Summary Under the proposed Cloud and AI Development Act (CADA), startups can access computing resources primarily through the Cloud and AI Leadership Initiatives, specifically by engaging with the network of Experience and Acceleration Centres for AI (Centres for AI). While the proposal does not grant automatic, individual rights to free compute, it establishes a structured ecosystem where the Union and Member States endeavour to provide sufficient computing resources for AI industrial innovation, physical AI, and public sector projects. Startups should align their projects with national strategies that invest in AI factories and AI gigafactories, and leverage the Centres for AI network to access infrastructure for model fine-tuning and compute where it is not available locally.

Detail

The proposed Cloud and AI Development Act (CADA), COM(2026) 502 final, establishes the Cloud and AI Leadership Initiatives as the central framework for scaling the EU's cloud and AI ecosystem. For startups, particularly small and medium-sized enterprises (SMEs) and small mid-caps (SMCs), the Act creates a multi-layered pathway to access the high-intensity computing resources necessary for training and deploying advanced AI models. This access is not a direct "free tier" but a strategic alignment mechanism involving national infrastructure, regional hubs, and Union-level resource matching.

The Role of Centres for AI in Compute Access

The most immediate point of contact for a startup is the network of Experience and Acceleration Centres for AI (Centres for AI). These centres, which build on the existing European Digital Innovation Hubs, are mandated to act as regional accelerators for AI adoption.

Under Article 5(2)(c) of the proposal, a core objective of the Centres for AI is to "leverage relevant infrastructure to accelerate the development and fine-tuning of AI models and systems." This provision is critical for startups that lack the capital to build proprietary data centres or access high-performance computing (HPC) clusters. By partnering with a Centre for AI, a startup can gain access to the computational resources required to train, fine-tune, and validate their AI solutions.

Furthermore, Article 5(6) establishes a network of these Centres to support collaboration and the exchange of best practices. Crucially, the text specifies that this network is designed to "provide specialised services across regions where the required skills or compute capacity are not available locally." This means that if a startup is located in a Member State or region with limited local infrastructure, the CADA framework envisions a mechanism to bridge that gap. The network allows for the routing of compute-intensive tasks to Centres in other regions that possess the necessary capacity, ensuring that geographic location does not become a barrier to innovation.

Union and Member State Commitments to Provide Compute

Beyond the regional Centres, CADA places specific obligations on the Union and Member States to ensure the availability of compute for strategic sectors. Article 9(3) explicitly states that "The Union and the Member States shall endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects."

For a startup developing industrial AI (e.g., in manufacturing, automotive, or energy) or physical AI (e.g., robotics, autonomous drones), this clause suggests a prioritised pathway for resource allocation. While the term "endeavour" indicates a strong political commitment rather than an absolute guarantee, it creates a legal basis for startups to seek support when their projects align with these strategic categories. The provision ensures that the Union and Member States actively work to match the demand for compute in these high-value areas.

This commitment is reinforced by the requirement for Member States to adopt national cloud and AI strategies under Article 7. Specifically, Article 7(2)(e) mandates that these national strategies must include "measures to invest in high-intensity computing infrastructure, including AI factories, AI gigafactories and quantum computers as strategic national and cross-border assets."

Startups can leverage these national strategies to identify where compute investments are being made. By aligning their projects with the specific "AI factories" or "gigafactories" designated in their Member State's strategy, startups can position themselves to access these high-intensity assets. These facilities are intended to serve as strategic national assets, and the Act envisions that they will support the broader ecosystem, including the development and deployment of AI models for the public sector and strategic industries.

Frontier AI and Strategic Project Matching

For startups working on cutting-edge technologies, the proposal offers a specific mechanism for resource matching through frontier AI priority projects. Article 8 outlines the criteria for recognising projects as "frontier AI priority projects," which must be pioneering efforts focused on scaling up frontier AI technologies and involve broad participation across the Union.

If a startup's project qualifies under these criteria, Article 9(2) provides a powerful incentive: "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 creates a potential "match-funding" mechanism for compute time. If a Member State commits resources to a startup's frontier AI project, the Union is obligated to match that commitment with its own HPC capacity. This significantly reduces the cost barrier for deep-tech startups working on frontier models, provided they can secure the initial commitment from their national authorities and meet the cross-border participation criteria.

