Does the Union give cash grants or compute time to AI projects under CADA?

Summary Under the proposed Cloud and AI Development Act (CADA), the Union does not provide direct cash grants to AI projects through the Regulation itself. Instead, Article 9 commits the Union and Member States to allocate in-kind AI computing resources (compute time) to designated "frontier AI priority projects." Direct financial funding for the development, deployment, and operational costs of these initiatives flows through separate, pre-existing EU programmes—specifically Horizon Europe, the Digital Europe Programme, and InvestEU—rather than through CADA's legal framework. CADA acts as a regulatory mechanism to guarantee access to infrastructure, while other instruments provide the liquidity.

Detail

The Cloud and AI Development Act (CADA) establishes a clear separation between the provision of computational capacity and the provision of financial capital. For technical leaders, CTOs, and project architects evaluating how to resource large-scale AI initiatives, it is critical to understand that CADA's primary mechanism for supporting frontier AI is the allocation of compute, not cash.

Article 9: The Commitment to Compute, Not Cash

The core of this distinction lies in Article 9 of the CADA proposal, titled "Computing support for AI projects." This article establishes a mandatory framework for the allocation of AI computing resources but explicitly avoids establishing a grant mechanism for cash disbursements.

Article 9(1) states: "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."

This provision creates a binding obligation to provide access to infrastructure—specifically, the processing power required to train and run large AI models. It does not mention financial transfers, grants, subsidies, or direct budgetary allocations to the project beneficiaries. The focus is squarely on removing the bottleneck of hardware scarcity by guaranteeing access to existing or newly built high-performance computing (HPC) and AI-optimized infrastructure.

Article 9(2) further clarifies the Union's role in this compute allocation: "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 "matching" mechanism reinforces the in-kind nature of the support. The Union is committing to provide compute time equivalent to what Member States contribute, ensuring that strategic projects have the necessary scale. This is a resource-sharing arrangement based on the European High Performance Computing Joint Undertaking (EuroHPC JU) access policies, not a funding agreement.

Article 9(3) extends this logic to broader categories of AI development, stating that the Union and Member States "shall endeavour to provide sufficient computing resource for AI industrial innovation, physical AI and public sector AI projects." Again, the obligation is to provide "computing resource," not financial capital.

Where Does the Cash Come From?

If CADA does not provide cash, how are these projects funded? The CADA proposal explicitly links its operational objectives to existing and future EU funding programmes. The financial support for the research, development, and deployment of AI technologies flows through separate legislative instruments.

Recital 28 of the CADA proposal notes: "The Cloud and AI Leadership Initiatives may be supported by funding from Union programmes, in particular from Horizon Europe and the digital Europe programme, as well as the InvestEU programme."

This means that while CADA provides the regulatory framework to allocate compute time (via Article 9), the cash grants, subsidies, and financial instruments required to build the infrastructure, hire staff, cover operational costs, and fund R&D come from:

Horizon Europe: Supports upstream research and innovation activities, including the development of frontier AI models and technologies.
Digital Europe Programme: Supports the deployment and uptake of digital technologies, including the construction of AI factories, high-performance computing infrastructure, and the training of AI models.
InvestEU Programme: Provides financial guarantees and equity investments to leverage private capital for digital infrastructure projects.

Additionally, Recital 29 mentions that initiatives may be supported by Member States through "research, development and innovation measures, in line with the applicable State aid rules," as well as private-sector investments. The CADA proposal itself does not create a new budget line for cash grants; it relies on the financial architecture of the Multiannual Financial Framework (MFF) and the specific regulations governing the programmes listed above.

The Distinction: In-Kind Compute vs. Grant Funding

For SMEs, large enterprises, and research consortia, the distinction between in-kind compute and grant funding is significant and dictates the strategy for project resourcing:

In-Kind Compute (CADA Article 9): This is a non-monetary benefit. It reduces the capital expenditure (CapEx) and operational expenditure (OpEx) associated with renting or building data centers. It provides access to specialized hardware (e.g., GPU clusters, AI-optimized servers) that might otherwise be inaccessible or prohibitively expensive due to global shortages. However, it does not provide liquidity for salaries, software licenses, data acquisition, or other overheads. It is a "resource" allocation, not a "funding" allocation.
Grant Funding (Digital Europe/Horizon Europe): This is direct financial support. It provides cash that can be used for any project-related expense, including personnel, hardware procurement, travel, and administrative costs. Grants often require co-financing from the beneficiary (typically 20-50% depending on the programme and beneficiary type) and are subject to strict reporting, eligibility, and audit criteria.

Under CADA, a "frontier AI priority project" (as defined in Article 8) must meet specific criteria, such as being a pioneering project focused on scaling frontier AI technologies, involving participation from at least three Member States, and pooling computing time. Once recognized by the Commission, these projects gain access to the compute resources mandated by Article 9. However, the project must still secure separate funding from the aforementioned programmes to cover the rest of its budget. The two streams are complementary but distinct: one provides the engine (compute), the other provides the fuel (cash).

What this means for you

For CTOs, project managers, and architects, this structure implies a two-track strategy for securing resources for large-scale AI projects:

Apply for Compute Access via CADA Criteria: To benefit from Article 9, your project must qualify as a "frontier AI priority project" under Article 8. This requires demonstrating that the project is pioneering, involves multiple Member States (at least three), and pools computing time and resources. You should engage with national authorities and European Digital Infrastructure Consortia (EDICs) to position your project for this recognition. The goal here is to secure guaranteed access to high-performance compute, which is a critical bottleneck for training large models. This is a regulatory pathway, not a grant application.
Secure Cash Funding Separately: Do not expect CADA to write a check. You must apply for financial support through the Digital Europe Programme, Horizon Europe, or national state aid schemes. Your project proposal should clearly distinguish between the compute resources you are seeking (potentially via CADA's framework) and the financial resources you need (via other programmes). The "matching" provision in Article 9(2) is particularly relevant for consortia: if you are part of a consortium that includes Member State contributions, the Union will match those compute resources. However, SMEs should still prioritize applying for grants from the Digital Europe Programme, which has specific provisions and funding windows for AI deployment and innovation.

Common misconceptions

"CADA provides direct grants for AI development." This is incorrect. CADA is a regulatory framework that mandates the allocation of compute resources and sets sovereignty standards. Financial grants are administered under separate regulations like the Digital Europe Regulation and the Horizon Europe Regulation.
"Article 9 guarantees unlimited compute." No. Article 9(1) explicitly states that allocation is "within the limits of available capacity." Compute is a finite resource, and access will be competitive and prioritized for projects meeting the strict criteria of Article 8.
"Only cash funding matters for AI projects." While cash is essential for operations, compute is the primary bottleneck for frontier AI. Securing in-kind compute through CADA's framework can be as valuable as cash, as it removes the need to invest billions in hardware infrastructure and ensures access to state-of-the-art resources.
"CADA replaces the Digital Europe Programme." No. CADA complements these programmes. The Digital Europe Programme provides the money to build infrastructure and fund projects; CADA provides the legal obligation to ensure that the compute capacity funded or built in the EU is accessible to strategic projects and that the Union matches Member State contributions.

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