Brussels, 3 July 2025
Statement by the European AI Forum on the:
Consultation regarding the EU Cloud and AI
Development Act
The EU is at a critical juncture in securing its digital and industrial future. The Commission’s proposal for a Cloud and AI Development Act therefore comes at a pivotal time, as global computing capacity has become a major bottleneck for AI innovation, and digital sovereignty has become essential for competitiveness and achieving strategic ambitions.
Since the urgency to scale Europe’s AI capabilities is clear, it is essential to ensure that these efforts rest on strong, enduring foundations. Investment in AI infrastructure must go hand in hand with building high-quality, context-rich data ecosystems that reflect real-world services, users, and legal frameworks. Without these foundations, there is a risk that AI will amplify inefficiencies and uncertainty rather than deliver meaningful, trustworthy innovation.
Despite the solid progress made through EuroHPC and other national initiatives, the total EU-HPC power is only around three exaflops. This is less than half of the capacity of the US and well behind China’s rapid expansion.1 With that in mind, the Draghi Report also warns that Europe’s innovation engine will stall without a sharp increase in capacity. This assessment is underlined by the latest Stanford AI Index 2025: in 2024, the US produced 40 significant AI models, compared to 15 produced by China and only 3 produced by Europe. Globally, the capacity for training doubles every five months. Unless Europe scales up at the same pace, it risks falling further behind.2 Thus, computing capacity has become a form of strategic infrastructure, on a par with energy or semiconductors. Dependence on non-EU cloud providers jeopardises the resilience of critical services and the competitiveness of European (AI) companies.
1 The Next Platform (2025). Peeling The Covers Off Germany’s Exascale “Jupiter” Supercomputer.
https://www.nextplatform.com/2025/06/11/peeling-the-covers-off-germanys-exascale-jupiter-supercomputer/ 2 Stanford University (2025). The 2025 AI Index Report. https://hai.stanford.edu/ai-index/2025-ai-index-report
In conclusion, the EU remains structurally underpowered compared to the world's leading AI nations. However, time is running out to close this gap, which requires coordinated, ambitious and swift action. The European AI Forum therefore strongly supports the Commission’s objective of establishing a sovereign, innovation-enabling computing infrastructure. While we recognise the EU’s efforts to establish a competitive and trustworthy AI ecosystem, we also emphasise the importance of complementing regulatory ambition with industrial policy tools to ensure that European innovators have practical access to computing power, energy, and markets. We believe that it will be essential to form a robust, forward-looking Act - one that will unlock scale, equity and strategic independence for the next generation of AI companies in Europe - rather than merely administering scarcity.
Policy Recommendations:
1. Increase (private) Investments in AI Compute Infrastructure
Europe’s first priority is still to anchor strategic, sovereign-grade compute infrastructure. The EIB’s landmark report pegs the funding gap for competitive exascale infrastructure at roughly €500-700 billion per investment cycle - well beyond the reach of any single Member State. Thus, this calls for a public-value investment approach: pool Digital-Europe, national and regional envelopes, top them up with a dedicated “EU-AI 2035” bond line, and invite industry in on a PPP basis. Further incentives, such as targeted tax credits for AI infrastructure investments (harmonised across Member States) and streamlined regulations, will reduce costs and uncertainty, thereby encouraging greater investment participation.
The second task must be to pull market demand forward, especially from SMEs. Dedicated capacity and favourable terms (e.g. compute vouchers or special access quotas) should be reserved for start-ups and SMEs so that young innovators can utilise state-of-the-art facilities from day one. Furthermore, the EU and national authorities should also leverage pooled procurement and 'European preference' policies in public tenders to increase demand for Europe-based chips, data centres, and cloud services, thereby improving the business case for private investors.
Finally, the supply side needs room (and incentives) to behave commercially. Many public centres are still bound by rules that cap commercial revenues; loosening those limits would let them build mixed public-private business models. On the finance flank, the 2024 Solvency II review cut the capital charge on long-term equity in regulated
infrastructure to 22 percent, opening a deep pool of insurer equity for data-centre funds.3 Complementary reforms in securitisation, for example, extending STS treatment to GPU lease receivables, would let banks refinance mezzanine debt and recycle capital for the next build-outs.4
2. Harmonise EU Legal and Tax Regulations for REITs to Boost Data Centre Investment
Data centres are essential strategic infrastructure for AI development and data processing. Yet, investment in this sector is held back by fragmented legal and tax regimes across the EU. Currently, there is no harmonised framework for Real Estate Investment Trusts (REITs) that invest in data centres, even though these assets are treated as real estate in the same way as hotels or logistics facilities. Each Member State maintains its own legal and tax rules for REITs, creating barriers to cross-border investment and undermining the EU’s capacity to scale critical digital infrastructure.
