Summary As proposed, the Cloud and AI Development Act (CADA) would create a formal synergy between the EU's strategic cloud framework and the European High-Performance Computing (EuroHPC) Joint Undertaking. Under Article 9, the Union would "at least match" AI computing resources contributed by Member States to designated "frontier AI priority projects" using the Union's share of EuroHPC access time. This mechanism, grounded in Recital 31's reference to Council Regulation (EU) 2021/1173, ensures that Europe's supercomputing capacity is strategically aligned with frontier AI development without displacing existing research commitments.

Detail

The proposed Cloud and AI Development Act (CADA) addresses the EU's critical shortage of computing capacity by establishing a coordinated ecosystem approach. A central pillar of this approach is the "Cloud and AI Leadership Initiatives," designed to bridge the gap between advanced research capabilities and their sustainable exploitation. Within this framework, the relationship with the EuroHPC Joint Undertaking (JU) — established under Council Regulation (EU) 2021/1173 — is explicitly defined to ensure that Europe's supercomputing assets are strategically aligned with the development of frontier AI technologies.

Synergies with EuroHPC Infrastructure

Recital 31 of the CADA proposal explicitly states that the Cloud and AI Leadership Initiatives should enhance synergies with actions currently supported by the Union, including those under Council Regulation (EU) 2021/1173 (the EuroHPC JU). The proposal recognizes that EuroHPC provides the high-performance computing infrastructure necessary for the most computationally intensive AI workloads. By coordinating with EuroHPC, CADA seeks to avoid duplication of funding and promote alignment of priorities across governance levels, ensuring that public investments in supercomputing directly feed into the development of European AI sovereignty.

The proposal emphasizes that these measures are without prejudice to the rules and procedures laid down in Council Regulation (EU) 2021/1173. This means the EuroHPC JU remains the operational vehicle for delivering these compute resources, while CADA provides the strategic direction and funding leverage for specific AI use cases.

Frontier AI Priority Projects and Compute Allocation

The mechanism for this synergy is operationalized through Article 8 and Article 9, which establish a dedicated pathway for "frontier AI priority projects."

Article 8 sets out the criteria for the Commission to recognize specific projects as frontier AI priority projects. To qualify, a project must:

  1. Be a pioneering project focused on the support and scaling-up of frontier AI technologies.
  2. Be undertaken by a European digital infrastructure consortium (EDIC) established pursuant to Decision (EU) 2022/2481 or another legal entity eligible for funding under Union law.
  3. Involve the participation of at least three Member States.
  4. Demonstrate that participating Member States pool computing time and other relevant resources to support the project.

This structure ensures that frontier AI development is not fragmented but driven by collaborative, cross-border efforts with significant industrial and research backing.

Article 9 dictates how compute resources are allocated to these designated projects, creating a direct financial and technical link to EuroHPC. The key provisions are:

  • Guaranteed Allocation: Under Article 9(1), the Union and Member States shall ensure that sufficient AI computing resources from their compute capacities are allocated to support frontier AI priority projects that meet Article 8 criteria, "within the limits of available capacity."
  • EU Matching Mechanism: Crucially, Article 9(2) states that 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." This means that if Member States commit a certain amount of compute time to a priority project, the EU will add an equivalent amount from its EuroHPC allocation, effectively doubling the available compute for these strategic initiatives.
  • Broader Support: Beyond frontier AI, Article 9(3) notes that the Union and Member States shall "endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects."

Operational Integration and Access Policy

The proposal acknowledges that this new allocation model must operate within the existing EuroHPC governance structure. Recital 35 clarifies that the EuroHPC JU's access policy should be "accommodated to reflect the allocation of such computing resources in an efficient, transparent and timely manner." Importantly, this accommodation must occur "without prejudice to the continuity of ongoing operations and the rights of projects already benefiting from allocated EuroHPC AI computing resources." This ensures that the new CADA-driven demands for frontier AI compute do not disrupt existing scientific and industrial users of EuroHPC supercomputers.

