IONOS
Compute
StackIt
Compute
General purpose
VMs ✓ ✓ ✓ ✓ ✓ ✓
244
Functionality AWS EC2 Google
Compute
Azure
Compute
OVH
Compute
IONOS
Compute
StackIt
Compute
Compute-
optimized ✓ ✓ ✓ ✓ ✓ ~
Memory-
optimized ✓ ✓ ✓ ✓ ✓ ~
Bare metal
instances ✓ ✓ ✓ ✓ ✓
(Dedicated
servers)
✕
Custom vCPU /
RAM ratio
~ ✓ ~ ~ ~ (flexible
within tiers)
~ (fixed
ratios)
Reserved
instances ✓ ✓ ✓ ✓ ✓ ✕
Confidential
compute ✓ ✓ ✓ ~ ✕ ✓
Legend: ✓ Supported ~ Partial / limited ✕ Not available
Table 76. Comparative table Storage (selected sample of providers)
Functionality AWS Google Azure OVH IONOS StackIt
Block storage
(persistent disks) ✓ ✓ ✓ ✓ ✓ ✓
Object storage
integration ✓ ✓ ✓ ✓ ✓ ✓
Shared / NFS file
storage ✓ ✓ ✓ ~ ~ ~
Snapshot & image
management ✓ ✓ ✓ ✓ ✓ ✓
Multi-zone disk
replication ✓ ✓ ✓ ~ ~ ~
Encrypted storage
at rest ✓ ✓ ✓ ✓ ✓ ✓
Legend: ✓ Supported ~ Partial / limited ✕ Not available
Table 77. Comparative table Network (selected sample of providers)
Functionality AWS Google Azure OVH IONOS StackIt
Virtual private
network (VPC) ✓ ✓ ✓ ✓ ✓ ✓
IPv6 support ✓ ✓ ✓ ✓ ✓ ✓
Load balancer
(managed) ✓ ✓ ✓ ✓ ✓ ✓
CDN ✓ ✓ ~ ✓ ✓ ✓
Private
interconnect /
VPN
✓ ✓ ✓ ✓ ✓ ~
245
Functionality AWS Google Azure OVH IONOS StackIt
DNS service
(managed) ✓ ✓ ✓ ✓ ✓ ✓
DDoS protection
(built-in) ✓ ✓ ✓ ✓ ✓ ✓
Floating / elastic
IPs ✓ ✓ ✓ ✓ ✓ ✓
Legend: ✓ Supported ~ Partial / limited ✕ Not available
Table 78. Comparative table Containers (selected sample of providers)
Functionality AWS
(ECS/EKS/Fargate)
(GKE/Cloud
Run)
Azure
(AKS /
ACI)
OVH
(MKS)
IONOS
(Managed
Kubernetes)
StackIt
(SKE)
Managed
Kubernetes
service
✓ ✓ ✓ ✓ ✓ ✓
Serverless /
managed
container service
(non-K8s)
✓ ✓ ✓ ✕ ✕ ✕
CNCF certified
Kubernetes ✓ ✓ ✓ ✓ ✓ ✓
Vanilla /
upstream
Kubernetes
~ ✓ ~ ✓ ✓ ✓
Latest K8s
version
availability speed
~ ✓ ~ ✓ ~ ~
Extended version
support (LTS) ✓ ✓ ✓ ~ ✕ ✕
Managed / free
control plane
~ ✓ ✓ ✓ ✓ ✓
High-availability
control plane ✓ ✓ ✓ ✓ ✓ ✓
Multi-AZ control
plane ✓ ✓ ✓ ✓ ~ ~
Dedicated control
plane (isolated) ✓ ✓ ✓ ✓ ~ ~
Automatic
control plane
upgrades
✓ ✓ ✓ ✓ ✓ ✓
Private (API
server not public) ✓ ✓ ✓ ✓ ✓ ✓
Cluster self-repair
/ auto-healing ✓ ✓ ✓ ✓ ✓ ✓
Multiple clusters
per project ✓ ✓ ✓ ✓ ✓ ✓
246
Functionality AWS
(ECS/EKS/Fargate)
(GKE/Cloud
Run)
Azure
(AKS /
ACI)
OVH
(MKS)
IONOS
(Managed
Kubernetes)
StackIt
(SKE)
Multiple node
pools per cluster ✓ ✓ ✓ ✓ ✓ ✓
Custom node pool
sizing
(CPU/RAM)
✓ ✓ ✓ ✓ ✓ ✓
GPU node pools ✓ ✓ ✓ ✓ ✓ ✕
Spot /
preemptible node
pools
✓ ✓ ✓ ✓ ~ ✕
Node auto-repair ✓ ✓ ✓ ✓ ✓ ✓
Automatic node
OS upgrades ✓ ✓ ✓ ✓ ✓ ✓
