Seven Factories
The European Union has unveiled plans to build seven AI giga-factories across the continent, each housing over 100,000 high-performance AI chips. The European Commission has earmarked around €10 billion in public funding and aims to attract an additional €20 billion in private investment, bringing the total to more than €30 billion.[1]
The original plan called for five giga-factories. The number was increased to seven due to higher-than-expected interest from member states. Ten countries have formally expressed interest: Germany, Italy, France, Poland, Czech Republic, Denmark, Finland, Greece, Portugal, and Spain. France has already announced its intention to submit a solo bid. Applications are open until November 12, with selected projects announced in early 2027 and operational facilities expected approximately 18 months later.[2]
€30 billion is a serious number. It is also, in the context of AI infrastructure spending, not that serious. The United States has seen private AI investment dwarf this figure through individual companies alone. China has been building AI capacity at a pace that makes European timelines look glacial. The EU is trying to close a gap that widens faster than it can fill it, and the mechanism it has chosen, public funding plus private co-investment, is the European way of doing things: deliberate, structured, and slow enough that the technology moves on before the buildings are finished.
The 18-month timeline from selection to operation is optimistic for European infrastructure projects. Permitting, environmental reviews, construction, procurement of 100,000 chips in a market where chip supply is constrained and competition for GPUs is fierce, staffing each facility with researchers and engineers who understand AI at scale, all of this takes time. The chips alone are a bottleneck. NVIDIA cannot make enough of them for the companies already buying them. Adding seven EU facilities to the demand queue does not make the queue shorter.
That said, the ambition is right. AI infrastructure is not something you can rent your way out of forever. Cloud providers can offer compute, but sovereignty over AI compute is becoming a strategic concern, and the EU has been more vocal about this than anyone. The giga-factories are framed as hubs for researchers, businesses, and start-ups to develop and train new AI models, with applications in healthcare, industry, research, and mobility. That is a lot of scope. It is also a lot of competing priorities for facilities that do not exist yet.
The question is not whether €30 billion is enough. It is whether the EU can move at the speed the technology demands, or whether by the time these factories open, the frontier they are trying to reach will have moved somewhere else entirely. The US and China are not waiting for 2027. Neither, for that matter, are the AI models that keep breaking out of their sandboxes.[3]
← All posts- RTL Today, "EU announces plans for seven AI giga-factories," July 31, 2026. today.rtl.lu ^
- Ibid. Member state interest, application timeline, and operational details from the same report. ^
- Ibid. Funding breakdown and scope from the same report. Cross-reference to Anthropic/OpenAI sandbox incidents covered in "The Models Keep Breaking Out," July 31, 2026. ^