Furthermore, the proposal emphasizes that these measures are without prejudice to the rules and procedures laid down in Council Regulation (EU) 2021/1173, meaning the EuroHPC JU remains the operational vehicle for delivering these compute resources, while CADA provides the strategic direction and funding leverage for specific AI use cases.

What this means for you

For CTOs, architects, and SMEs evaluating the practical impact of CADA, the interaction with EuroHPC represents a significant opportunity to access high-end compute resources that are typically scarce and expensive.

For Frontier AI Developers: If you are involved in developing frontier AI models (those approaching or exceeding the current state of the art), the pathway to EuroHPC compute becomes more structured. By aligning your project with the "frontier AI priority project" criteria in Article 8 — particularly by partnering with at least two other Member States and forming a European Digital Infrastructure Consortium (EDIC) — you can unlock the EU matching mechanism in Article 9. This effectively doubles your available compute budget from EuroHPC sources, reducing the financial barrier to training large-scale models.

For Industrial and Physical AI Innovators: While the strict "matching" mechanism applies to frontier AI, Article 9(3) indicates a broader commitment to providing compute for industrial and physical AI. SMEs and industrial actors should monitor the implementation of the Cloud and AI Leadership Initiatives, as these sectors are explicitly prioritized for compute support. Engaging with national strategies (required under Article 7) and local Experience and Acceleration Centres for AI (Article 5) will be crucial for identifying how to access these allocated resources.

For Infrastructure Architects: The requirement to "accommodate" EuroHPC access policies (Recital 35) suggests that the technical interfaces for accessing AI compute on EuroHPC supercomputers may evolve to better support AI-specific workloads (e.g., GPU optimization, specific AI frameworks). Architects designing AI pipelines should prepare for potential changes in how EuroHPC resources are booked and managed under the CADA framework, ensuring their systems are compatible with the standardized access procedures that will likely emerge from the EuroHPC JU's updated policies.

Strategic Planning: The emphasis on pooling resources (Article 8(c)) means that solo efforts are less likely to succeed. Organizations should proactively seek cross-border partnerships to meet the "at least three Member States" threshold. This collaborative requirement is designed to strengthen the European AI ecosystem as a whole, but it also means that access to premium EuroHPC compute will be increasingly tied to collaborative, pan-European industrial consortia.

Common misconceptions

Misconception 1: CADA replaces EuroHPC. No. CADA does not replace the EuroHPC Joint Undertaking or its infrastructure. Instead, it leverages EuroHPC's existing supercomputing capacity. EuroHPC remains the operational entity managing the hardware and access, while CADA provides the strategic framework and additional funding leverage (via the matching mechanism) for specific AI projects.

Misconception 2: All AI projects automatically get EuroHPC compute. The guaranteed matching mechanism in Article 9(2) applies specifically to "frontier AI priority projects" that have been formally recognized by the Commission under Article 8. General AI projects, including many industrial applications, may receive support through other Cloud and AI Leadership Initiatives measures, but they do not automatically qualify for the EU matching of Member State contributions within EuroHPC access time.

Misconception 3: Existing EuroHPC users will lose access. The proposal explicitly states that the accommodation of access policies must occur "without prejudice to... the rights of projects already benefiting from allocated EuroHPC AI computing resources" (Recital 35). The goal is to expand and strategically direct capacity, not to displace existing scientific or industrial users.

Misconception 4: CADA provides direct cash grants for hardware. CADA's interaction with EuroHPC is primarily about access to existing supercomputing resources (compute time), not the direct construction of new data centers for AI. While CADA does have provisions for data center deployment (Title III), the specific synergy with EuroHPC is focused on allocating compute capacity for AI training and development, not building new physical infrastructure outside the EuroHPC framework.

Official sources

Related

This is general information about a draft EU regulation, not legal advice.