Operational Objectives and Strategic Sectors

The Cloud and AI Leadership Initiatives are further detailed in Article 4, which lists eight operational objectives. For startups, Operational Objective 5 (accelerating industrial AI) and Operational Objective 4 (advancing physical AI) are particularly relevant. The Initiatives are designed to "facilitate access to the necessary computing resources and AI tools required to develop and operationalise AI models and systems tailored to industrial sector needs."

This confirms that the framework is not merely about building infrastructure but about actively facilitating access to it for specific use cases. Startups developing solutions for strategic sectors such as healthcare, transport, manufacturing, and defence are explicitly targeted by these objectives, which aim to "accelerate the development and uptake of sectoral AI models and systems across the Union's strategic industrial sectors."

What this means for you

For CTOs, founders, and architects in startups, the proposed CADA framework shifts compute access from a purely commercial transaction to a strategically supported ecosystem. Here is how you should prepare to leverage these provisions:

  1. Engage with Your Local Centre for AI: Identify the Experience and Acceleration Centre for AI in your Member State immediately. As mandated by Article 5, these centres are your primary entry point for accessing infrastructure for model fine-tuning. Prepare technical documentation that demonstrates how your AI model aligns with the Centre's mission to support strategic industrial or public sector use cases. Ask specifically about their capacity to leverage the network for compute where local resources are scarce.
  2. Align with National Strategies: Review your Member State's national cloud and AI strategy, which is required under Article 7. Look for specific mentions of AI factories, AI gigafactories, and quantum computers. If your startup's technology complements these national investments, position your compute requests as part of a broader national innovation effort. This alignment is essential to unlock access to the high-intensity infrastructure described in Article 7(2)(e).
  3. Target Strategic Classifications: If your startup is developing industrial AI, physical AI, or frontier AI, ensure your project documentation explicitly highlights these classifications. Under Article 9(3), these categories are the specific areas where the Union and Member States shall endeavour to provide sufficient computing resources. This alignment can improve your chances of receiving support or prioritised access to EuroHPC resources.
  4. Monitor for Priority Project Calls: Keep an eye on open calls for expression of interest for frontier AI priority projects under Article 8. If your startup qualifies as a pioneering project and can demonstrate broad Union participation, the potential for matched compute resources from the Union under Article 9(2) can significantly accelerate your R&D phase.
  5. Leverage the Network Effect: Do not assume your local region has the necessary capacity. Utilise the network established under Article 5(6) to access specialised services and compute capacity in other regions where it is available. The Act explicitly envisions this cross-regional support to ensure no startup is left behind due to local infrastructure gaps.

Common misconceptions

Misconception 1: CADA guarantees free compute for all startups. CADA does not guarantee free or unlimited compute access for every startup. Instead, it establishes a framework where access is tied to strategic objectives, such as AI industrial innovation, frontier AI, or public sector adoption. Access is often facilitated through Centres for AI or national strategies, and may involve competitive selection, partnership requirements, or alignment with specific national priorities.

Misconception 2: Startups can directly apply to the Commission for compute. The proposal does not create a direct application portal for individual startups to request compute from the European Commission. Instead, access is mediated through Member State national strategies, the network of Centres for AI, and specific calls for frontier AI priority projects. The Union's role is to match resources for designated projects and ensure Member States invest in infrastructure, rather than acting as a direct compute provider for individual SMEs.

Misconception 3: Only large corporations can benefit from AI factories. While AI factories and AI gigafactories are large-scale assets, Article 7(2)(e) and Article 5(2)(c) indicate that these infrastructures are intended to support the broader ecosystem, including SMEs and startups, through the Centres for AI. The goal is to democratise access to high-performance compute, not to restrict it to incumbents. The Centres act as the bridge, allowing smaller entities to utilise these massive resources for fine-tuning and development.

Misconception 4: The Union will provide compute regardless of project type. The Union's commitment to provide resources is specific. Article 9(3) limits the "endeavour to provide sufficient computing resource" to AI industrial innovation, physical AI, and public sector AI projects. Startups working on non-strategic or purely consumer-facing applications may not fall within the scope of this specific resource allocation mandate, though they may still benefit from general market measures.

Related

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