Summary As proposed, the Cloud and AI Development Act (CADA) does not legally mandate that frontier AI compute projects must exclusively use semiconductors produced under the Chips Act. Instead, CADA establishes a funding and coordination framework (Articles 8 and 9) that explicitly seeks synergies with the Chips Act 2.0 and the EuroHPC Joint Undertaking. The proposal aims to align hardware supply chains with AI compute demand, creating a practical dependency where CADA-funded projects are expected to leverage European semiconductor capabilities to strengthen technological sovereignty. While no strict procurement obligation exists for the AI model layer, the ecosystem is designed so that "frontier AI priority projects" align with the Chips Act's grand challenges to access Union compute resources.

Detail

The relationship between CADA's frontier AI initiatives and the EU's semiconductor strategy is one of strategic alignment rather than direct regulatory compulsion. While CADA does not impose a strict procurement obligation requiring frontier AI projects to source exclusively from Chips Act beneficiaries, it creates a structural ecosystem where the two instruments are designed to reinforce each other. The proposal treats AI compute capacity and semiconductor manufacturing as two halves of the same strategic equation, linking the demand side (AI models) with the supply side (chips).

Frontier AI Priority Projects and Compute Allocation

Under Article 8, the Commission is empowered to recognize specific initiatives as "frontier AI priority projects." To qualify, a project must meet strict criteria: it must be a pioneering project focused on the support and scaling-up of frontier AI technologies; it must be undertaken by a European digital infrastructure consortium (EDIC) or another eligible legal entity; and it must involve the participation of at least three Member States. Crucially, the participating Member States must pool computing time and other relevant resources to support the implementation of the designated project.

Article 9 then dictates how compute resources are allocated to these recognized projects. The Union and Member States are required to ensure that sufficient AI computing resources from their available 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." The Union commits to matching the AI computing resources contributed by Member States to these projects "to the extent that sufficient AI computing capacity is available within the Union's share of European high performance computing access time."

This mechanism creates a centralized pool of compute resources dedicated to strategic AI development. While Article 9 does not explicitly specify the hardware architecture (e.g., requiring a specific chip architecture), the nature of "frontier AI" implies a heavy reliance on high-performance, energy-efficient accelerators and processors. The availability of these resources is directly linked to the broader EU industrial base, which includes the semiconductor sector. The text explicitly links the allocation of these resources to the EuroHPC JU access policy, which is governed by Regulation (EU) 2021/1173.

Strategic Synergies with the Chips Act 2.0

The proposal explicitly links its operational objectives to the semiconductor supply chain through its recitals and operational objectives. Recital 31 states that the Cloud and AI Leadership Initiatives should enhance synergies with actions currently supported by the Union, including "Regulation (EU) 2026/XXX [Chips Act 2.0] on a framework of measures for strengthening Europe's semiconductor ecosystem." The recital notes that these initiatives should build on the "grand challenges" established under the Chips Act 2.0, which are aimed at enabling semiconductor technologies underpinning AI, cloud computing, data centres, and edge infrastructures.

This recital is critical for understanding the practical interplay. It signals that the EU views AI compute capacity and semiconductor manufacturing as interdependent. The Chips Act 2.0 is designed to boost the production and design of advanced semiconductors in Europe, while CADA is designed to create the demand and application layer for those chips. By referencing the Chips Act 2.0's grand challenges, CADA ensures that the AI projects it funds are aligned with the technological roadmap of the semiconductor industry.

Furthermore, Recital 14 highlights the importance of developing AI-optimized servers and baseline software based on processors, accelerators, and quantum accelerators "designed and manufactured in the Union." It explicitly mentions that these should be developed in line with the Chips Act 2.0. This creates a policy expectation that CADA-supported projects will prioritize or actively collaborate with European hardware developers. Article 4(2)(b) reinforces this by setting an operational objective to "develop AI-optimised servers and baseline software based on processors, accelerators and quantum accelerators designed and manufactured in the Union."

The Role of EuroHPC and Joint Undertakings

The proposal also leverages existing structures to bridge the gap between AI software/models and hardware. Recital 26 and Recital 31 reference the European High Performance Computing Joint Undertaking (EuroHPC JU), established under Regulation (EU) 2021/1173. The EuroHPC JU is a key vehicle for acquiring and deploying supercomputing capacity in Europe, much of which relies on advanced semiconductor components.

CADA's framework for frontier AI priority projects is designed to integrate with the EuroHPC access policy. Recital 35 notes that the EuroHPC JU access policy should be accommodated to reflect the allocation of computing resources for frontier AI projects. Since EuroHPC systems are increasingly built with a mix of European and non-European hardware, and the Chips Act aims to increase the share of European components, the synergy is clear: CADA drives the workload, EuroHPC provides the infrastructure, and the Chips Act supports the underlying hardware supply chain. The proposal explicitly states that the Union and Member States shall provide sufficient compute time for AI industrial innovation, physical AI and public sector AI projects, "without prejudice to the rules and procedures laid down in Council Regulation (EU) 2021/1173."

