Summary To secure recognition as a "frontier AI priority project" under the proposed Cloud and AI Development Act (CADA), a consortium must be formed by a European digital infrastructure consortium (EDIC) or an eligible Union legal entity, involving at least three Member States. The consortium must respond to a Commission open call, demonstrating that the project supports Grand Challenge 3 (frontier AI models) and that Member States have committed to pooling computing resources. Upon a positive Commission decision under Article 8, the project unlocks a matching mechanism under Article 9, where the Union matches Member State contributions with AI computing resources from the European High-Performance Computing Joint Undertaking (EuroHPC JU), subject to available capacity.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes a targeted mechanism to accelerate the development of frontier AI technologies within the Union. Recognizing that frontier AI requires unprecedented computational power and cross-border collaboration, the proposal creates a specific statusβ€”"frontier AI priority project"β€”to facilitate access to high-performance computing (HPC) resources.

This mechanism is governed by Article 8 (Criteria for frontier AI priority projects) and Article 9 (Computing support for AI projects). Unlike general innovation funding, this pathway is strictly reserved for projects that address the most advanced technological frontiers and require a multi-national, pooled-resource approach.

The following guide details the mandatory steps, legal requirements, and practical sequencing for obtaining this recognition.

Step 1: Establish the Correct Legal Consortium Structure

The first and most fundamental barrier to entry is the legal structure of the applicant. CADA explicitly prohibits single-entity or bilateral applications for this specific status.

According to Article 8(1), point (b), the project must 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." The proposal further clarifies that this entity must be "capable of representing a meaningful share of the Union's interest."

This requirement ensures that the project has a governance structure capable of managing cross-border data, resources, and strategic objectives. A single cloud provider, a university, or a national research institute acting alone cannot apply. They must first establish or join an EDIC or a similar cross-border legal vehicle.

The Three-Member State Mandate Crucially, Article 8(1), point (b) mandates that the project "involves the participation of at least three Member States." This is not merely a suggestion for broad collaboration; it is a cumulative criterion for recognition. The consortium must demonstrate active involvement from entities in at least three distinct EU Member States. This requirement aligns with the CADA's broader objective of reducing fragmentation and ensuring that strategic AI assets are developed across the Union rather than concentrated in a single national hub.

Practical Implication: If your organization is a private cloud provider or a data centre operator, you cannot apply directly. You must partner with public authorities or research entities in at least three Member States to form or join an EDIC. This consortium will serve as the legal applicant.

Step 2: Align with Grand Challenge 3

Not all AI projects qualify for priority status. The proposal restricts this mechanism to projects that address the most advanced, strategic technological gaps.

Article 8(1) states that the Commission may recognize projects that "support grand challenge 3 set out in Annex I."

Referring to Annex I of the proposal, Grand Challenge 3 is defined as:

"Developing the next generation of multimodal frontier AI models and systems and pioneering novel capabilities."

The annex elaborates that the focus is on:

"The architectural design and development of next-generation multimodal models and systems that push the boundaries of current algorithmic capabilities for achieving superior performance in advanced reasoning, cross-modal understanding and agentic capabilities."

It further includes investigating "novel approaches to model efficiency, cognitive modelling, and alternative computational structures."

What this means for your proposal: Your application must clearly articulate how the project moves beyond the deployment of existing models. It must target:

  • Frontier capabilities: Advanced reasoning, agentic behavior, and cross-modal understanding.
  • Strategic sectors: The annex notes potential applications in "foundational science such as scientific discovery and complex data interpretation."
  • Novelty: The project must pioneer capabilities that are not yet state-of-the-art.

Projects focused solely on industrial AI (Grand Challenge 5), physical AI (Grand Challenge 4), or the deployment of existing models for public administration (Grand Challenge 8) do not qualify for this specific "frontier AI priority project" recognition, even if they are valuable.

Step 3: Secure Pooled Resource Commitments

A unique feature of the CADA frontier AI mechanism is the requirement for Member States to commit resources before recognition is granted. The proposal envisions a collaborative model where national investments are pooled to support a Union-level strategic asset.

Article 8(1), point (c) requires that "the participating Member States pool computing time and other relevant resources to support the implementation of the designated project."

This is a critical sequencing step. You cannot simply propose a project and ask the EU to fund it later. The consortium must secure commitments from the participating Member States to contribute:

  1. Computing time: Likely from national high-performance computing (HPC) infrastructures.
  2. Other relevant resources: This could include data, talent, or specific infrastructure access.

The proposal emphasizes that this pooling is a prerequisite for the Commission's decision. The application must demonstrate that these commitments are real and binding, not aspirational.

Step 4: Respond to the Commission Open Call

The Commission does not accept unsolicited applications for frontier AI priority project status. The process is triggered exclusively by a formal call.

Article 8(1) specifies that projects are "selected through open calls for expression of interest."