Bare metal
worker nodes ✓ ✓ ✓ ✓ ✕ ✕
Cluster / node
autoscaler ✓ ✓ ✓ ✓ ✓ ✓
Horizontal Pod
Autoscaler (HPA) ✓ ✓ ✓ ✓ ✓ ✓
Vertical Pod
Autoscaler (VPA) ✓ ✓ ✓ ~ ~ ~
Event-driven
autoscaling
(KEDA)
✓ ✓ ✓ ~ ~ ~
Autopilot /
serverless node
provisioning
~ ✓ ~ ✕ ✕ ✕
Scale-to-zero /
scheduled
shutdown
✓ ✓ ✓ ~ ~ ✓
VPC / private
network
integration
✓ ✓ ✓ ✓ ✓ ✓
Managed load
balancer (ingress) ✓ ✓ ✓ ✓ ✓ ✓
Network policies
(CNI) ✓ ✓ ✓ ✓ ✓ ✓
Cilium CNI
support ✓ ✓ ✓ ✓ ✕ ✕
Service mesh
integration
(Istio/Linkerd)
✓ ✓ ✓ ~ ~ ~
IPv6 cluster
support ✓ ✓ ✓ ✓ ~ ~
247
Functionality AWS
(ECS/EKS/Fargate)
(GKE/Cloud
Run)
Azure
(AKS /
ACI)
OVH
(MKS)
IONOS
(Managed
Kubernetes)
StackIt
(SKE)
Ingress / Gateway
API support ✓ ✓ ✓ ✓ ~ ~
Private nodes (no
public IPs) ✓ ✓ ✓ ✓ ✓ ✓
Persistent
volumes (block
storage)
✓ ✓ ✓ ✓ ✓ ✓
ReadWriteMany
(RWX) volumes ✓ ✓ ✓ ~ ~ ~
Managed NFS /
file storage ✓ ✓ ✓ ✓ ~ ~
Dynamic volume
provisioning ✓ ✓ ✓ ✓ ✓ ✓
Encrypted storage
at rest ✓ ✓ ✓ ✓ ✓ ✓
Volume
snapshots ✓ ✓ ✓ ~ ~ ~
Managed private
container registry ✓ ✓ ✓ ✓ ✓ ✓
OCI & Helm
artifact support ✓ ✓ ✓ ✓ ✓ ✓
Image
vulnerability
scanning
✓ ✓ ✓ ✓ ~ ~
Geo-replication
of registry ✓ ✓ ✓ ~ ✕ ✕
Registry
integrated with
K8s cluster
✓ ✓ ✓ ✓ ✓ ✓
RBAC (role-
based access
control)
✓ ✓ ✓ ✓ ✓ ✓
OIDC / SSO
authentication ✓ ✓ ✓ ✓ ~ ~
Pod Security
Standards (PSS) ✓ ✓ ✓ ✓ ✓ ✓
Secrets
management
(KMS integrated)
✓ ✓ ✓ ~ ~ ~
Network policy
enforcement ✓ ✓ ✓ ✓ ✓ ✓
Image signing /
admission control ✓ ✓ ✓ ~ ~ ~
248
Functionality AWS
(ECS/EKS/Fargate)
(GKE/Cloud
Run)
Azure
(AKS /
ACI)
OVH
(MKS)
IONOS
(Managed
Kubernetes)
StackIt
(SKE)
Confidential
Kubernetes / TEE
nodes
✓ ✓ ✓ ~ ✕ ✓
Audit logging ✓ ✓ ✓ ✓ ~ ~
Built-in
monitoring
(metrics)
✓ ✓ ✓ ~ ~ ~
Built-in logging ✓ ✓ ✓ ~ ~ ~
Prometheus /
Grafana
integration
✓ ✓ ✓ ✓ ✓ ✓
Alerting &
dashboards ✓ ✓ ✓ ~ ~ ~
kubectl / API
access ✓ ✓ ✓ ✓ ✓ ✓
Terraform / IaC
support ✓ ✓ ✓ ✓ ✓ ✓
GitOps / ArgoCD
support ✓ ✓ ✓ ✓ ~ ~
CLI tooling ✓ ✓ ✓ ✓ ✓ ✓
Native CI/CD
pipeline
integration
✓ ✓ ✓ ~ ~ ~
PaaS / Cloud
Foundry layer ✕ ✕ ✕ ✕ ✕ ✓
Multi-cluster
management
console
✓ ✓ ✓ ✓ ✓ ✓
Legend: ✓ Supported ~ Partial / limited ✕ Not available
Table 79. Comparative table Office automation tools (selected marek of providers)
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
DOCUMENT CREATION & WORD PROCESSING
Word
processor ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
249
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
Advanced
text
formatting ✓ ~ ~ ~ ~ ~ ✓ ~
Desktop
application
(offline) ✓ ✕ ✕ ✓ ✕ ✕ ✓ ✕
