The European Commission opened a formal tender on July 30 to select consortia that will build up to seven “AI Gigafactories” across the European Union, backing the project with roughly €10 billion in public funding and aiming to unlock more than €20 billion in private investment, according to the European Commission and the EuroHPC Joint Undertaking, which is running the process.
What is an AI Gigafactory
Each site is expected to pack at least 100,000 state-of-the-art AI chips — about four times the power of a typical AI data center now running in the bloc, according to the Commission. The gigafactories go well beyond the EU’s existing network of 19 smaller AI Factories, and officials say the new sites should more than double the bloc’s overall AI computing capacity once operational. The model echoes what Japan has already built with Nvidia — see Nvidia and Japan’s national AI factory — though Brussels’ version is larger and publicly anchored from the start.
Money and timeline
Public funding from the EU and member states will act as an “anchor customer” for each project, a structure meant to de-risk the private capital needed to actually build and run the sites, according to the EuroHPC call. Consortia — which can combine companies, public bodies and investors, and may span more than one member state — have until November 12 to submit bids. The Commission expects to select winners in early 2027, with each facility required to become operational within 18 months of that decision.
The plan’s scope already grew once, from an original five gigafactories to seven, after an earlier informal call drew 77 proposals across 16 member states — a sign of how much demand exists for EU-based AI training capacity.
Why it matters
“Access to the raw scale of computing power within AI gigafactories is a strategic necessity for Europe as AI development accelerates,” said Henna Virkkunen, the Commission’s executive vice president for tech sovereignty. The initiative reflects a persistent worry in Brussels: that its AI regulation has moved faster than its compute, leaving European researchers and companies dependent on American and Chinese infrastructure to train frontier models.