# Europe bets €10 billion on AI gigafactories but Nvidia still holds the keys > Source: > Published: 2026-07-30 12:00:13+00:00 *The European Commission opened tenders today for seven AI gigafactories backed by €10 billion in public funding, part of a broader €30 billion push that critics say cannot paper over a fundamental chip dependency problem.* Brussels launched its formal call for bids on July 30, 2026, offering €10 billion ($11.4 billion) in public financing to build seven large-scale AI compute facilities across the EU. The plan is part of the broader InvestAI initiative targeting €200 billion in AI investment by 2030. Private capital is expected to match and exceed the public figure: the Commission wants that €10 billion to pull in an additional €20 billion from industry, hitting roughly €30 billion total. Each winning facility must deploy at least 100,000 next-generation AI chips - making them about four times more powerful than anything currently running on EU soil. Applications close November 12, with construction expected to begin in early 2027 and facilities coming online by mid-2028. Henna Virkkunen, the Commission's executive vice president for tech sovereignty, framed it bluntly: "Access to the raw scale of computing power within AI gigafactories is a strategic necessity for Europe as AI development accelerates." The subtext is hard to miss. American hyperscalers - Amazon Web Services and Microsoft Azure chief among them, with Google Cloud not far behind - currently hold approximately 70 percent of the EU cloud market. European AI startups training frontier models have, until now, had essentially one practical option: rent compute from the same US firms Brussels is trying to reduce its dependence on. The Commission has signed letters of intent with three hardware suppliers: Nvidia, AMD, and Qualcomm. All three are American companies. Analysts at CEPS have identified the chip layer as the sovereignty blind spot in the entire programme. Put plainly, Europe is building AI factories that will run on foreign silicon, sourced from vendors it has no meaningful influence over. The Commission says consortia bidding for gigafactory slots will have streamlined access to hardware through those agreements - a practical concession, given that there is no European alternative at the scale required. Some 76 expressions of interest from 16 EU member states, covering 60 different sites, have already been submitted. That suggests genuine appetite from national governments and industry consortia. A slice of the procurement is reserved for startups under three years old and scale-ups moving beyond early-stage risk. Smaller players get a real path to subsidised compute access. They don't just watch incumbents build the infrastructure around them. But here's the thing: €10 billion of public money sounds large until you put it beside what the US is spending. Google, Amazon, Meta, and Microsoft combined are on track to spend roughly $725 billion on AI infrastructure in 2026 alone - up 77 percent from $410 billion the year before, according to analyst estimates. Microsoft's data center lease commitments reportedly exceed $130 billion. The EU's entire gigafactory programme, public and private money combined, is a single quarter of what one US hyperscaler commits in annual capex. ## The sovereignty paradox Europe hasn't solved Critics - including researchers writing for The Next Web and the Internet Policy Review - have identified what they call a structural trap in the gigafactory model. Training large frontier models requires enormous compute. Inference, the actual running of those models at scale for commercial use, requires a different and even larger infrastructure footprint. Europe may end up building training compute that produces models which then have to run on AWS or Azure anyway, because the inference capacity doesn't exist at home. Build the gigafactories, and you potentially subsidise capacity that the only customers large enough to fill it - the leading US AI firms - end up using. That's a real tension. The Commission doesn't have a clean answer to it. What it does have is a deadline problem: European AI labs and startups are making infrastructure decisions right now. The first gigafactories don't come online until mid-2028. Until then, the practical compute options remain largely American. The InvestAI programme dates to the Paris AI Action Summit in February 2025, where the €200 billion mobilisation target was first announced. The gigafactory tender launched today is the largest single procurement action inside that framework. Whether it moves the needle depends entirely on execution speed, private co-investment actually materialising, and whether consortia can source chips at the volumes required. The Commission declined to specify which member states are favoured. There is no guarantee the facilities end up distributed across Europe rather than concentrated in the countries with existing data center infrastructure - Germany and the Netherlands being the obvious candidates, with Ireland close behind. For European AI startups, the practical upside is real if the timeline holds: sovereign compute access at competitive rates, without routing training runs through a US hyperscaler's billing department. For chipmakers and cloud providers bidding on the buildout, Nvidia is almost certainly supplying the GPUs regardless of who wins. The question is which European energy providers, construction firms, and cloud operators end up as the primary infrastructure partners. That contest is only just beginning. **Also read:** [The EU is betting €30 billion on AI gigafactories but it's playing catch-up in a race it's already losing](https://startupfortune.com/the-eu-is-betting-30-billion-on-ai-gigafactories-but-its-playing-catch-up-in-a-race-its-already-losing/) • [Microsoft posted record fiscal 2026 revenue as Azure crossed $100 billion and proved its AI bet is paying off](https://startupfortune.com/microsoft-posted-record-fiscal-2026-revenue-as-azure-crossed-100-billion-and-proved-its-ai-bet-is-paying-off/) • [Encore AI raises $30M after its bank and insurer customers decided to become investors](https://startupfortune.com/encore-ai-raises-30m-after-its-bank-and-insurer-customers-decided-to-become-investors/)