Summary Under the proposed Cloud and AI Development Act (CADA), there is no detailed application form or procedure for frontier AI priority project recognition spelled out in the text. Article 8 sets out a single mechanism: the Commission "may, by means of a decision, recognise" projects "selected through open calls for expression of interest." To be in scope, a project must support grand challenge 3 in Annex I and meet three criteria — it is pioneering and focused on scaling up frontier AI; it is run by a European Digital Infrastructure Consortium (EDIC) or another legal entity eligible for funding under Union law, with at least three Member States participating; and the participating Member States pool computing time and other relevant resources. CADA is a proposal and not in force, so much of the practical process is yet to be defined.

Detail

The Cloud and AI Development Act (CADA) is a proposal (COM(2026) 502 final) and is not in force. If adopted in its current form, it would create a route for designating "frontier AI priority projects" — projects that support grand challenge 3 set out in Annex I, which concerns developing next-generation multimodal frontier AI models and systems and pioneering novel capabilities. The relevant provision is Article 8 of the proposed Regulation.

It is important to be precise about what the proposal actually says, because the draft text is short and does not lay out a bureaucratic application process. There are no forms, no submission deadlines, no evaluation timelines, no scoring methodology, and no appeals process in the text. What Article 8 provides is a power, not a procedure. It states that 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," subject to three criteria being fulfilled. The word "may" matters: recognition is discretionary, not automatic, even where the criteria are met.

The mechanism in the text

Reading Article 8 closely, the recognition pathway has two moving parts that are actually stated in CADA:

1. Selection through an open call for expression of interest. Article 8 ties recognition to projects "selected through open calls for expression of interest." This tells you that the entry point is a Commission-led call rather than a permanent, open-door submission system. The proposal does not, however, specify who issues each call, how often calls run, what an expression of interest must contain, or how selection decisions are made. Those operational details are left to be defined outside the Article 8 text — likely through the calls themselves and any implementing arrangements that follow if CADA is adopted.

2. Recognition by Commission decision. Where a project has been selected and the criteria are met, the Commission "may, by means of a decision, recognise" it as a frontier AI priority project. That decision is the legal act that confers the status. Again, the proposal does not set out a timeline for that decision or a route to challenge it.

The three criteria you must meet

To fall within Article 8, a project must satisfy all of the following cumulative criteria, quoted verbatim from the proposal:

  • (a) Pioneering project. It must be "a pioneering project, focused on the support and scaling-up of frontier AI technologies." This points away from incremental work on existing systems and toward genuinely advanced, large-scale efforts aligned with grand challenge 3.
  • (b) Eligible entity and at least three Member States. 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 and it involves the participation of at least three Member States." Note carefully: an EDIC is one option, but it is not the only one — any legal entity eligible for funding under Union law can qualify. What is mandatory is the participation of at least three Member States.
  • (c) Pooled resources. "The participating Member States pool computing time and other relevant resources to support the implementation of the designated project." This is a commitment that has to come from the Member States, not from a private provider acting alone.

Aligning with grand challenge 3

Because Article 8 requires the project to "support grand challenge 3 set out in Annex I," the substance of an expression of interest would need to map onto what that grand challenge describes. Annex I frames grand challenge 3 around "developing the next generation of multimodal frontier AI models and systems and pioneering novel capabilities," with a focus on next-generation multimodal models that "push the boundaries of current algorithmic capabilities" in advanced reasoning, cross-modal understanding and agentic capabilities, alongside novel approaches to model efficiency and alternative computational structures. A persuasive submission would therefore be expected to articulate, in concrete technical terms, how the project pursues those objectives rather than applying existing models to a narrow use case. This connects back to the "pioneering" wording in Article 8(a): the proposal is reserving this status for genuinely frontier work, and Annex I is the yardstick the Commission would use to judge that.

It is also worth noting that the resource-pooling commitment in Article 8(c) is described in the text as coming from "the participating Member States." In practice, an EDIC or other eligible entity would likely need to evidence binding commitments from the participating states before a call closes, since the pooled computing time and "other relevant resources" are part of the qualifying criteria, not an afterthought added once recognition is granted.

