While the headlines have focused on the US and China's AI race, Europe has been quietly assembling its own answer: sovereign compute, on European soil, under European jurisdiction. The numbers involved are no longer symbolic. They run into the tens of billions of euros and tens of thousands of GPUs, and they are backed by governments, sovereign wealth funds, and some of the continent's largest regulated companies.
The €109 billion moment
At the Paris AI Action Summit in February 2025, France announced €109 billion in AI infrastructure investment. The commitments behind that figure came from a mix of sources: the UAE pledged €50 billion toward an AI campus, Brookfield committed €20 billion for data centers, and French companies Iliad, Orange, and Thales added their own contributions. This was not a single company placing a bet. It was a coordinated signal that Europe intends to build, not just consume, AI infrastructure.
Germany moved in parallel, launching its "Industrial AI Cloud": 10,000 NVIDIA Blackwell GPUs delivering 0.5 ExaFLOPS of compute, operational since early 2026 and already running at over one-third utilization. That utilization number matters. Infrastructure announcements are easy to make; infrastructure that gets used within months of going live is a different signal entirely.
Mistral's bet on sovereign compute
The clearest commercial expression of this push is Mistral AI's Mistral Compute platform, built on 18,000 NVIDIA Grace Blackwell chips housed in a 40 MW data center in Essonne, just south of Paris. In September 2025, Mistral raised €1.7 billion at an €11.7 billion valuation, with semiconductor equipment manufacturer ASML taking an 11% stake and becoming the company's largest shareholder.
ASML's involvement is worth sitting with. This is not a venture fund chasing an AI valuation; it is the company that makes the lithography machines the entire chip industry depends on, betting on a European AI compute provider. That is a supply-chain player putting capital behind sovereignty, not just software.
Why regulated enterprises are opting in
The demand side of this story is arguably more telling than the investment side. Named enterprise clients moving toward sovereign infrastructure include BNP Paribas, Orange, SNCF, Thales, Veolia, and Schneider Electric: banks, utilities, transport operators, and defense-adjacent industrials. These are organizations with the least tolerance for regulatory ambiguity, and they are the ones voting with their infrastructure budgets.
The jurisdictional argument driving this shift is concrete. One widely cited figure states that over 80% of digital tools used by European governments are designed, operated, or controlled outside of Europe. Combine that with the US CLOUD Act, which allows US authorities to compel American providers to hand over data regardless of where that data is physically stored, and the exposure becomes obvious: a European company running critical workloads on a US-controlled cloud is subject to US law, not just its own.
The energy question behind the compute question
None of this works without power, and Europe is treating that constraint seriously. France's Fluidstack partnership plans to deploy 250 MW of AI compute in 2026, backed by €10 billion in investment, with ambitions to scale "beyond 1 GW by 2028." EDF has already offered four sites with a combined 3 GW of available power to support the buildout.
That is the part of the story that rarely makes it into AI headlines: sovereignty at this scale is as much an energy and industrial-planning problem as it is a chips-and-models problem. The countries and companies that solve for power alongside compute are the ones that will actually be able to deliver on these commitments.
What this means if you operate in Europe
For any company handling regulated, sensitive, or strategically important data, the calculus is shifting. Sovereign compute is no longer a niche option reserved for governments; it is becoming a mainstream infrastructure choice for banks, utilities, industrials, and increasingly, mid-sized companies that simply do not want their data governed by a foreign legal regime by default.
This matters well beyond the infrastructure layer. Whether you are building on top of sovereign cloud providers or evaluating where your own AI systems and data should live, jurisdiction is now a design decision, not an afterthought. If you are weighing where and how to build your AI systems with these constraints in mind, get in touch with our team. We help companies design AI solutions that hold up under real regulatory scrutiny, not just in a pitch deck.