EU Commits 10 Billion Euros to Build 7 AI Gigafactories, Each Packing 100K+ Top-End Chips
The EU opened bids on July 30: 10 billion euros in public funding, seven AI gigafactories, each packed with at least 100,000 cutting-edge AI chips — roughly four times the compute power of any data center currently operating on the continent.
Three Key Takeaways
10 billion public, 30 billion total. The European Commission formally launched the AI Gigafactory call for proposals on July 30, putting up €10 billion ($11.4 billion) in public funding and targeting €20 billion in private co-investment for a total of €30 billion. The seven gigafactories will be distributed across EU member states, each housing at least 100,000 state-of-the-art AI chips for training, inference, and fine-tuning frontier models. Bidding closes November 12, with awards expected in early 2027 and construction starting the same year.
Where the chips come from: AMD, Nvidia, and Qualcomm have signed on. The Commission confirmed it has signed letters of intent with AMD, Nvidia, and Qualcomm to give gigafactory consortia "seamless access to necessary hardware." This means the EU is keeping a pragmatic procurement channel alongside its chip-sovereignty ambitions — stack up compute with American silicon now, while waiting for domestic semiconductor capacity to catch up.
Built for startups and SMEs, not big tech. Unlike US hyperscalers building data centers for their own models, the EU's AI Gigafactories are explicitly positioned as open infrastructure, prioritizing access for European AI startups, small and medium enterprises, and research institutions. The logic: Europe does not lack AI talent or papers — it lacks the large-scale compute needed to move models from the lab to production.
WangDou's Take
Thirty billion euros sounds massive until you line it up against the numbers next door: Nvidia backstopped $250 billion for a single OpenAI data center in Ohio, and Meta alone is spending $130–145 billion on AI capex this year. All seven EU gigafactories combined might not match the GPU cluster Meta adds in a single quarter. That said, the EU got one thing right this time: it is not trying to train its own frontier model — it is building roads so others can drive. Europe's AI problem has never been a shortage of smart people; it is that smart people finish their experiments, discover they cannot get GPU time, and book a one-way ticket to San Francisco. Whether this €10 billion actually keeps them home depends less on chip counts and more on queue times.
Source: Bloomberg, Washington Post, Mobile World Live