Summary No, the proposed Cloud and AI Development Act (CADA) does not establish a fixed numerical cap on the number of frontier AI priority projects the European Commission can recognise. Under Article 8, the Commission may recognise an unlimited number of projects selected through open calls, provided they meet strict qualitative criteria. However, the practical availability of support is constrained by Article 9, which mandates that computing resources be allocated only "within the limits of available capacity." Consequently, while recognition is not numerically capped, the actual delivery of compute support is scarcity-driven and dependent on physical infrastructure limits.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes a framework to strengthen Europe's cloud and AI ecosystem. A central pillar of this framework is the support for "frontier AI"β€”defined in Article 2(4) as "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." To accelerate these technologies, the proposal introduces the mechanism of "frontier AI priority projects."

No statutory numerical limit under Article 8

The proposal explicitly avoids setting a quantitative ceiling on the number of projects that can be designated. Article 8 grants the Commission the power to recognise projects as frontier AI priority projects "by means of a decision." The text states:

"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..."

The criteria listed in Article 8 are purely qualitative and structural:

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

Nowhere in Article 8 does the text mention a maximum number of projects, an annual quota, or a limit on the total volume of recognitions. The mechanism relies on "open calls for expression of interest," implying that any number of eligible projects could theoretically be recognised if they meet the criteria and are selected by the Commission. The Commission's discretion is exercised in evaluating the merit of applications against the strategic objectives of the Cloud and AI Leadership Initiatives, not in restricting the total count of recognitions.

The practical constraint: Article 9 and available capacity

While the recognition of projects is not capped, the support provided to them is strictly bounded by physical reality. Article 9, titled "Computing support for AI projects," introduces the critical operational constraint.

Article 9(1) states:

"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."

This phraseβ€”"within the limits of available capacity"β€”is the de facto bottleneck. It acknowledges that the Union's high-performance computing (HPC) infrastructure, particularly the capacity managed by the EuroHPC Joint Undertaking, is finite. The proposal does not guarantee that every recognised project will receive full support if the total demand exceeds the available supply.

Furthermore, Article 9(2) clarifies the matching mechanism:

"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 confirms that the Union's contribution is conditional. Even if a project is recognised under Article 8, the actual allocation of EU compute time is capped by the "Union's share of European high performance computing access time." If the sum of all recognised projects' requirements exceeds this share, the Union cannot fulfill all requests.

Selection via open calls and scarcity-driven prioritisation

The selection process is designed to be transparent and competitive. The Commission selects projects through "open calls for expression of interest." This suggests a merit-based evaluation where the Commission assesses which projects best align with the strategic goals of the Cloud and AI Leadership Initiatives, particularly Grand Challenge 3 (Frontier AI) outlined in Annex I.

Because there is no numerical cap on recognition but a hard cap on resources, the environment is inherently scarcity-driven. In practice, this means:

  • Prioritisation: The Commission may prioritise projects that demonstrate the highest strategic value, such as those addressing cybersecurity, scientific discovery, or critical industrial applications.
  • Resource Commitment: Projects that bring significant pooled resources from Member States (as required by Article 8(c)) may be favoured, as they reduce the burden on the Union's limited capacity.
  • Phasing: Support may be allocated in phases, with some projects receiving full support while others receive partial support or are placed on a waiting list until capacity becomes available.

The proposal does not specify a ranking methodology for cases where demand exceeds supply, leaving this to the Commission's discretion in the implementing acts or the evaluation of specific calls. However, the explicit reference to "limits of available capacity" in Article 9 serves as a clear warning that recognition does not equate to a guaranteed resource allocation.

What this means for you

For legal counsel, compliance officers, and project leaders in the AI sector, the absence of a numerical cap combined with the resource constraint in Article 9 creates a specific strategic landscape.

  1. Focus on Eligibility and Multi-State Collaboration: Since there is no "first-come, first-served" limit on the number of projects, the primary barrier to entry is structural. Ensure your consortium is legally eligible (e.g., an EDIC) and that you have secured binding commitments from at least three Member States to pool resources. Without this multi-state foundation, a project cannot be recognised, regardless of its technical merit.
  2. Resource Planning is Critical: Do not assume that recognition guarantees access to EU supercomputing resources. Article 9 explicitly ties support to "available capacity." When preparing your expression of interest, provide robust evidence of the computing time you have already secured or can contribute. Projects that demonstrate a clear path to leveraging existing EuroHPC capacity or that bring substantial national contributions may be viewed more favourably in a resource-constrained environment.
  3. Monitor Open Calls and Strategic Alignment: The Commission will issue open calls for expression of interest. Compliance teams should monitor these calls closely. Prepare documentation that clearly demonstrates how your project is "pioneering" and supports the scaling-up of frontier AI technologies, as required by Article 8(a). Align your proposal with the specific objectives of Grand Challenge 3 in Annex I to maximise your chances of selection.
  4. Strategic Positioning for Scarcity: Given that compute capacity is the limiting factor, position your project as essential to the Union's strategic autonomy. The proposal emphasises reducing dependencies on third-country technologies. Highlighting how your project contributes to this goal may strengthen your case during the Commission's selection process, especially if the Commission must prioritise among multiple eligible projects.

Common misconceptions

Misconception 1: There is a fixed quota for frontier AI priority projects. Correction: The CADA proposal does not set a numerical limit. The number of recognised projects will depend on the quality of applications received through open calls and the availability of computing resources.

Misconception 2: Recognition guarantees unlimited access to EU supercomputing resources. Correction: Recognition does not equate to an open-ended resource allocation. Article 9 explicitly states that support is provided "within the limits of available capacity." The Union and Member States are only obliged to allocate resources up to the point where their compute capacities are exhausted.

Misconception 3: Any AI project can apply for recognition. Correction: The criteria are strict. Projects must be "pioneering," support Grand Challenge 3, involve at least three Member States, and be undertaken by an eligible legal entity like an EDIC. General AI projects that do not meet the "frontier" threshold or lack multi-state collaboration will not qualify.

Misconception 4: The Commission has sole discretion without criteria. Correction: While the Commission makes the final decision, it is bound by the specific criteria in Article 8. It cannot recognise projects that fail to meet the requirements of being pioneering, multi-state, and structurally eligible.

Official sources

Related

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