Browser-
based
editing ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
Mobile
editing (iOS
/ Android) ✓ ✓ ✓ ~ ~ ✓ ✕ ~
Track
changes /
review
mode
✓ ✓ ✓ ✓ ✕ ✓ ✓ ~
Templates
library ✓ ✓ ~ ✓ ~ ~ ✓ ~
Mail merge ✓ ~ ~ ✓ ✕ ✕ ✓ ✕
Macro
support
(scripting) ✓ ~ ✕ ~ ✕ ✕ ✓ ✕
Digital
signatures /
e-signing ✓ ✓ ~ ~ ✕ ~ ✕ ~
SPREADSHEETS
Spreadsheet
editor ✓ ✓ ✓ ✓ ✓ ✓ ✓ ✓
Advanced
formulas &
functions ✓ ~ ~ ~ ~ ~ ✓ ~
Pivot tables ✓ ✓ ~ ✓ ✕ ~ ✓ ✕
Macros /
VBA
support ✓ ✕ ✕ ~ ✕ ✕ ✓ ✕
Charts &
data
visualizatio
n
✓ ✓ ✓ ✓ ~ ~ ✓ ~
250
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
Data
validation
rules ✓ ✓ ~ ✓ ~ ~ ✓ ~
Conditional
formatting ✓ ✓ ✓ ✓ ~ ~ ✓ ~
Large
dataset
performanc
e
✓ ~ ~ ~ ✕ ~ ✓ ✕
Import/exp
ort Excel
(.xlsx) ✓ ✓ ✓ ✓ ~ ✓ ✓ ~
PRESENTATIONS
Presentatio
n editor ✓ ✓ ✓ ✓ ✓ ✓ ✓ ~
Slide
templates &
themes ✓ ✓ ~ ✓ ~ ~ ✓ ✕
Animations
&
transitions ✓ ~ ~ ✓ ~ ~ ✓ ✕
Presenter
view ✓ ✓ ✓ ✓ ✕ ~ ✓ ✕
Live
audience
interaction ✓ ~ ✕ ✕ ✕ ✕ ✕ ✕
Export to
PDF / video ✓ ✓ ✓ ✓ ~ ✓ ✓ ~
Embed
media
(audio/vide
o)
✓ ~ ~ ✓ ✕ ~ ✓ ✕
PPTX
import/expo
rt fidelity ✓ ~ ~ ✓ ~ ✓ ~ ✕
REAL-TIME COLLABORATION
Simultaneo
us co-
editing ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
251
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
Live cursor
/ presence
display ✓ ✓ ✓ ✓ ✓ ✓ ✕ ~
In-
document
comments
& threads
✓ ✓ ✓ ✓ ✓ ✓ ✕ ~
Document
version
history ✓ ✓ ✓ ✓ ✓ ✓ ✕ ~
Suggesting
/ review
mode ✓ ✓ ✓ ✓ ~ ✓ ✓ ~
Paragraph-
locking
(conflict
free)
~ ✕ ✕ ✓ ✓ ✕ ✕ ✕
Link
sharing (no-
login
editing)
✓ ✓ ~ ✓ ✓ ✓ ✕ ~
Guest /
external
collaborator
s
✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
EMAIL & COMMUNICATION
Email client
/ webmail ✓ ✓ ✕ ~ ✕ ✓ ✕ ✓
Calendar ✓ ✓ ✓ ~ ✕ ✓ ✕ ✓
Contacts /
address
book ✓ ✓ ✓ ~ ✕ ✓ ✕ ✓
Tasks / to-
do
managemen
t
✓ ✓ ✓ ~ ✕ ✓ ✕ ✓
Team chat /
messaging ✓ ✓ ✓ ~ ✓ ~ ✕ ~
Video
conferencin
g ✓ ✓ ✓ ✕ ✓ ✓ ✕ ~
252
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
Webinar /
large
meeting
hosting
✓ ✓ ~ ✕ ✕ ~ ✕ ✕
Screen
sharing ✓ ✓ ✓ ✕ ✓ ✓ ✕ ~
FILE STORAGE & MANAGEMENT
Cloud file
storage ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
Desktop
sync client ✓ ✓ ✓ ✓ ✕ ✓ ✕ ~
Offline file
access ✓ ~ ✓ ✓ ✕ ~ ✓ ~
External
file sharing
links ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
File
versioning
& restore ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
Folder-level
permissions ✓ ✓ ✓ ✓ ✓ ✓ ✕ ✓
Self-hosted
/ on-
premise ✕ ✕ ✓ ✓ ✓ ✕ ✓ ✓
Included
cloud
storage
(entry plan)
1TB 15GB Self-
hosted 5GB 1GB 15GB Local
Self-
hosted
AI & PRODUCTIVITY TOOLS
