Summary To establish a consortium eligible for recognition as a frontier AI priority project under the proposed Cloud and AI Development Act (CADA), you must organize through a European Digital Infrastructure Consortium (EDIC) or another legal entity eligible for Union funding. Crucially, Article 8(b) mandates that the project involves the participation of at least three Member States, and Article 8(c) requires these states to pool computing time and other relevant resources. As proposed, the Commission may recognize such projects via a decision following open calls, unlocking access to Union computing capacity and strategic support.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes a targeted mechanism to accelerate the development of "frontier AI" within the Union. Unlike general AI initiatives, frontier AI priority projects are designed to tackle specific "grand challenges" that require massive scale, cross-border coordination, and significant computational resources. The legal framework for forming the consortia that will deliver these projects is defined in Article 8 of the proposal.

This article sets out three cumulative criteria that must be met for the Commission to recognize a project as a "frontier AI priority project." Failure to meet any single criterion disqualifies the application. The following sections detail the structural, geographic, and operational requirements necessary to build a compliant consortium.

1. The Legal Vehicle: EDICs and Eligible Entities

The first structural hurdle is the legal form of the consortium. Article 8(b) explicitly states that the project 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."

The Preferred Vehicle: European Digital Infrastructure Consortium (EDIC)

The regulation identifies the EDIC as the primary vehicle for these strategic projects. Established under the Digital Decade Policy Programme (Decision (EU) 2022/2481), an EDIC is a legal entity created by a group of Member States to develop, deploy, and operate digital infrastructure of European interest.

  • Why use an EDIC? Using an EDIC signals a high degree of political and operational commitment from the participating states. It provides a pre-existing legal framework for cross-border cooperation, simplifying the governance of shared assets like computing capacity and data.
  • Structure: An EDIC is typically formed by a group of Member States (and potentially associated countries) to coordinate infrastructure projects. For a frontier AI project, the EDIC would act as the legal applicant, managing the consortium's activities and ensuring alignment with Union objectives.

The Alternative: "Another Legal Entity Eligible for Union Funding"

Recognizing that not all consortia may fit the strict EDIC model immediately, Article 8(b) offers a fallback option: "another legal entity eligible for funding under Union law."

  • Scope: This could include joint undertakings (such as the EuroHPC JU or the Smart Networks and Services Joint Undertaking), public-private partnerships, or other legal structures established under Union law that are eligible to receive funding from programs like Horizon Europe or the Digital Europe Programme.
  • Eligibility Check: The entity must be capable of receiving Union funding. This implies it must have the legal capacity to enter into grants or contracts with the Commission and the financial and administrative capacity to manage large-scale research and innovation projects.
  • Strategic Note: While the alternative offers flexibility, the EDIC remains the most robust vehicle for ensuring the "European dimension" required by the Act. Entities choosing the alternative route must demonstrate clearly that they meet the eligibility criteria for Union funding and can effectively coordinate the multi-state participation required.

2. The Geographic Mandate: Minimum Three Member States

A frontier AI priority project cannot be a national initiative. Article 8(b) imposes a strict geographic threshold: the project must "involve the participation of at least three Member States."

What "Participation" Means

The requirement for "participation" implies more than just having partners located in three countries. It requires active involvement in the project's governance and execution.

  • Governance: The consortium's decision-making bodies must reflect the involvement of the three Member States. This ensures that the project serves the strategic interests of the Union rather than a single national agenda.
  • Resource Contribution: The participation is directly linked to the resource-pooling requirement (see below). The three Member States must be the ones contributing the computing time and resources.
  • Strategic Rationale: This threshold is designed to prevent fragmentation. By requiring a minimum of three states, CADA ensures that frontier AI development is a collective European effort, fostering interoperability and creating a critical mass of resources that no single Member State could easily replicate alone.

Implications for Consortium Building

When assembling your consortium, you must secure commitments from at least three distinct Member States. This often involves:

  1. Identifying National Competent Authorities: Engaging with the relevant ministries or agencies in three different Member States that are willing to champion the project.
  2. Formalizing Commitments: Ensuring that the participation is formalized, likely through the EDIC agreement or the founding documents of the alternative legal entity.
  3. Geographic Balance: While the Act does not specify a geographic distribution (e.g., North/South/East/West), a balanced consortium is often more competitive in the Commission's evaluation, as it aligns with the broader goal of reducing regional disparities in digital capacity.

3. The Operational Core: Pooling Computing Time and Resources

The defining operational characteristic of a frontier AI priority project is the pooling of resources. Article 8(c) mandates that "the participating Member States pool computing time and other relevant resources to support the implementation of the designated project."

Pooling Computing Time

Frontier AI models require immense computational power, often exceeding the capacity of individual national supercomputing centers.