Funding and Implementation Mechanisms

The financial architecture of CADA further cements this link. The proposal indicates that the Cloud and AI Leadership Initiatives may be supported by funding from Union programmes, including Horizon Europe and the Digital Europe Programme, as well as the InvestEU Programme. Recital 28 and Recital 31 emphasize the need for consistency and complementarity between CADA initiatives and national or regional strategies, including those implemented under national reform programmes and the Chips Act.

While CADA does not create a new fund specifically for semiconductor procurement, it directs existing funds towards projects that must demonstrate strategic relevance. A project seeking recognition as a frontier AI priority project under Article 8 would likely need to demonstrate how it contributes to the Union's technological sovereignty. Given the explicit policy direction in Recital 31, using European-designed or manufactured semiconductors would be a significant factor in demonstrating this contribution, even if not a strict legal prerequisite in the text of the regulation itself. The proposal aims to "foster the co-design and cross-optimisation of hardware and software development and the integration of AI computing infrastructures," directly linking the AI stack to the semiconductor stack.

What this means for you

For CTOs, architects, and SMEs evaluating the practical impact of CADA, the distinction between legal mandate and strategic expectation is vital.

  1. Hardware Agnosticism with Strategic Bias: You are not legally barred from using non-European semiconductors in your AI models under CADA. However, if you are applying for recognition as a "frontier AI priority project" under Article 8, your proposal will be evaluated in the context of the EU's broader industrial strategy. Demonstrating a supply chain that includes European semiconductors (supported by the Chips Act) will likely strengthen your application's alignment with the "technological autonomy" goals of the Regulation. The operational objectives in Article 4 explicitly target the development of AI-optimised servers based on Union-manufactured processors.
  2. Access to Compute Resources: Article 9 guarantees access to AI computing resources for recognized projects. These resources are drawn from EuroHPC capacity. As the EuroHPC infrastructure evolves, it will increasingly incorporate chips developed under the Chips Act 2.0 roadmap. Architects designing AI systems should prepare for a future where the available "sovereign" compute infrastructure is tightly coupled with European hardware stacks. The proposal explicitly links the allocation of resources to the EuroHPC JU access policy.
  3. Collaboration Opportunities: The proposal encourages the formation of European digital infrastructure consortiums (EDICs) for frontier AI projects. These consortiums are expected to include stakeholders across the value chain, including semiconductor manufacturers. Article 8 requires that frontier AI priority projects involve participation from at least three Member States and be undertaken by an EDIC or eligible entity. SMEs specializing in AI applications should consider partnering with hardware providers or semiconductor firms to form cohesive bids that address the full stack, from chip to cloud.
  4. Long-Term Planning: The synergies with the Chips Act 2.0 suggest a long-term shift in the European AI ecosystem. Projects that rely solely on third-country hardware may face increasing pressure to diversify or migrate to sovereign alternatives, especially if they depend on public funding or public sector procurement. Early integration of European semiconductor solutions can future-proof your AI infrastructure against potential supply chain disruptions or geopolitical risks. The proposal aims to "foster the development of cloud computing stacks alternatives for strategic sectors" and "develop AI-optimised servers... designed and manufactured in the Union."

Common misconceptions

  • Misconception: CADA bans non-European chips for AI.
    • Reality: CADA does not prohibit the use of non-European semiconductors. It establishes a framework for supporting European capabilities. While public procurement rules (Article 30) may require certain assurance levels for cloud services, there is no explicit ban on the underlying hardware in the AI model development phase. The preference for European chips is driven by strategic funding alignment and operational objectives, not a legal prohibition.
  • Misconception: Frontier AI projects must be built on Chips Act-funded factories.
    • Reality: The Chips Act supports the design and manufacturing of semiconductors, but it does not mandate that all AI projects use chips from specific "Chips Act" facilities. The link is through the broader ecosystem and the strategic goals of the Cloud and AI Leadership Initiatives. Projects are encouraged to use technologies developed under the Chips Act's grand challenges, as noted in Recital 31, but this is a matter of strategic alignment and eligibility for certain types of support, not a rigid supply chain mandate.
  • Misconception: CADA replaces the Chips Act.
    • Reality: CADA and the Chips Act are complementary. The Chips Act focuses on the supply side (semiconductor production and design), while CADA focuses on the demand and application side (AI models, cloud infrastructure, and compute capacity). They are designed to work together, as explicitly stated in Recital 31. CADA does not duplicate the Chips Act's measures but rather creates the market and infrastructure context in which those semiconductors can be utilized effectively.

Official sources

Related

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