This fragmentation discourages capital allocation by introducing complexity and uncertainty for investors. Divergent corporate tax exemptions, withholding taxes on dividends, and capital gains regimes create an uneven playing field that deters long-term investment in high-growth sectors like data centres. Inconsistent listing requirements on EU stock exchanges - covering disclosure obligations, minimum capital thresholds and governance standards - further limit REITs’ ability to raise funds across borders and constrain their scalability. As a result, Europe is leaving substantial private investment potential untapped at precisely the moment it needs to expand its AI and cloud capacity.
We therefore propose that the European Commission develop a single, EU-wide regulatory framework for REITs, establishing clear rules for their creation, operation and governance. This framework should standardise eligibility criteria, such as a minimum allocation to real estate assets like data centres, require the distribution of a fixed percentage of profits as dividends to improve investor returns, and introduce a streamlined approval process for cross-border operations. A harmonised EU tax regime is also essential to encourage investment in strategic infrastructure, offering consistent corporate tax exemptions, standardised withholding tax rates, and targeted incentives such as accelerated depreciation or credits for sustainable, energy-efficient data
3 DWS (2024). Solvency II Review and Long-Term Equity Investments: A New Era for Private Equity and Infrastructure Investments?. https://www.dws.com/en-US/AssetDownload/Index?assetguid=ad3a641b-a256-4d93-a119-0606c1297050 4 Financial Times (2025). How to fix Europe’s securitisation market.
https://www.ft.com/content/f01d18c6-16dd-41be-9b89-4465f01a47d6
centres. In addition, the EU should work with national stock exchanges and ESMA to create uniform listing requirements, ensuring consistent disclosure standards, governance rules and minimum capital thresholds to strengthen investor confidence and facilitate cross-border fundraising.
Finally, targeted policy support will be critical to unlocking the potential of REITs for European data centre development. This includes exploring EU grants or subsidies for projects aligned with the Digital Decade goals, fostering public-private partnerships to reduce investment risk in underserved regions, and prioritising green data centres that advance the EU’s climate neutrality objectives. By harmonising legal and tax rules and providing a clear, predictable investment environment, the EU can channel significant private capital into data centre infrastructure, strengthen its digital sovereignty and ensure that European innovators have the computing resources they need to compete globally.
3. Simplify EU Rules to Accelerate AI Infrastructure and Innovation
Investment in Europe’s AI, cloud, and compute infrastructure is significantly hindered by a fragmented and complex regulatory environment. Overlapping compliance regimes, slow permitting procedures and rigid state aid rules create friction at every stage of the innovation process, delaying infrastructure projects, increasing costs and deterring private capital. For start-ups and infrastructure providers equally, this means spending more time navigating bureaucracy and less time building, ultimately resulting in lower competitiveness. Even before projects begin, Europe’s reputation for heavy regulation can discourage ambition. Founders and investors often avoid high-impact AI ventures or choose to scale them up outside the EU, anticipating lengthy approval processes and legal uncertainty. This risk-averse climate undermines the Union’s ability to transform its research excellence into globally competitive AI solutions.
Europe’s public procurement sector, representing around 15% of EU GDP, has the potential to be a powerful catalyst for AI innovation.5 However, this will only be possible if procurement and funding rules are modernised. Streamlining tender processes and promoting innovation-oriented procurement strategies, such as lighter qualification requirements, challenge-based tenders and 'European preference' for critical AI technologies, would significantly improve accessibility for start-ups and emerging players. At the same time, state aid and funding approvals, particularly for strategic
5 European Commission (n.d.) Access to public procurement.
https://single-market-scoreboard.ec.europa.eu/business-framework-conditions/public-procurement_en
infrastructure, must be accelerated. The lengthy reviews that currently delay projects should be shortened, as recommended in the Commission’s Competitiveness Compass. Simplifying access to EU grants and programmes through clearer guidance, reduced administrative burdens and faster disbursement would further lower barriers for AI start-ups and SMEs.
However, procurement reform alone is insufficient. Regulatory simplification must be approached holistically. This does not mean lowering standards, but rather making them clearer, more consistent and easier to navigate. A framework for AI innovation that is fit for the future must ensure EU-wide coherence, proportionality, and legal certainty.