Integrated
AI writing
assistant ✓ ✓ ~ ✓ ✕ ✕ ✕ ✕
AI email
drafting /
summarizat
ion
✓ ✓ ✕ ✕ ✕ ✕ ✕ ✕
AI meeting
notes /
transcriptio
n
✓ ✓ ✕ ✕ ✕ ✕ ✕ ✕
253
Functionali
ty
Microsof
t 365
Workspa
ce
Nextclo
ud +
Collabo
ra
OnlyOffi
ce
CryptP
ad
Infomani
ak kSuite
LibreOff
ice
Open-
Xchan
ge
US
Propriet
ary
US
Proprieta
ry
EU
Open
Source
EU
Open
Source
EU
Open
Source
EU
Proprieta
ry
EU Open
Source
EU
Open
Source
AI
spreadsheet
/ formula
help
✓ ✓ ✕ ~ ✕ ✕ ✕ ✕
No-code
app builder ~ ✓ ✕ ✕ ✕ ✕ ✕ ✕
Workflow
automation ✓ ✓ ~ ~ ✕ ~ ✕ ~
Forms /
survey tool ✓ ✓ ✓ ✓ ✓ ~ ✕ ~
Whiteboard
/ drawing
tool ✓ ✓ ✓ ✓ ✓ ~ ✕ ~
Notes / wiki
tool ✓ ✓ ✓ ~ ~ ~ ✕ ~
Legend: ✓ Supported ~ Partial / limited ✕ Not available
EN EN
EUROPEAN COMMISSION
Brussels, 3.6.2026
SWD(2026) 503 final
COMMISSION STAFF WORKING DOCUMENT
EXECUTIVE SUMMARY OF THE IMPACT ASSESSMENT REPORT
Accompanying the document
Proposal for a
REGULATION OF THE EUROPEAN PARLIAMENT AND OF THE COUNCIL
establishing a framework of measures for strengthening Europe's cloud and AI
ecosystem (Cloud and AI Development Act)
{COM(2026) 502 final} - {SEC(2026) 502 final} - {SWD(2026) 502 final}
1
Identification of the problem and EU-level dimension
Cloud computing and AI have become driving forces reshaping industry, public services, and
daily life. They are powered by computing capacity located in data centres. The availability of
computing capacity in the EU is limited and geographically concentrated, negatively affecting
competitiveness. Moreover, the EU is dependent on cloud and AI computing services
supplied by non-European providers. This can result in risks of control over data and
operational autonomy.
The deployment of computing capacity in the EU is currently characterised by national
policies, leading to fragmentation. The dependence on non-European providers of cloud and
AI computing services has the same root causes across the EU, affecting the public and
private sector in all Member States.
Aim of the initiative
The Cloud and AI Development Act (CADA) aims to ensure the functioning of the internal
market for cloud and AI computing services and to secure the conditions necessary for the
Union’s competitiveness and strategic autonomy. It sets out to:
• Increase computing capacity deployed in the EU through innovative and sustainable
technologies
• Ensure attractive conditions for such deployment
• Reducing the overall reliance on non-sovereign cloud and AI computing services
• Contribute to the protection of public order by enhancing the resilience of supply of cloud
