Summary No, frontier AI priority projects are not the same as AI factories. Under the proposed Cloud and AI Development Act (CADA), frontier AI priority projects are specific research and innovation initiatives focused on scaling advanced AI models, governed by Article 8. In contrast, AI factories (and AI gigafactories) are the physical high-intensity computing infrastructure required to train and run those models, referenced in Article 7(2)(e) as strategic assets. While frontier AI priority projects would utilise compute resources from AI factories, the Act treats them as distinct legal categories: one is a collaborative R&D programme, the other is a data-centre facility.

Detail

To understand the operational landscape of the proposed CADA, it is essential to distinguish between the strategic research initiatives and the physical infrastructure that supports them. The proposal creates a clear framework separating the "what" (the research goals) from the "where" (the compute capacity).

What are Frontier AI Priority Projects?

Frontier AI priority projects are defined as specific, high-level research and innovation initiatives designed to push the boundaries of AI capabilities. Article 8 of the CADA proposal establishes the strict criteria for a project to be recognised by the Commission as a "frontier AI priority project."

These projects are explicitly linked to Grand Challenge 3 of the Cloud and AI Leadership Initiatives (set out in Annex I), which focuses on "Developing the next generation of multimodal frontier AI models and systems." The goal is to support pioneering projects that develop frontier AI models as strategic assets, particularly in key sectors such as cybersecurity.

To qualify under Article 8, a project must meet four cumulative criteria:

  1. Pioneering Nature: It must be focused on the support and scaling-up of frontier AI technologies.
  2. Legal Structure: It 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.
  3. Cross-Border Participation: It must involve the participation of at least three Member States.
  4. Resource Pooling: The participating Member States must pool computing time and other relevant resources to support the implementation of the designated project.

These projects represent the "what"β€”the specific technological objectives, the governance of the research consortium, and the scientific goals (e.g., advanced reasoning, cross-modal understanding, and agentic capabilities).

What are AI Factories?

AI factories (and AI gigafactories) refer to the physical and digital infrastructure itself. They are large-scale data centres equipped with high-performance computing (HPC) capacity, specifically optimised for AI workloads. In the context of CADA, these facilities fall under the data centre capacity framework outlined in Title III of the proposal.

The legal basis for these facilities is found in Article 7(2)(e), which mandates that Member States include in their national cloud and AI strategies "measures to invest in high-intensity computing infrastructure, including AI factories, AI gigafactories and quantum computers as strategic national and cross-border assets."

These facilities represent the "where"β€”the hardware, energy, network connectivity, and physical data-centre assets that provide the necessary computational power (measured in FLOPs) to train and deploy AI models. They are the strategic assets that Member States must plan for and deploy to ensure the Union has sufficient capacity.

The Relationship: Projects Use Factories

The distinction is crucial because frontier AI priority projects cannot exist without the underlying compute capacity. Article 9 of the CADA proposal bridges this gap by stating that the Union and Member States shall ensure that sufficient AI computing resources from their compute capacities are allocated to support the development of frontier AI priority projects.

In practice, the relationship functions as follows:

  • The Project (Article 8): A consortium of three or more Member States applies to the Commission to be recognised as a frontier AI priority project. They define the scientific scope, the consortium partners, and the governance.
  • The Infrastructure (Article 7): Member States develop their national strategies to build or designate AI factories and gigafactories as strategic assets.
  • The Allocation (Article 9): Once a project is recognised, the Union and Member States match the AI computing resources contributed by the Member States to that project. The project then utilises the compute time from these AI factories or gigafactories.

Thus, an AI factory is the facility; a frontier AI priority project is the governance and funding structure that uses the facility. The Act does not merge them; it creates a pipeline where the infrastructure (Title III) supports the initiatives (Title II).

What this means for you

For public-sector authorities, national strategy planners, and potential consortium leaders, understanding this distinction is vital for compliance with the proposed CADA framework.

1. Strategic Planning and National Strategies

When developing or updating your national cloud and AI strategy (required by Article 7), you must address infrastructure and R&D as separate but interdependent tracks.

  • Infrastructure Track: You must identify where your AI factories will be located, how they will connect to the grid, and what sustainability standards they will meet (as per Article 7(2)(d) and Title III).
  • R&D Track: You must identify which national entities will lead or participate in frontier AI priority projects (as per Article 7(2)(f)). These are two different budget lines and policy objectives. You cannot simply "build a factory" to satisfy the requirement for a "frontier AI priority project"; you must also establish the cross-border consortium and research agenda.

2. Procurement of Compute vs. Procurement of R&D Services

The procurement processes differ significantly based on the legal nature of the activity:

  • Procuring for a Frontier AI Priority Project: You are likely procuring research services, consortium management, and specific AI model development outcomes. This falls under the Cloud and AI Leadership Initiatives.
  • Procuring for an AI Factory: You are procuring data centre construction, hardware (GPUs, TPUs), energy contracts, and network connectivity. This falls under the data centre deployment framework. Furthermore, the procurement of cloud computing services (which might host the outputs of an AI factory) is subject to the Union assurance levels (Article 16) and risk assessments (Article 29). However, the procurement of the physical data centre infrastructure itself is governed by the acceleration zone and strategic project rules in Title III.

3. Resource Allocation and Matching

Article 9 specifies that the Union will match AI computing resources contributed by Member States to frontier AI priority projects. As a public authority, you need to track how much compute capacity you are contributing from your national AI factories to these priority projects. This requires clear internal accounting between the infrastructure providers (the factories) and the project beneficiaries (the R&D consortia). The Union's matching is conditional on the project being a designated "frontier AI priority project" under Article 8.

4. Eligibility and Consortium Building

If your authority is involved in a frontier AI priority project, ensure you meet the Article 8 criteria, particularly the requirement for participation from at least three Member States. This is a cross-border R&D collaboration, not a unilateral infrastructure build. You will need to coordinate with other Member States to pool resources and define the project scope, which is distinct from the unilateral or bilateral agreements often used for data centre site selection.

Common misconceptions

Misconception 1: "An AI factory is a frontier AI priority project." This is incorrect. An AI factory is a facility (infrastructure). A frontier AI priority project is a research initiative (R&D). You can have an AI factory that hosts commercial AI training, open-source model training, or industrial AI projects that are not designated as "frontier AI priority projects." Conversely, a frontier AI priority project is a governance and funding structure that uses the factory, but is not the factory itself.

Misconception 2: "All AI projects in an AI factory are frontier AI priority projects." No. The designation of "frontier AI priority project" is a specific status granted by the Commission under Article 8 for projects that meet strict criteria, including cross-border participation (three Member States) and a focus on scaling frontier AI technologies. Many other AI initiatives, such as industrial AI applications in manufacturing or healthcare, may use AI factory compute but do not qualify as frontier AI priority projects.

Misconception 3: "AI gigafactories are only for frontier AI priority projects." While AI gigafactories provide the massive scale needed for frontier AI, they are not exclusively reserved for Article 8 projects. They are part of the broader EU compute capacity strategy. The CADA proposal aims to triple EU capacity, and this capacity will serve various needs, including general-purpose AI, industrial AI, and public sector AI, not just the specific frontier AI priority projects.

Misconception 4: "Procuring for a frontier AI priority project means buying servers." Not necessarily. Procuring for a frontier AI priority project often involves buying compute time or access to existing infrastructure, or funding the R&D consortium. Procuring for an AI factory involves buying the physical assets (servers, cooling, power). The procurement processes, legal bases, and sovereignty requirements may differ significantly between buying a service (compute time) and buying goods (hardware/infrastructure).

Official sources

Related

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