From Legislative Rationale to Cloud and AI Leadership

A proposal, not enacted law

This is Proposal 2026/0138 (COD) for the Cloud and AI Development Act, with EEA relevance. The supplied text begins with the formal heading and recitals.

Recitals supply reasons, context, and interpretive direction. They use language such as "should"; this extract does not contain the later articles' detailed binding rules or deadlines.

Status on July 20, 2026

The proposal remains ongoing in the ordinary legislative procedure and is not yet an enacted Regulation. Treat it as a proposal, not as directly applicable law.

Its sequence moves from strategic rationale, to legal bases and definition, then to initiatives for infrastructure, AI, sectoral use, public services, adoption, and skills.

Recital (1): AI as a competitiveness driver

The text links AI to faster innovation, efficiency, smarter decisions, and substantial economic, environmental, and societal gains. It frames AI as fundamental to competitiveness.

The Union should advance innovation while strengthening competitiveness, security of supply, technological sovereignty, and strategic autonomy in cutting-edge digital technologies.

Existing foundations

Recital (2) situates the proposal beside the AI continent action plan, Apply AI Strategy, Data Act, and Data Union Strategy, especially a secure interoperable data market.

The stated dependency problem

Reinforcing the Union's capacity to develop and deploy cloud and AI technologies within its territory has become a strategic priority.

Recitals (4)-(6) associate dependence on limited third-country cloud providers with market-share, autonomy, resilience, security, and extraterritorial-legislation risks.

Two objectives, two legal bases

Recital (7) says the framework pursues separate objectives. Its Treaty bases are "in particular Article 114 and Article 173(3) thereof".

Article 173(3) TFEU

Recital (8) addresses competitiveness, capacity, and resilience of the industrial and technological base. It says these measures should not harmonise national laws or regulations.

Article 114 TFEU

Recital (9) supports a coherent Union framework harmonising certain conditions for cloud-service providers and deployers, including capacity building and service provision.

Cloud-service scope

The definition covers remote, on-demand access to a scalable and elastic pool of shareable computing resources, even across several locations.

Only the delivery and making available of an AI system forms part of the service. The AI system itself and its underlying model are excluded from the scope of this definition.

Scenario

A hospital remotely accesses an AI image-analysis tool. Computing resources can scale with demand and operate across several data-centre locations.

That setup fits the recital's service concept: on-demand administration plus broad remote access to a scalable and elastic pool of shareable computing resources.

Separate the layers

Distinguish: 1. delivery of the remotely hosted service; 2. the AI system; and 3. the underlying model. The recital treats only the first as part of the service.

Do not infer size, user, location, or sectoral thresholds. Recital (10) supplies no such numerical test; it defines the scope boundary between delivery and the system or model.

The headline objective

increase the cloud and data centre capacity of the Union, while advancing cutting-edge cloud and AI technologies together with broad cloud and AI adoption

Recitals (11)-(12) move from broad goals to complementary, actionable initiatives at Union level, including sector-specific frontier, physical, and industrial AI research.

Sustainable data centres

Recital (13) lists efficiency by design and in operation: energy, water, quantum technologies, AI server efficiency, grids, clean energy, and on-site generation.

Cloud stacks and middleware

Recital (14) supports strategic-sector cloud-stack alternatives, EU AI-optimised hardware and software, co-design, and secure middleware for common European data spaces.

Open-source capacity

the creation of open-source software foundations supporting the design, development and maintenance of open-source components

Frontier AI

Recital (16) treats frontier AI as a strategic asset. The goal is to scale breakthroughs, reduce dependencies, and align the AI transition with Union values, safety, and long-term interests.

Physical AI

AI systems and models capable of perceiving the physical environment and executing complex actions within that environment

For physical AI, the text calls for high-quality data, compute access, and testing and validation in diverse real-world environments to establish robustness and reliability.

Industrial deployment

Priority areas include healthcare, transport, aerospace, automotive, manufacturing, defence, space, climate and environment, agri-food, services, and scientific discovery.

AI agents

Recital (21) envisages sovereign, secure platforms and orchestration frameworks for advanced AI agents, with transparency, accountability, testing, and experimentation.

Public-sector objective

increase the development and adoption of AI models and systems across the Union's public sector.

The stated uses are better decisions, simpler administration, and fewer unnecessary burdens. Healthcare data reuse should be facilitated while ensuring security and data protection.

Scale through reuse

Recital (22) supports better public data plus sharing and reuse of training data and AI models to avoid fragmentation and scale successful user-oriented solutions.

Adoption and skills

Recital (23) uses Centres for AI to support broad adoption, including regional and local levels, and envisages a dedicated cloud and AI skills curriculum.