  • The Mechanism: The participating Member States must commit to sharing their available High-Performance Computing (HPC) capacity. This is not merely a promise of future access; it is a requirement to actively pool existing or planned resources.
  • EuroHPC Integration: This pooling mechanism is closely linked to Article 9, which states that the Union shall "match, on a proportional basis and within the limits of available European high-performance computing ('EuroHPC') capacity, the AI computing resources contributed or committed by the Member States."
  • Strategic Value: By pooling resources, the consortium creates a unified, high-capacity compute environment. This allows for the training of larger, more complex models that would be impossible on isolated national clusters.

"Other Relevant Resources"

While computing time is the primary resource, Article 8(c) explicitly includes "other relevant resources." This broadens the scope of the consortium's obligations and opportunities:

  • Data: Access to high-quality, diverse datasets is critical for training frontier AI. Member States may pool access to public sector data, research datasets, or anonymized private data, subject to data protection rules.
  • Expertise: The consortium may pool human capital, including researchers, data scientists, and domain experts from across the participating states.
  • Infrastructure: Beyond raw compute, this could include access to specialized hardware (e.g., AI-optimized accelerators, quantum prototypes) or testing facilities.
  • Funding: While the Act focuses on computing resources, the pooling of financial resources from national budgets to support the project is often a practical necessity to sustain the high costs of frontier AI development.

4. The Commission's Role and the Recognition Process

The Commission acts as the gatekeeper and facilitator of this framework.

  • Open Calls: The process begins with an "open call for expression of interest." The Commission will publish criteria and timelines for applicants to submit their proposals.
  • Evaluation: Proposals are evaluated against the cumulative criteria in Article 8. The Commission assesses whether the legal entity is eligible, whether three Member States are participating, and whether the resource-pooling plan is robust.
  • Decision: If the criteria are met, the Commission recognizes the project as a "frontier AI priority project" by means of a decision. This recognition is the key that unlocks the benefits of the Act, including the matching of computing resources under Article 9.
  • Grand Challenge Alignment: It is important to note that the project must support "grand challenge 3" set out in Annex I. This challenge focuses on "developing the next generation of multimodal frontier AI models and systems." The consortium's proposal must clearly demonstrate how its objectives align with this specific strategic goal.

What this means for you

For technology leaders, research institutions, and public authorities, the CADA proposal outlines a clear but demanding path to accessing Union-level support for frontier AI.

1. Secure the Legal Vehicle First Before approaching Member States, ensure you have a legal vehicle in place. If you are a private entity, you cannot apply alone. You must either join an existing EDIC or form a new legal entity that is eligible for Union funding. If you are a public authority, work with your national ministry to establish or join an EDIC. The legal structure is the foundation; without it, the project cannot proceed.

2. Build a Tri-National Coalition Identify at least two other Member States willing to participate. This is a hard constraint. Look for states with complementary strengths: one might have strong HPC infrastructure, another might have advanced AI research capabilities, and a third might have specific domain expertise (e.g., in healthcare or manufacturing). The consortium must be a genuine partnership, not just a nominal listing of partners.

3. Develop a Concrete Pooling Plan Your expression of interest must include a detailed plan for resource pooling. Do not just state that you will "pool resources." Specify:

  • How much computing time (in FLOPs or hours) each Member State will contribute.
  • What "other resources" (data, expertise, hardware) are being pooled.
  • How these resources will be managed and accessed by the consortium.
  • How the Union's matching resources (under Article 9) will be integrated.

4. Align with Grand Challenge 3 Ensure your project's scientific and technical objectives are explicitly aligned with "grand challenge 3." Your proposal should detail how your project will advance "multimodal frontier AI models," "advanced reasoning," "cross-modal understanding," or "agentic capabilities." Projects that focus on incremental improvements or non-frontier applications will likely be rejected.

5. Prepare for a Competitive Process Recognition is not automatic. The Commission will evaluate proposals based on their strategic value and feasibility. Prepare a high-quality proposal that demonstrates the consortium's ability to deliver on its promises. Engage with the Commission early, attend stakeholder consultations, and monitor the open calls closely.

Common misconceptions

"Any AI project can be a frontier AI priority project." No. The Act is specific. Only projects that support "grand challenge 3" (next-generation multimodal frontier AI models and systems) and meet the strict criteria of Article 8 can be recognized. General AI projects, even if innovative, do not qualify.

"A consortium of two Member States is sufficient." No. Article 8(b) explicitly requires the participation of at least three Member States. A consortium limited to one or two states cannot be recognized as a frontier AI priority project.

"Private companies can apply individually." No. The project must be undertaken by an EDIC or another legal entity eligible for Union funding. Private companies must participate through these consortia. They cannot apply as a standalone entity.

"Pooling resources is a suggestion, not a requirement." No. Article 8(c) mandates that participating Member States pool computing time and other relevant resources. This is a cumulative criterion. Without a concrete plan for pooling, the project cannot be recognized.

"Recognition guarantees funding." While recognition unlocks access to computing resources and strategic support, it does not automatically guarantee direct financial grants. The project must still compete for funding under relevant Union programs (e.g., Horizon Europe, Digital Europe) or rely on the pooled national resources. However, recognition significantly strengthens the project's profile and eligibility.

Official sources

Related

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