Summary Under the proposed Cloud and AI Development Act (CADA), Article 9(3) establishes a non-binding obligation for the Union and Member States to "endeavour" to provide sufficient computing resources for AI industrial innovation, physical AI, and public sector AI projects. Unlike "frontier AI priority projects" (which benefit from a strict matching mechanism under Article 9(1) and 9(2)), industrial AI projects rely on this softer, best-effort commitment to access high-performance computing capacity. This provision is designed to support the strategic deployment of AI across key European industriesβ€”such as automotive, healthcare, and manufacturingβ€”by ensuring that compute availability does not become a bottleneck for industrial digital transformation, while explicitly linking to Grand Challenge 5 in Annex I.

Detail

The Cloud and AI Development Act (CADA), as proposed in COM(2026) 502 final, establishes a comprehensive framework to strengthen Europe's cloud and AI ecosystem. A central pillar of this framework is the allocation of high-performance computing (HPC) resources to support strategic AI development. While Article 9(1) and 9(2) provide robust, matching-fund mechanisms for "frontier AI priority projects," Article 9(3) addresses a broader, yet equally critical, category of strategic initiatives: AI industrial innovation, physical AI, and public sector AI.

The "Endeavour" Obligation in Article 9(3)

The text of Article 9(3) is precise in its legal phrasing, distinguishing it from the mandatory matching obligations found in the preceding paragraphs. It states:

"The Union and the Member States shall endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects."

This use of the term "endeavour" creates a best-effort obligation rather than a strict, enforceable entitlement to specific compute quotas. In the context of EU legislative drafting, this signals a high political priority for these sectors while acknowledging the practical reality that compute capacity is a finite resource. The Union and Member States are committed to prioritizing these projects, but the provision does not guarantee that every application will be met with a specific allocation of EuroHPC time in the same way frontier AI projects are.

This distinction is crucial for providers and applicants. While frontier AI priority projects (designated under Article 8) trigger a mechanism where the Union matches Member State contributions "within the limits of available capacity" (Article 9(2)), industrial AI projects under Article 9(3) rely on the general commitment of the Union and Member States to coordinate and provide sufficient resources. This approach ensures that while frontier AI receives guaranteed support to maintain global competitiveness, the broader industrial and public sector AI ecosystems are not left without access to the computational power required for training and inference.

Link to Grand Challenge 5: Industrial AI

The reference to "AI industrial innovation" in Article 9(3) is not generic; it is directly tied to the operational objectives of the Cloud and AI Leadership Initiatives, specifically Grand Challenge 5: Industrial AI, as outlined in Annex I of the CADA proposal.

Grand Challenge 5 focuses on accelerating the development and deployment of European industrial AI across the Union's strategic sectors. The proposal identifies specific areas where industrial AI is critical and where compute support is essential to bridge the gap between research and large-scale deployment:

  • Healthcare: Improving clinical decision accuracy and transforming pharmaceutical processes.
  • Automotive: Supporting the development of software-defined vehicles and autonomous driving systems.
  • Manufacturing: Enabling specialized models that optimize production processes.
  • Defence and Space: Enhancing security capabilities and transforming how space assets are operated.
  • Climate and Environment: Leveraging Earth observation data for geospatial AI.
  • Agri-food: Using AI to optimize irrigation, fertiliser use, and yield forecasting.

The proposal explicitly states that initiatives launched under this grand challenge "should rely on specialised computing resources and testing facilities necessary to validate AI systems in real-world environments before supporting their large-scale deployment and uptake." By linking compute support to these specific industrial applications, Article 9(3) aims to ensure that the "grand challenges" have the necessary infrastructure to succeed.