When the Commission publishes such a call, the consortium must submit an expression of interest that explicitly addresses the three cumulative criteria of Article 8:

  1. Pioneering Nature: Demonstrate that the project is focused on the support and scaling-up of frontier AI technologies (Article 8(1), point (a)).
  2. Consortium Structure: Prove that the project is undertaken by an EDIC or eligible entity involving at least three Member States (Article 8(1), point (b)).
  3. Resource Pooling: Provide evidence that the participating Member States have committed to pooling computing time and other resources (Article 8(1), point (c)).

The open call will likely specify the technical format, deadlines, and specific evidence required to prove the "pooling" of resources.

Step 5: Commission Decision and Recognition

Once the open call closes, the Commission evaluates the expressions of interest against the strict criteria in Article 8.

If the project meets all requirements, the Commission adopts a formal legal act. Article 8(1) states:

"The Commission may, by means of a decision, recognise as frontier AI priority projects, projects selected through open calls for expression of interest..."

This decision is the definitive legal instrument that grants the project the status of a "frontier AI priority project." It is not a grant agreement for cash, but a recognition of strategic priority that triggers specific support obligations under Article 9.

Step 6: Receive Computing Support via Article 9

The primary value of recognition is the unlocking of significant AI computing resources through a matching mechanism. Article 9 outlines the obligations of the Union and Member States once a project is recognized.

Article 9(1) establishes the general obligation:

"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."

The Matching Mechanism (Article 9(2)) The core benefit is the matching rule. Article 9(2) states:

"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:

  • If your consortium secures 100 petaflops of compute time from three Member States, the Union (via the EuroHPC JU) will match this with at least 100 petaflops from the Union's share of EuroHPC capacity.
  • This effectively doubles the compute resources available to the project, leveraging national investments with Union resources.
  • Limitation: The matching is subject to "the extent that sufficient AI computing capacity is available." It is not an unlimited guarantee but is tied to the actual availability of EuroHPC access time.

Article 9(3) notes that the Union and Member States shall also "endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects," but the mandatory matching mechanism in paragraph 2 is specific to frontier AI priority projects.

What this means for you

For cloud service providers, data centre operators, and AI research entities, the CADA frontier AI mechanism represents a high-stakes, high-reward pathway.

  1. Consortium Building is Non-Negotiable: You cannot apply alone. If you are a private provider, your immediate priority must be to identify potential partners in at least three Member States and establish an EDIC or join an existing one. This legal structure is the gateway to the process.
  2. Focus on "Frontier" Capabilities: Ensure your technical roadmap aligns strictly with Grand Challenge 3. Projects focused on incremental improvements or standard industrial applications will be rejected. You must demonstrate a leap in "advanced reasoning," "agentic capabilities," or "multimodal understanding."
  3. Secure National Commitments Early: The requirement to "pool computing time" under Article 8(1)(c) means you must engage with national HPC authorities and ministries before the open call. You need binding commitments, not just letters of intent.
  4. Monitor Open Calls: The process is not continuous. You must track Commission announcements for the specific "open calls for expression of interest" under Article 8. Missing a call means waiting for the next cycle.
  5. Leverage the Match: Once recognized, the Article 9 matching mechanism is a powerful tool. It allows you to scale your compute access significantly beyond what any single Member State could provide. Coordinate closely with the EuroHPC JU to ensure your technical architecture can integrate with the allocated resources.

Common misconceptions

"Any AI project can apply for priority status."

  • Correction: No. Only projects supporting Grand Challenge 3 (frontier AI models and systems) are eligible. Projects focused on industrial AI, physical AI, or public sector AI, while supported by other parts of CADA, do not qualify for this specific "frontier AI priority project" status and its matching compute mechanism.

"A single company can apply directly."

  • Correction: No. Article 8(1)(b) explicitly requires the project to be undertaken by an EDIC or an eligible Union legal entity involving at least three Member States. Single-entity applications are ineligible.

"Recognition guarantees unlimited compute."

  • Correction: No. Article 9(1) and Article 9(2) specify that resources are allocated "within the limits of available capacity." The Union matches contributions only to the extent that sufficient AI computing capacity is available within the Union's share of EuroHPC access time. If EuroHPC capacity is fully utilized, the match may be partial or delayed.

"The Commission provides direct cash funding for the project."

  • Correction: The recognition under Article 8 unlocks computing resources (access time on HPC infrastructure), not necessarily direct cash grants for R&D salaries or hardware purchases. While other funding streams (like Horizon Europe or the Digital Europe Programme) may complement the project, the specific benefit of this mechanism is in-kind compute support.

"The project must be entirely owned by the EU."

  • Correction: No. The project must be undertaken by an EDIC or eligible entity, but the consortium can include private sector partners. The key is the cross-border structure and the pooling of national resources, not total public ownership.

Official sources

Related

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