Summary Under the proposed Cloud and AI Development Act (CADA), frontier AI priority projects do not run exclusively on "AI gigafactory" infrastructure by legal mandate, but they are designed to leverage a hybrid ecosystem of national and Union resources. Article 7(2)(e) obliges Member States to invest in high-intensity assets, explicitly including "AI gigafactories," within their national strategies. The actual compute allocation for recognized projects follows a specific pathway defined in Article 9: Member States contribute resources (potentially from gigafactories), and the Union matches these contributions "to the extent that sufficient AI computing capacity is available within the Union's share of European high performance computing access time." Thus, while gigafactories form the national backbone, the Union-matched portion is sourced from EuroHPC capacity, creating a sovereign, multi-source compute pathway for frontier AI.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes a dual-track approach to securing the compute resources necessary for Europe's strategic autonomy in artificial intelligence. One track focuses on the physical deployment of sovereign infrastructure through national obligations; the other focuses on the allocation mechanism for specific high-priority projects. To determine whether frontier AI priority projects run on AI gigafactory infrastructure, one must analyze the interplay between the national investment mandates in Article 7 and the resource allocation rules in Article 9.

The National Infrastructure Mandate: Article 7(2)(e)

The foundation of the CADA compute strategy is the requirement for Member States to build a robust domestic capacity. Article 7 mandates that Member States establish national cloud and AI strategies within one year of the Regulation's entry into force. Crucially, Article 7(2)(e) specifies the content of these strategies, requiring them to include "measures to invest in high-intensity computing infrastructure, including AI factories, AI gigafactories and quantum computers as strategic national and cross-border assets supporting research, development and industrial AI deployment across strategic sectors."

This provision does not create a direct legal link forcing a specific project to run on a specific gigafactory. Instead, it creates a systemic obligation: the national ecosystem must contain these high-capacity assets. The term "AI gigafactories" is used to denote large-scale, dedicated facilities for AI model training and deployment, distinct from general-purpose cloud data centres. By mandating their inclusion in national strategies, CADA ensures that Member States are building the physical "high-intensity" capacity required to support frontier AI, reducing reliance on third-country hyperscalers.

The Allocation Mechanism: Article 9 and the Matching Principle

Once a project is designated as a frontier AI priority project (a status granted by the Commission under Article 8 based on criteria such as pioneering nature and multi-state participation), the mechanism for securing compute is governed by Article 9.

Article 9(1) establishes the primary 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... within the limits of available capacity."

The critical operational detail lies in Article 9(2), which defines the "matching" principle:

"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 clause clarifies the source of the compute:

  1. Member State Contribution: Member States must pool computing time and resources. These resources could logically include capacity from the "AI gigafactories" they are mandated to build under Article 7(2)(e).
  2. Union Matching: The Union does not create a new, isolated cloud for these projects. Instead, it matches the Member State contribution using its share of European high-performance computing (EuroHPC) capacity.
  3. The Limitation: The Union's matching is strictly capped by "available capacity" within the EuroHPC framework.

The Role of EuroHPC and Recital 35

The relationship between the national gigafactories and the Union-matched compute is further clarified in Recital 35 of the explanatory memorandum. The recital states that the allocation of resources is of strategic importance and that the Union should match resources "within the limits of available European high-performance computing ('EuroHPC') capacity."

It further notes that the "EuroHPC JU access policy should be accommodated to reflect the allocation of such computing resources in an efficient, transparent and timely manner." This indicates that the EuroHPC Joint Undertaking (JU) is the primary vehicle for delivering the Union-matched portion of the compute. The infrastructure supporting these projects is therefore a composite:

  • National Layer: High-intensity assets (AI gigafactories, AI factories) built by Member States as per Article 7(2)(e).
  • Union Layer: EuroHPC capacity, which serves as the matching resource under Article 9(2).

The practical implication is that a frontier AI priority project would likely access a hybrid environment. The Member States' contribution might come from their national gigafactories, while the Union's matching contribution comes from the EuroHPC network. The project does not run exclusively on gigafactories, but it runs on a sovereign ecosystem where gigafactories are a mandated component of the national contribution.

The Practical Compute Pathway

For a project developer or a Member State authority, the pathway to securing compute under CADA involves a structured sequence:

  1. Designation: The project must be recognized as a "frontier AI priority project" by the Commission. This requires meeting the criteria in Article 8, including being a pioneering project, involving at least three Member States, and pooling resources.
  2. National Strategy Alignment: The participating Member States must ensure their national strategies (under Article 7) include the necessary investment in high-intensity infrastructure, such as AI gigafactories, to support the project.
  3. Resource Pooling: The Member States contribute their computing resources (which may include gigafactory capacity) to the project.
  4. Union Matching: The Commission allocates matching resources from the Union's share of EuroHPC capacity, as per Article 9(2).
  5. Access Policy Adaptation: The EuroHPC JU adjusts its access policy to accommodate these strategic allocations, ensuring the project receives the matched capacity efficiently.

This pathway ensures that frontier AI development is supported by a mix of newly built national capacity (gigafactories) and existing or planned Union capacity (EuroHPC), creating a resilient and sovereign compute foundation.

What this means for you

For technology leaders, CTOs, and infrastructure architects, the proposed CADA framework signals a shift towards a sovereign, multi-source compute model for frontier AI.

  • Align with National Strategies: Since Article 7(2)(e) mandates Member States to invest in AI gigafactories, your infrastructure strategy should align with these national plans. Engage early with national authorities to understand where these assets will be deployed and how they will be integrated into the broader compute pool.
  • Prepare for Hybrid Access: Do not assume a single infrastructure provider. The compute pathway under Article 9 implies a hybrid model where national contributions (potentially gigafactories) are matched by EuroHPC capacity. Your technical architecture should be flexible enough to operate across these different sovereign environments.
  • Leverage the Matching Mechanism: If you are leading a frontier AI initiative, ensure your consortium includes at least three Member States to qualify for the Union matching under Article 9(2). This matching is a critical lever to secure the high-intensity compute needed for frontier models without relying on non-EU providers.
  • Monitor EuroHPC Policy: The EuroHPC JU access policy will be adapted to prioritize these projects. Stay informed about changes to these policies, as they will dictate the practical availability of the Union-matched capacity.

Common misconceptions

"CADA mandates that all frontier AI projects must run exclusively on AI gigafactories." Reality: CADA does not mandate exclusive use of gigafactories. Article 7(2)(e) requires Member States to invest in gigafactories as part of their national strategies, but Article 9 sources compute from a combination of Member State contributions (which may include gigafactories) and Union-matched EuroHPC capacity. The infrastructure is a composite ecosystem, not a single mandated facility.

"The Union provides unlimited compute to any project labeled 'frontier'." Reality: Support is strictly conditional. First, the project must be formally recognized as a "frontier AI priority project" by the Commission under Article 8. Second, Article 9(2) explicitly limits the Union's matching to "the extent that sufficient AI computing capacity is available within the Union's share of European high performance computing access time." There is no guarantee of unlimited resources; availability is capped by actual EuroHPC capacity.

"AI gigafactories are a new legal entity created by CADA." Reality: CADA does not create a new legal entity for gigafactories. Instead, it sets a policy target for Member States to include them in their national strategies under Article 7(2)(e). The infrastructure will be built and operated by a mix of public and private entities, potentially integrated into existing EuroHPC networks or new national initiatives.

Related

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