Strategic Sectors and the Scope of Support

Article 9(3) explicitly groups three distinct categories of projects under the "endeavour" obligation:

  1. AI Industrial Innovation: As detailed above, this covers the strategic sectors listed in Grand Challenge 5. The focus is on sector-specific AI models designed to meet operational requirements, such as those in healthcare, transport, manufacturing, defence, space, climate, and agri-food.
  2. Physical AI: This refers to AI systems and models capable of perceiving the physical environment and executing complex actions within that environment, such as robotics, autonomous drones, and self-driving vehicles. The proposal notes that physical AI requires a "dedicated approach to data and computing infrastructure," necessitating access to high-quality data and significant compute resources for training and validation in diverse real-world environments.
  3. Public Sector AI: This supports the development of AI models and systems for critical public domains such as healthcare, public administration, law, and crisis management. The goal is to improve public service delivery, simplify administrative procedures, and enhance decision-making. The proposal encourages the sharing and reuse of training data and AI models across the Union's public services to avoid fragmentation and enable the scaling of successful solutions.

Role of EuroHPC and Member State Cooperation

The provision of compute resources under Article 9(3) is expected to leverage existing European High Performance Computing (EuroHPC) infrastructure. While the text does not mandate a specific quota for industrial AI, it implies that Member States and the Union should coordinate to ensure that these projects have access to sufficient compute time.

This coordination is part of the broader Cloud and AI Leadership Initiatives (Title II), which aim to integrate networks, cloud, AI, and software into coherent ecosystems. The proposal emphasizes that the Union and Member States shall "endeavour to provide sufficient computing resource," suggesting a collaborative approach where national contributions are pooled and aligned with Union-level capacity.

Furthermore, the proposal highlights the importance of the Centres for AI (Experience and Acceleration Centres for AI) in facilitating access to these resources. Under Article 5, these centres are tasked with helping organizations accelerate their digital transformation by connecting them with European providers of cloud and AI technologies, including access to compute capacity. This is particularly relevant for SMEs and SMCs (small mid-caps) that may lack the resources to navigate complex HPC allocation processes independently.

What this means for you

For cloud service providers, data centre operators, and industrial AI developers, Article 9(3) and the associated Grand Challenge 5 have several strategic implications:

  1. Demand for Specialized Compute: Expect increased demand for compute resources tailored to industrial and physical AI workloads. These may require specific configurations, low-latency access, or integration with edge computing capabilities to support real-world testing and validation in sectors like automotive and robotics.
  2. Partnership Opportunities: The proposal encourages collaboration between cloud providers, industrial actors, and public sector bodies. Providers may find opportunities to partner with Centres for AI or participate in joint initiatives that demonstrate the feasibility of sovereign cloud stacks for industrial applications.
  3. Sovereign Cloud Recognition: As public sector bodies are required to procure cloud services with specific Union assurance levels under Article 30, providers who achieve recognition under the CADA sovereignty framework (Article 17) will be better positioned to serve public sector AI projects. This is particularly relevant for projects handling sensitive data in healthcare, defence, or administration.
  4. Data Sovereignty and Security: Industrial AI projects often involve commercially sensitive or operationally critical data. Providers must ensure their services meet the strict data localization and security requirements outlined in the CADA sovereignty framework, particularly for Union assurance levels 2, 3, and 4 (Annex II).
  5. Support for SMEs: The proposal emphasizes supporting SMEs and SMCs in accessing AI technologies. Cloud providers can differentiate themselves by offering user-friendly interfaces, pre-configured industrial AI models, and simplified access to HPC resources through the Centres for AI network.

Common misconceptions

  • "Article 9(3) guarantees compute for all industrial AI projects." This is incorrect. Article 9(3) uses the term "endeavour," which is a best-effort obligation. It does not create a legally enforceable right to specific amounts of compute for every industrial AI project. Priority may be given to projects that align closely with the Grand Challenges or demonstrate significant strategic value, but there is no automatic matching mechanism like that for frontier AI.
  • "Industrial AI projects are treated the same as frontier AI projects." No. Frontier AI priority projects (designated under Article 8 and supported by Article 9(1)-(2)) benefit from a matching mechanism where the Union matches Member State contributions. Industrial AI projects under Article 9(3) do not have this guaranteed matching structure. They rely on the general commitment of the Union and Member States to provide sufficient resources.
  • "Only large corporations can benefit from Article 9(3)." While large industries are key targets, the proposal explicitly mentions supporting SMEs and SMCs. The Centres for AI are designed to help smaller organizations access these resources and expertise. Cloud providers are encouraged to facilitate this access to ensure a diverse and competitive ecosystem.

Related

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