Summary The European High Performance Computing Joint Undertaking (EuroHPC JU) is the EU body established by Council Regulation (EU) 2021/1173 that manages the Union's high-performance computing infrastructure. Under the proposed Cloud and AI Development Act (CADA), the EuroHPC JU provides the foundational compute capacity that the European Commission uses to match Member State contributions for designated frontier AI priority projects. Specifically, CADA Article 9(2) mandates that the Union shall match AI computing resources contributed by Member States to these priority projects, strictly limited to the extent that sufficient AI computing capacity is available within the Union's share of EuroHPC access time.

Detail

To understand how CADA supports AI development, one must first understand the infrastructure it relies upon: EuroHPC. The proposed regulation does not build new supercomputers from scratch; rather, it creates a regulatory and funding framework to allocate the compute time from existing and planned facilities managed by the EuroHPC JU.

What is EuroHPC?

EuroHPC is a strategic initiative of the European Union designed to boost the development, deployment, and use of high-performance computing (HPC) and artificial intelligence (AI) in Europe. The initiative is governed by the EuroHPC Joint Undertaking (EuroHPC JU), a public-private partnership established under Council Regulation (EU) 2021/1173.

As noted in CADA Recital 26, the implementation of CADA's Cloud and AI Leadership Initiatives may be entrusted to joint undertakings, explicitly naming the EuroHPC JU. The JU's primary role is to co-fund the construction of supercomputers (often referred to as "AI factories" or "AI gigafactories" in the CADA context) and to ensure their accessibility to researchers and industries across the Union. It acts as the operational vehicle for the EU's strategy to reduce dependency on non-European computing infrastructure and to foster technological sovereignty.

How CADA Leverages EuroHPC for Funding

CADA introduces a specific mechanism to accelerate the development of "frontier AI"β€”defined in CADA Article 2(4) as AI models or systems that approach, reach, or exceed the current state of the art. The regulation recognizes that developing these models requires unprecedented computational resources.

The link between CADA and EuroHPC is codified in Article 9: Computing support for AI projects. This article outlines how the Union and Member States will allocate compute resources to support strategic AI initiatives.

1. The Matching Mechanism for Frontier AI Priority Projects

CADA Article 8 defines "frontier AI priority projects" as pioneering projects focused on scaling up frontier AI technologies, undertaken by European digital infrastructure consortia (EDICs) or similar legal entities involving at least three Member States.

Once a project is designated as a frontier AI priority project, CADA Article 9(1) requires the Union and Member States to ensure sufficient AI computing resources are allocated from their respective compute capacities. However, the critical funding and resource-matching rule is found in Article 9(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."

This provision creates a direct financial and operational dependency on EuroHPC. The Commission's ability to "match" national contributions is capped by the availability of the Union's share of EuroHPC access time. In practical terms, this means that if a Member State contributes a certain amount of compute hours from its national supercomputers to a frontier AI project, the Commission will contribute an equivalent amount of compute hours drawn from the EuroHPC JU's pool, provided that pool has available capacity.

2. The Role of Council Regulation (EU) 2021/1173

The legal basis for this matching mechanism is the existing Council Regulation (EU) 2021/1173, which established the EuroHPC JU. CADA Recital 35 explicitly references this regulation, clarifying that the Union's matching obligation is "within the limits of available European high-performance computing ('EuroHPC') capacity."

The EuroHPC JU manages the "Union's share" of access time. This share represents the portion of the supercomputer time that the EU has funded and controls, distinct from the time contributed by individual Member States or private partners. By anchoring the CADA matching obligation to this specific share, the proposal ensures that the EU does not overcommit resources beyond what the EuroHPC JU has actually secured and allocated.

3. Ensuring Sufficient Capacity and Access Policy

CADA Recital 35 elaborates on this mechanism, stating that the allocation of AI computing resources to frontier AI priority projects is of strategic importance. It clarifies that the Union will match resources "on a proportional basis and within the limits of available European high-performance computing ('EuroHPC') capacity."

This matching mechanism is designed to lower the barrier to entry for European consortia. By guaranteeing that national contributions will be doubled (matched) by Union resources from the EuroHPC pool, CADA incentivizes Member States to pool their compute time and resources for large-scale, cross-border AI initiatives.

Furthermore, Recital 35 notes that the EuroHPC JU access policy should be accommodated to reflect the allocation of computing resources in an "efficient, transparent and timely manner." This implies that the EuroHPC JU must adapt its existing access frameworks to prioritize and accommodate the specific compute demands of CADA-designated frontier AI priority projects, without prejudicing the continuity of ongoing operations or the rights of projects already benefiting from allocated EuroHPC AI computing resources.

4. Broader Compute Support Beyond Frontier AI

While Article 9(2) specifically addresses frontier AI priority projects, Article 9(3) extends the obligation to ensure sufficient computing resources for other strategic areas. The Union and Member States shall "endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects."

While the strict "matching" formula of paragraph 2 applies specifically to frontier AI priority projects, the overarching goal of Article 9 is to leverage the EuroHPC infrastructure to support a broader ecosystem of AI development, including industrial and physical AI applications outlined in CADA's operational objectives.

What this means for you

For CTOs, architects, and SMEs involved in AI development, the interaction between CADA and EuroHPC has several practical implications:

  • Access to Subsidized Compute: If your organization is part of a consortium applying for designation as a "frontier AI priority project" under CADA Article 8, you may benefit from significantly increased compute capacity. The matching mechanism means that for every hour of compute your Member State contributes, the EU contributes an equivalent hour from the EuroHPC pool. This effectively doubles the available resources for your project, reducing the cost burden on individual participants.
  • Strategic Alignment with EuroHPC: To qualify for this support, projects must involve broad participation from entities across the Union, preferably through European Digital Infrastructure Consortia (EDICs) or similar structures. SMEs should look for opportunities to join these consortia rather than applying for compute resources individually.
  • Capacity Constraints: Be aware that the Union's matching commitment is not unlimited. It is strictly bounded by the "Union's share of European high performance computing access time." As demand for frontier AI compute grows, competition for this matched capacity will intensify. Early engagement with national authorities and the EuroHPC JU access policies will be crucial.
  • Beyond Frontier AI: Even if you are not working on frontier models, CADA Article 9(3) signals a political commitment to provide compute resources for industrial innovation and physical AI. While the matching mechanism is specific to frontier projects, the broader policy direction suggests that EuroHPC resources will increasingly be earmarked for strategic industrial AI applications, potentially creating new access pathways for SMEs in sectors like manufacturing, healthcare, and logistics.

Common misconceptions

  • Misconception: CADA creates new supercomputers. CADA does not directly fund the construction of new hardware. Instead, it leverages existing and planned infrastructure managed by the EuroHPC JU (under Council Regulation (EU) 2021/1173). CADA's role is to create a regulatory and funding framework that allocates the compute time from these existing facilities to specific strategic projects.
  • Misconception: The matching mechanism applies to all AI projects. The strict "matching" obligation in Article 9(2) applies only to designated "frontier AI priority projects." For other AI projects, such as industrial or public sector AI, Article 9(3) uses softer language ("endeavour to provide"), meaning there is no guaranteed 1:1 match of compute resources.
  • Misconception: EuroHPC access is automatic for EU companies. Access to EuroHPC compute time is competitive and governed by the EuroHPC JU's access policy. CADA prioritizes this access for frontier AI priority projects, but it does not grant blanket access to all European companies. Participation in recognized consortia and alignment with CADA's grand challenges are key prerequisites.

Related

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