The European Commission opened bidding this week for up to seven AI gigafactories across the continent, the latest step in its push to build domestic compute capacity. Public money is the seed, not the total: 10 billion euros pooled from the EU and national governments, structured to pull in at least 20 billion euros from private investors, which brings the program’s total near 30 billion euros. Participation spans eighteen member states, among them Germany and France.
The shared infrastructure is aimed at four kinds of users, in no particular order of priority: startups, government agencies, academic and industrial research groups, and larger companies that need training-scale compute without building their own data centers. Applications close November 12, 2026, and construction on the first sites is scheduled to begin in 2027. The initiative sits inside the EU’s broader “AI Continent” strategy, its answer to years of warnings that Europe trains and deploys its models on infrastructure it does not own.
Funding alone does not guarantee hardware, so Brussels lined up letters of intent from three chipmakers, AMD, Nvidia, and Qualcomm, aimed at locking in supply for the buildout. That detail matters more than the funding total. A gigafactory is a shell without allocated chips, and letters of intent are not delivery contracts. Nvidia’s order book already runs years deep with hyperscaler commitments; Europe is asking to be inserted into a supply chain it does not control, on a timeline that depends on suppliers headquartered outside the bloc.
The Decoder, which first reported the bidding announcement, drew the comparison: this year alone, the major American hyperscalers will pour upward of 600 billion dollars into their own data center buildouts, with no sign of that pace slowing. Measured against Europe’s roughly 30 billion euro package, the American figure comes out to about twenty times larger. That framing is accurate as a magnitude check, but the two numbers are not measuring the same thing.
The US figure is annual private capital expenditure by a handful of hyperscalers building out their own proprietary infrastructure. The European figure is a program total: 10 billion euros of public money, structured as EU and national co-investment, meant to catalyze 20 billion euros of private capital across up to seven sites over a multi-year build that will not break ground until 2027. One number describes a single year of corporate spending. The other describes a subsidized, multi-year public-private program still at the bidding stage. Dividing one by the other produces a ratio, not an equivalence.
Even accounting for that distinction, the gap in committed capital is real. Compressing the EU total into rough annual terms across a multi-year buildout still leaves a fraction of what individual US hyperscalers report quarter to quarter. The comparison is directionally sound. It is the precision, not the direction, that the 20 times figure overstates.
The more useful question is what 30 billion euros actually buys. Money finances buildings, power contracts, and cooling systems. It does not manufacture GPUs. Every gigafactory’s usable capacity depends on allocations from AMD, Nvidia, and Qualcomm, companies that answer first to whichever customers pay the most and wait the least. If those allocations lag behind construction, Europe will own data center shells cofunded by the state while the compute inside runs on hardware terms set in Santa Clara and Austin. The AI Continent strategy is a bet that public co-investment can pull chip supply toward Europe faster than the market would on its own. That bet is unproven until the first shipments arrive.
This buildout lands the same week other governments are making their own moves to control the inputs and outputs of AI, from an open-weight Chinese model cutting the cost of running a national AI stack to a US restriction on certain Chinese hardware.
Operators building for European enterprise customers should treat November 12 as the date that reveals whether this program attracts serious private co-investment or mostly public commitments, since that ratio will determine whether usable capacity arrives on the 2027 timeline or slips behind it.
The Decoder’s Matthias Bastian reported the gigafactory bidding announcement and the US spending comparison on July 31, 2026.