What recognition would unlock

The benefit of recognition is addressed in Article 9. In broad terms, the Union and Member States are to ensure sufficient AI computing resources within the limits of available capacity, and the Union "shall at least match" Member State contributions to the extent that sufficient capacity is available within the Union's share of European high-performance computing (EuroHPC) access time. EuroHPC was established by Council Regulation (EU) 2021/1173.

Two points follow. First, the matching is conditional on available capacity — it is not an open-ended guarantee of compute. Second, the matching is tied to Member State contributions, which connects directly back to the resource-pooling requirement in Article 8(c).

A note on EDICs and broad participation

A European Digital Infrastructure Consortium (EDIC) is a legal vehicle established under Decision (EU) 2022/2481 (the Digital Decade Policy Programme) that lets Member States jointly run multi-country digital infrastructure projects. Recital 34 of the proposal signals an intent to encourage broad participation from across the Union — in particular through EDICs, or any other legal structure representing a meaningful share of the Union's interest — and notes cybersecurity in particular. This reinforces that Article 8(b) is deliberately open to structures beyond an EDIC.

What this means for you

If you are a cloud service provider or data centre operator, the most useful thing to understand is that you cannot pursue this status as a single commercial entity. The status attaches to a multi-Member-State project run by an EDIC or another Union-funding-eligible legal entity, with Member States pooling resources. Your role is as a technical partner within that structure, not as a sole applicant.

Because CADA is a proposal and the detailed procedure is not yet defined, the steps below are practical expectations rather than legal requirements stated in the text. An applicant would likely be well served to:

  1. Build cross-border alliances early. Article 8(b) requires at least three participating Member States and an EDIC or other eligible legal entity. In practice that means engaging national authorities, research institutions, and industry partners well before any call appears. Your value proposition is the infrastructure and operational expertise to support and scale frontier AI work.
  2. Frame the project as pioneering. Article 8(a) uses the word "pioneering." You would be prepared to show how the project advances the state of the art and aligns with grand challenge 3, rather than optimising existing services.
  3. Line up resource pooling. Article 8(c) requires participating Member States to pool computing time and other relevant resources. You should be prepared to show how your infrastructure can integrate with shared resources, including EuroHPC capacity, although the proposal does not prescribe how this is done.
  4. Watch for calls. Recognition is tied to "open calls for expression of interest." Since the proposal does not set their timing or frequency, monitoring official Commission communications would be sensible once CADA is in force.

Treat any specific documents, formats, or deadlines you encounter in a future call as coming from that call, not from Article 8 itself — the Article leaves those details open.

Common misconceptions

Misconception 1: A single AI or cloud company can apply directly. Reality: Article 8(b) requires the project to be undertaken by an EDIC or another legal entity eligible for funding under Union law, with at least three Member States participating. A lone commercial provider cannot obtain this status by itself, though it can be a partner within a qualifying project.

Misconception 2: Only an EDIC can lead. Reality: An EDIC is one route, but Article 8(b) expressly also allows "another legal entity eligible for funding under Union law." The non-negotiable element is the participation of at least three Member States.

Misconception 3: CADA sets out a step-by-step application procedure. Reality: It does not. The proposal contains no forms, deadlines, evaluation timelines, scoring, or appeals. It states only that the Commission "may, by means of a decision, recognise" projects "selected through open calls for expression of interest." The operational detail is left to be defined.

Misconception 4: Recognition guarantees unlimited compute. Reality: Article 9 frames the Union contribution as a matching of Member State contributions to the extent sufficient capacity is available within the Union's share of EuroHPC access time. It is conditional on available capacity, not an unlimited grant.

Misconception 5: Meeting the criteria means recognition is automatic. Reality: Article 8 says the Commission "may" recognise qualifying projects. Recognition is discretionary, and it is conferred by a Commission decision.

Official sources

Related

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