and AI computing services, in particular in the public sector
Value added at EU-level
To boost computing capacity, an EU-level action provides a common approach enabling
coherent planning and deployment of capacity in a geographically balanced way, while
avoiding a race to the bottom and making the regulatory environment clearer for relevant
businesses. To reduce the dependence on services supplied by non-European providers, EU-
level action delivers benefits beyond what Member States can achieve individually by pooling
resources and creating economies of scale.
It also strengthens the Union's strategic autonomy of digital infrastructures, safeguarding
digital sovereignty against risks related to the use of non-sovereign services and external
dependencies. By fostering a robust internal market for cloud and AI computing services, the
EU can secure vital digital assets and boost its global competitiveness. CADA supports the
development of a resilient and secure ecosystem conducive to innovation via EU-level R&D
funding challenges targeting the development of highly innovative Cloud and AI enabling
technologies, offering a stable environment to attract investments and to nurture growth.
Ultimately, it can enhance the collective bargaining power of Member States in the
international arena.
Options evaluated & Preferred option
The document evaluates six policy options, grouped into two categories: three addressing the
limited and geographically concentrated availability of computing capacity in the EU (PO1)
and three tackling the dependence on cloud and AI computing services supplied by non-EU
providers (PO2).
PO1-A enhances the existing collaborative framework between Member States, EU
institutions and relevant industries to support data centre expansion. PO1-B provides rules and
2
financial support for faster data centre deployment, implemented nationally. PO1-C shifts this
implementation to the EU-level.
PO2-A contains supporting measures to increase transparency and visibility of sovereign
cloud and AI computing services. PO2-B creates a voluntary framework for advancing such
services and proposes the federation of resources among the public sector to boost efficiency.
PO2-C establishes an EU-coordinated procurement and mandatory support framework for
strategic autonomy and sovereign services. In this later aspect, it provides measures for the
public sector as well as for private sector essential entities operating under the sectors listed in
Annex I of NIS2.
