What opened on 30 July
The European High Performance Computing Joint Undertaking launched the AI Gigafactories call on 30 July, inviting tenders to select the consortia that will build and operate up to seven AI gigafactories in the European Union. The financing structure is stated plainly: up to 10 billion euro in EU and national funding, at least 20 billion euro expected from private investors, and a total the Commission describes as more than 30 billion euro of investment unlocked. Applications close on 12 November 2026.
The facilities are meant to give start-ups, scale-ups, smaller companies, industrial firms, universities and public authorities access to infrastructure for training and fine-tuning advanced AI models, combining AI processors, cloud and software systems, high-speed connectivity and energy-efficient data centres. Bids may come from consortia or special purpose vehicles made up of companies, public bodies, investors and other partners, and a gigafactory may sit in one member state, on one site, across several locations, or span borders as distributed computing infrastructure.
Count the days, then read the entry requirements
From 30 July to 12 November is 105 days. In that window an applicant is expected to have formed a consortium or a special purpose vehicle containing companies, public bodies and investors, agreed how billions of euro of private capital will be committed, and identified sites with the power and connectivity a facility of this size needs. Assembling that from a standing start in fifteen weeks is not a realistic proposition, and it is worth saying so directly rather than treating the deadline as an ordinary administrative detail.
The implication is not that the process is unfair, it is that the process already ran. The Commission held a non-binding call for expression of interest in 2025, which drew 76 respondents proposing sites in 16 member states across 60 locations. That exercise carried no financial commitment and was explicitly a mapping exercise, but it is also where the groupings, the site options and the relationships were formed. A company reading today's announcement as an open invitation has misread the timeline. The realistic reading is that this call converts an existing pipeline into binding bids, and that anyone who skipped the 2025 round is now looking for a seat in someone else's vehicle rather than building their own.
A third of the money is public, which sets the price
Ten billion euro of public support against at least twenty billion of private capital puts the state at roughly a third of the total. That ratio is the most useful number in the announcement, because it tells you what these facilities will be. A project financed two-thirds privately has to return capital to private investors, which means the compute will be sold at commercial rates to whoever can pay, under contracts written to satisfy the people who funded it.
So the mental model of a European public compute utility, available cheaply to European companies because European taxpayers paid for it, does not survive contact with the financing structure. What public money buys here is de-risking and bankability, not price control. Spread across seven facilities, the public contribution is roughly 1.4 billion euro each against a total nearer 4.3 billion each. An owner planning to use this capacity should budget for market pricing and treat any preferential access for European small and medium companies as something to verify in the eventual terms rather than assume from the policy language.
Where the compute lands is not yet a geography
One provision deserves more attention than it will get. A gigafactory may be established across multiple locations and may span several member states as distributed AI computing infrastructure. That flexibility is sensible, since few single European sites can supply the power a facility of this scale needs, and it lets bidders assemble capacity where grid connections actually exist.
It also means the word gigafactory may describe a network rather than a building, and that has consequences for anyone planning around it. Latency, data residency and the practical question of which national jurisdiction holds the hardware all become properties of the winning bid rather than properties you can plan against now. If your interest in this scheme is sovereignty in the strict sense, that a specific workload sits under a specific national law, then the announcement does not yet answer your question, and the answer will arrive with the awards rather than with the call.
The date to put in your plan is not November
For the small number of organisations that will bid, 12 November is the only date that matters and the work is already late. For everyone else, which is almost everyone, the useful exercise is to work out when this capacity can actually be used. The call closes in November, awards follow, consortia then finance, site, build, power and commission facilities of unprecedented scale for Europe. None of those steps is short, and nothing in the announcement compresses them.
That produces the honest conclusion for an operator: this is a late-decade supply story, and it does not relieve a compute or memory constraint you are feeling in 2026. Plan procurement on the market that exists, treat European gigafactory capacity as an option that may improve your negotiating position several years from now, and revisit the assumption when awards are published rather than when calls are announced. The gap between a policy timeline and a procurement timeline is where a great deal of European planning goes wrong, and this scheme is a clean example of the distance between the two.
Read next: Nine European Vendors Sit On A US Retaliation List | Your AI Vendor Is About to Get a European Auditor



