EU scales back data center tender as Europe's AI infrastructure gap widens

The European Union has reduced the scope of its latest data center procurement, a move that highlights the growing mismatch between the bloc’s stated AI ambitions and its ability to deliver on them.

On June 3, the European Commission unveiled its Technological Sovereignty Package. The package is a legislative bundle targeting cloud computing, semiconductors, and AI development.

That package set an investment target of around €422 billion ($490 billion) over the next decade across data centers, chips, cloud, AI, and open-source software.

This time around, the bloc has scaled back the number of advanced chips per data center. Initially, it planned for 100,000 advanced chips for five data centers, but now it is seeking bids to build seven data centers, with the first phase seeing four data centers taking at least 25,000 GPUs and three taking at least 40,000 processors.

This time around, the bloc plans that the private sector will be responsible for most of the financing, according to people familiar with the development.

A continent playing catch-up

Europe currently accounts for approximately 5% of global AI compute capacity. 

The United States is responsible for around 80%. American tech firms have poured more than $400 billion into AI infrastructure expansion in 2025 alone, a sum that dwarfs the EU’s €200 billion plan, much of which consists of repackaged funding spread across multiple years.

AWS, Microsoft Azure, and Google Cloud together command about 70% of Europe’s cloud market. 

The US Cloud Act gives American authorities the power to compel those providers to hand over data regardless of where it is stored. An Allianz report published in late May found that US firms control roughly 80% of Europe’s cloud compute market and close to 60% of its enterprise software revenue.

Is Anthropic’s restriction of advanced models a wake-up call to Europe?

The urgency behind Europe’s sovereignty push intensified on June 13, when the US government restricted foreign access to Anthropic’s most advanced AI models, Mythos and Fable. 

That move confirmed the fear that European policymakers had long had over Washington’s ability to effectively flip a switch and cut off access to critical AI systems.

“We cannot rely on tools developed by foreign powers. France must have its own tools,” French Prime Minister Sebastien Lecornu reportedly said.

Telecoms firm Orange, in its response to the US government’s move to restrict access to Anthropic’s advanced models, stated that access to AI that can “never be switched off on a whim” is now a strategic necessity for Europe.

The incident gave fresh ammunition to advocates of the Commission’s Cloud and AI Development Act, which aims to triple EU data center capacity over five to seven years. Under the proposal, strict sovereignty requirements would apply to sensitive public-sector workloads in healthcare, finance, and justice.

Critics question whether the plan matches the problem

Competition economist Cristina Caffarra, who leads the EuroStack Industry Initiative, stated that the Commission’s measures were very feeble. Caffarra added that stronger procurement mechanisms had been watered down under pressure from Washington.

Under the proposed regulation, only about 10% of cloud contracts would carry a strong European sovereignty standard. The remaining 90% would remain open to all suppliers, including American hyperscalers.

In March, Siemens CEO Roland Busch warned that prioritizing domestic AI infrastructure over deploying existing tools would slow Europe’s adoption of the technology. He accused the EU of having a “miscalibrated” approach to AI regulation and compared America’s embrace of the technology to a “fast flowing river” against Europe’s “standing water.”

The cost of going it alone is also high, with Capgemini COO Karine Brunet pointing out that European cloud alternatives carry premiums of up to 40% over their American counterparts, and this forces companies to weigh sovereignty against their budgets.

Is Mistral Europe’s answer for the AI race?

France’s Mistral remains Europe’s most prominent AI contender. The startup is reportedly in talks to raise $3.5 billion at a $23.2 billion valuation. 

That figure is still a fraction of OpenAI’s valuation, which is north of $500 billion.

Former Google CEO Eric Schmidt warned at Davos in January that without significant investment in open-source AI labs and lower energy costs, Europe would default to using Chinese AI models, an outcome he called “probably not good,” as Cryptopolitan reported.

The Commission says it plans to launch a call for AI gigafactories in July and will consult with member states and the European Investment Bank on funding mechanisms.

If you're reading this, you’re already ahead. Stay there with our newsletter.