The preferred option is the combination of PO1-B, PO2-C and two individual policy measures
of PO1-C, related to the funding for the development and deployment of novel technologies
for sustainability along the cloud value chain (PM8, PM9). This package achieves the best
balance of costs and benefits, while providing a good balance in terms of social and
environmental impacts and respecting subsidiarity and proportionality principles.
Stakeholders’ support
Data centre operators and cloud service providers have voiced support for PO1-B and PM8
and PM9, aiming to facilitate quicker deployment of data centres by reducing red tape and
encouraging investments, particularly those that promote sustainable technologies. They
welcome this package as a streamlining of administrative processes to deploy data centres.
Cloud and AI computing service providers have expressed support for PO2-C, notably the
joint public procurement mechanisms it puts in place; the marketplace, which allows smaller
providers to scale and be more visible towards incumbents; and the audit mechanism for
sovereign services, which would end sovereign-washing. They welcome these measures as
they enhance transparency and visibility of service offers, including from smaller providers,
and they result in opportunities to distinguish their services. Non-EU service providers have
repeatedly stressed their interest in retaining an open European market, including for public
procurement.
Public sector bodies support PO2-C, notably the public sector federation of cloud and AI
computing resources as this would increase utilisation rates of idle resources. They also
favour the coordinated approach to public procurement and the clarity it provides on how to
safeguard data confidentiality and ensure operational autonomy for highly critical use cases.
Benefits and costs of the preferred option
The preferred policy package (PO1-B, PO2-C, PM8 and PM9) is expected to have benefits in
the short and medium term that outweigh the generated costs over the assessment period for
different stakeholders. Data centre operators would benefit from streamlined processes, fast-
track permitting and administrative simplification, while cloud and AI computing service
providers would benefit from cost savings stemming mainly from the ability to get a single
audit for sovereign services, valid throughout the EU. This is relevant for the public
procurement of services for highly critical use cases across the whole EU. National public
authorities are also expected to benefit from cost savings, mainly through increased use of
open source, joint public procurement and the federation of capacity. This combination of
options also entails implementation and transition costs. There are mainly linked to adapting
systems and procedures, familiarising stakeholders with the new requirements and ensuring
compliance. Most of these costs are one-off or transitional in nature, while several of the
expected benefits are recurring over time. This combination of options is expected to deliver a
3
positive balance of costs and benefits, while achieving the objectives in a proportionate and
cost-effective manner.
Impact on SMEs and competitiveness
The impacts on SMEs and competitiveness are predominantly positive, as the comprehensive
EU-level intervention aims at enhancing the cloud and AI computing landscape. The preferred
package is designed to facilitate access to cloud and AI computing resources for SMEs
looking to innovate or expand their digital capabilities. By fostering a more unified and open
market, EU-based SMEs can leverage a harmonised regulatory environment which simplifies
business processes and enhances market opportunities.
For competitiveness, the intervention aims to boost the EU’s internal market by reducing
dependency on non-European service providers. This shift creates opportunities for local
service providers and SMEs, which they can leverage to compete globally.
Other significant impacts
If implemented, CADA is expected to lead to both environmental and social impacts related
to increased data centre capacity.
While expanding data centre capacity results in increased resource use, key environmental
benefits include the use and development of novel energy-efficient technologies in data
centres. By supporting such advancements, CADA is expected to help the EU alleviate the
carbon footprint of its data centre industry. This will contribute to the EU's broader climate
goals and commitments, aligning digital infrastructure expansion with sustainability
principles and fostering environmentally responsible growth in the tech sector.
Socially, CADA is expected to support innovation and create employment opportunities in the
tech sector, leading to high value job creation and skills development. Additionally, sovereign
cloud and AI services are expected to enhance resilience and security for public sector
operations, protecting sensitive data and ensuring service reliability for citizens.
Follow up
The Commission should carry out an evaluation five years after the adoption date to assess to
which extent the objectives of the initiative have been reached.