Lastly, We propose creating one-stop shops at the EU and national levels to centralise regulatory guidance, streamline permitting and coordinate assessments. We recommend integrating these centres into future AI infrastructure hubs, therefore making them part of the Cloud and AI Development Act. Second, expand smart sandboxes that offer early-stage guidance and fast-track approvals for high-potential AI projects, allowing controlled experimentation within clear boundaries.
4. Launch a EU Compute-for-Equity Facility to Empower AI Startups
Europe already owns the raw material for the next generation of smart subsidies: the EuroHPC supercomputers that idle for hours every day. The Draghi Report explicitly urges the EU to “offer free computing capacity in exchange for equity options, royalties or dividends to be reinvested in capacity and maintenance”, turning sunk infrastructure costs into recyclable risk capital.
This Act should translate that idea into law by creating a Compute-for-Equity Facility. EuroHPC would run competitive calls twice a year; winning start-ups receive a capped block of GPU time priced at a quarterly fair value, recorded as a convertible note or revenue-share claim. A narrow amendment to the General Block Exemption Regulation would deem such in-kind aid compatible with Articles 107 and 108 TFEU, provided the award emerges from a transparent tender and stays below a defined intensity ceiling. Returns flow into a ring-fenced Compute Sovereign Wealth Fund, financing the next hardware refresh.
With training costs for a mid-range frontier model falling by roughly forty per cent, AI startups could reach market sooner while the public sector gains a modest stake in their upside. A pilot that sets aside five per cent of the Leonardo and Jupiter machines in 2026
would validate demand, valuation and governance; scaling to twenty per cent of EuroHPC capacity by 2028 could recycle about €150 million a year into new super-computers. To ensure these benefits reach innovators across all Member States, the Facility should be complemented by a network of Regional AI Access Hubs. These hubs (integrated with national AI strategies and Digital Innovation Hubs) would offer onboarding, technical support and regulatory guidance, particularly for start-ups and SMEs in less-connected or under-resourced regions. Converting idle GPU-hours into equity is therefore the quickest, fiscally neutral way to match the cloud-credit firepower of foreign rivals and anchor Europe’s compute sovereignty on sustainable financial footing.
5. Modernise the EU’s Energy Infrastructure for Scalable AI Deployment
In addition to expanding Europe’s AI infrastructure, the EU must address energy system challenges to prevent bottlenecks that could constrain AI development. Demand for electricity from AI and cloud computing is growing rapidly. In 2018, EU data centres consumed 76.8TWh of electricity, with projections showing a 30% increase by 2030.6 AI-specific facilities, especially those used for training large language models (LLMs), are accelerating this trend and will require significant adaptation of Europe’s energy infrastructure and regulatory frameworks to remain competitive.
Data centres already account for roughly 3% of the EU’s total electricity consumption, but grid capacity and infrastructure distribution vary significantly across Member States, creating unequal opportunities and potential market distortions. Grid connection capacity is now a primary constraint for expansion. Traditional first-come, first-served allocation systems are proving inadequate for the strategic needs of AI infrastructure. In some countries, such as Poland, the process for reserving power connections is opaque and inefficient, with operators securing large capacities that remain unused for years. For example, approximately 30GW of power is reserved but not utilised in Poland, blocking new, viable data centre projects from accessing the grid and creating artificial scarcity. The lack of transparency about who holds these reservations further complicates planning and accountability.
To address these barriers, the European Commission should prioritise EU-wide coordination of grid investment planning and ensure that AI infrastructure needs are
6 European Commission (2020). Energy-efficient Cloud Computing Technologies and Policies for an Eco-friendly Cloud Market. https://digital-strategy.ec.europa.eu/en/library/energy-efficient-cloud-computing-technologies-and-policies-eco-friendly-cloud-market
integrated into energy market design, supporting cross-border electricity allocation and reducing fragmentation. A competitive support framework should help Member States implement energy pricing policies that reduce operational costs for AI infrastructure, including planned AI Gigafactories, while coordinating national investment programmes to avoid duplication and maximise synergies.
Improving transparency and efficiency in power reservation is essential. Consequently, the EU should introduce time-limited reservation periods that require demonstrable progress (such as construction permits or investment commitments) to retain allocations. Unused capacity should be automatically released for reallocation. Establishing a centralised, EU-wide digital registry of power reservations, managed by national regulators or a designated EU body, would increase oversight by publishing details such as the reserving entity, location, capacity size, and utilisation status in real time.
Finally, harmonised EU guidelines should prioritise feasible, sustainable projects and ensure consistent assessment of reservation requests. Financial disincentives for holding unused capacity, dynamic allocation models, and AI-driven grid management tools can further optimise utilisation. Closer cooperation among regulators, grid operators and data centre developers will improve forecasting and planning, ensuring Europe’s energy infrastructure can support the scale and speed required for sovereign, competitive AI development.
