Europe’s AI plan has reached the data centre queue

The AI race sounds abstract until somebody asks where the servers will sit. Then it becomes a queue: grid connection, land, cooling, planning permission, procurement rules, supplier risk and a budget line large enough to survive contact with a data-centre quote.

That is the useful way to read the European Commission’s 3 June tech sovereignty package. The Commission framed it as a push to strengthen Europe’s ability to act independently in the digital world, with two legislative proposals, a Chips Act 2.0 and a Cloud and AI Development Act, plus an Open Source Strategy and a roadmap for digitalisation and AI in energy. The language is grand. The machinery is not. It is about whether Europe can build, buy and maintain the digital infrastructure it says it needs.

The starting point is dependency. The Commission says the EU relies on non-EU countries for over 80% of key digital products, services, infrastructure and intellectual property. That figure does not mean every cloud contract is suddenly unsafe or every foreign supplier is a problem. It does mean the AI boom has made an old question harder to ignore: who controls the compute, chips, software layers and service contracts that modern public services and businesses now depend on?

The proposed Cloud and AI Development Act is the clearest sign that AI policy has moved out of the demo room. The Commission says Europe faces long permitting procedures, limited access to energy, land, water and financing, and over-reliance on non-EU cloud providers. The proposal aims to make it faster to deploy sustainable cloud and data-centre infrastructure and to at least triple EU data-centre capacity within the next five to seven years.

Tripling capacity is easy to type and hard to build. A data centre is not just a box of servers. It is an electricity user, a heat problem, a local planning issue and, increasingly, a political argument about who gets scarce grid capacity. The Commission’s plan recognises that by linking cloud and AI expansion to permitting, energy and public procurement. That matters because the bottleneck may not be the model. It may be the substation.

There is a second layer: chips. The Chips Act 2.0 page says the original European Chips Act helped mobilise more than €52 billion in public and private investment and created an estimated 46,000 direct and indirect jobs. Yet the Commission still describes dependency in areas such as advanced chip manufacturing and semiconductor design. The new proposal points to faster permitting, demand accelerators, strategic projects and support for AI chips. It also notes that the global semiconductor market is expected to reach €1.37 trillion by 2030, with AI-related components driving around 70% of that growth.

Those numbers explain the urgency, but they also warn against a tidy story. Europe is not trying to build every chip, cloud service and AI stack alone. The more realistic goal is leverage: more domestic capacity, more bargaining power, less single-supplier fragility and fewer moments where a critical service depends on decisions made far outside Europe’s legal and political reach.

The Open Source Strategy is the quieter part of the package, and possibly the most revealing. The Commission says open source can reduce dependence on non-EU proprietary technologies and increase control over critical digital infrastructure. That is not a romantic claim about volunteer coders saving the continent. Open source only helps if projects are funded, maintained, governed and made usable in public and private procurement. The strategy’s lifecycle language matters because abandoned code is not sovereignty. It is another dependency with fewer invoices.

For startups and smaller companies, the practical question is access. The Commission’s AI Factories programme says 19 AI Factories and 13 antennas are operational or selected, with priority access for startups and SMEs. It also describes an InvestAI Facility of €20 billion to create up to five AI Gigafactories. The promise is that European firms will not have to watch the largest players turn compute into a private moat. The test will be dull and decisive: prices, waiting lists, technical support, terms of use and whether smaller teams can actually train or adapt models without being buried in paperwork.

For public bodies, the package points toward harder procurement conversations. If a council, hospital system or ministry wants AI-enabled software, the cloud layer is no longer just a background IT choice. It can affect data control, resilience, exit rights and the ability to audit what a supplier is doing. None of that makes procurement simpler. It makes the trade-offs more visible.

The package is still a policy package, not a magic build-out. Legislative proposals have to pass through the EU process. Member states will still argue about energy, permitting and money. Local communities will still have views on data centres. Companies will still buy from providers that work, not providers that sound geopolitically satisfying in a strategy paper.

That is why the sober version is the stronger one. Europe’s AI plan is not just a promise to have better models. It is a bet that infrastructure, chips, cloud rules and maintained open-source software can be treated as part of the same system. If that sounds less exciting than a chatbot launch, good. The boring parts are where the next phase of AI will either become usable or get stuck in the queue.