The European Commission has opened the bidding to build up to seven AI “gigafactories,” vast computing hubs meant to give the bloc the raw power to train frontier AI models without depending on the United States.

The call for tenders, launched today, is pitched as a roughly €30 billion effort to close the gap with American and Chinese AI.

The structure is a public-private split. Brussels and member states would put in about €10 billion between them, with the other €20 billion expected from private investors, on the bet that state money can crowd in the capital these projects need.

The scale is deliberately enormous. Each gigafactory would house at least 100,000 cutting-edge AI chips, making it roughly four times more powerful than the largest data centres running in the EU today, and together they would more than double the bloc’s AI compute.

The “gigafactory” name is a deliberate step up. Europe already funds a network of 19 smaller “AI factories” attached to its supercomputers, but the new sites are meant to be an order of magnitude larger, built to train the kind of frontier models that today only a handful of US and Chinese labs can afford.

The appetite is clearly there. An earlier expression-of-interest round drew 76 responses from consortia across the continent, with ten countries, including Germany, France, Italy, Spain, and Poland, lining up to host, and France already signalling it will go it alone.

The reasoning is competitive anxiety. Europe has watched the US and China pour money into AI infrastructure while its own firms rent compute from American clouds, and Henna Virkkunen, the Commission’s tech-sovereignty chief, called that raw scale of computing power “a strategic necessity.”

The ambition, though, runs ahead of the cheque. Only about €1 billion of Brussels’ share is actually committed; the rest depends on the EU’s next long-term budget, the Multiannual Financial Framework, which has not been agreed.

A senior official said as much, unusually plainly. “We cannot pre-empt the decisions about the next MFF,” the official told reporters, describing the sum as a “best estimate” of what might be available rather than money in hand.

The timeline is tight against that uncertainty. Construction is meant to begin in early 2027 and the first gigafactories to come online by mid-2028, a schedule that assumes budgets, sites, and supply chains all fall into place.

History suggests caution. The gigafactory plan was first floated in early 2025 and has since slipped repeatedly, with delays that frustrated some of the very partners the Commission now needs to build the things.

There is also an awkwardness at the heart of the sovereignty pitch. The chips that would fill these European gigafactories will overwhelmingly come from Nvidia, AMD, and Qualcomm, all American, which means the bloc would be buying its independence from the companies it wants independence from.

Energy is the other constraint. European electricity costs two to three times what it does in the US or China, and data-centre power is already a limiting factor that no amount of subsidy can quickly wish away.

Critics have questioned the whole model. Some argue that renting sovereign-branded capacity built on foreign-owned hardware reinforces an illusion of independence rather than delivering it, and that Europe’s real gap is in chips and models, not buildings.

The counter is that infrastructure is a start, not a finish. Compute alone will not conjure a European OpenAI, but without it the continent’s startups have nowhere to train at scale, which is the gap the Commission is trying, expensively, to fill.

The Commission’s answer is that you have to start somewhere. Public funders would get a proportional share of the compute for research and public projects, and the hope is that building the infrastructure draws the talent, the startups, and eventually the chipmaking toward Europe.

For now, the door is open and the money is mostly notional. Firms can bid to build Europe’s AI future, on the understanding that the continent has announced €30 billion of ambition and secured, so far, about a thirtieth of it.

Get the TNW newsletter

Get the most important tech news in your inbox each week.