6. Strengthen Open Source AI and Governance
Open-source AI is crucial for ensuring that Europe’s digital transformation remains inclusive, adaptable and innovation-driven. Many sectors of the European economy, particularly those in industry, healthcare and public administration, require AI models tailored to their specific needs. Often, these use cases cannot be adequately served by large, general-purpose foundation models alone. Therefore, supporting a vibrant ecosystem of smaller, specialised and open AI models is essential to enable broad adoption, foster local innovation and preserve Europe’s technological sovereignty.
We therefore recommend that the Cloud and AI Development Act explicitly promotes and safeguards open-source AI development as public infrastructure. This should include expanding funding for open-source models through existing EU instruments and joint public–private partnerships. Access to computing power, particularly via
shared infrastructure and targeted voucher schemes, should be prioritised for open initiatives that deliver strategic value to the EU economy.
7. Improve the Regulatory Framework for Cross-Border-Data-Exchange
Europe’s competitiveness in AI will hinge on its ability to move data across borders. We therefore urge the Commission to weave today’s scattered rules into a single, technology-neutral “Trusted Cloud & AI Passport”. This passport would follow data, not servers, setting three clear risk tiers: friction-free intra-EU flows; streamlined transfers to “Adequacy +” jurisdictions that accept EU safeguards; and strictly conditioned processing elsewhere, backed by robust encryption and a binding submission to EU law.7 A consolidated rule-book lowers legal friction for startups, neutralises de facto localisation, and offers third-country partners a clear compliance ladder instead of a binary yes-or-no.
8. Promote and Strengthen European Digital Sovereignty
The Cloud and AI Development Act provides a vital opportunity to strengthen Europe’s digital sovereignty by fostering a secure, competitive and interoperable cloud and computing ecosystem. Critical public and industrial use cases - such as healthcare, defence, energy and banking - should take priority for sovereign infrastructure to safeguard sensitive data and maintain strategic autonomy. While non-EU hyperscale providers currently dominate cloud services in these sectors, even their so-called “sovereign cloud” offerings often fail to build intellectual property or long-term value within Europe. Relying on foreign providers risks keeping critical data and infrastructure under external influence, undermining the EU’s ability to control its own digital future.
To address this, the EU should establish clear ownership and operational requirements for cloud services in critical sectors. Regulations should define and enforce ultimate beneficial ownership rules to ensure that providers serving these sectors are ultimately controlled by EU-domiciled entities. Such requirements would help keep investments, technology and infrastructure aligned with European strategic interests. At the same time, cloud solutions deployed in critical sectors should be built on open-source
7 Art. 45 GDPR; “A transfer of personal data to a third country or an international organisation may take place where the Commission has decided that the third country, a territory or one or more specified sectors within that third country, or the international organisation in question ensures an adequate level of protection. Such a transfer shall not require any specific authorisation.”
frameworks or proprietary technologies owned by EU-based providers. This approach not only strengthens data sovereignty and security but also promotes the development of European intellectual property, fostering a self-sufficient digital ecosystem.
Supporting the development of the entire technology stack (from chips and AI hardware to sovereign platforms and services) will be essential. Targeted public investment, streamlined permitting, and cross-border coordination can help close the infrastructure gap and reduce Europe’s dependence on non-EU technology. Public procurement must act as a strategic lever by prioritising secure, Europe-based cloud services in tenders. Amending procurement directives to include strong standards for data sovereignty, security, technology ownership and open-source compatibility will ensure that public sector demand drives the growth of the European cloud ecosystem. This will de-risk private investment and encourage innovation, while providing clear incentives for the development of open-source and EU-owned solutions.
Finally, cloud interoperability and data portability must be guaranteed so that users, especially start-ups and SMEs, can avoid vendor lock-in and scale across borders. Open-source solutions, federated architectures and shared standards should be actively promoted to enable a diverse, resilient and innovative European digital ecosystem. By combining regulatory measures to secure critical sectors with support for EU-based providers and open technologies, the Act will be able to deliver a robust foundation for true European digital sovereignty.
For more information
European AI Forum: eaiforum.org
Contact: info@eaiforum.org
About the European AI Forum
The European AI Forum (EAIF) is the umbrella organisation of national European AI associations and clusters. Combined, we represent over 2.200 members, making us the largest European AI organisation. We represent members such as SMEs, corporates, organisations and individual experts in AI.