EU commits €10 billion to build seven AI gigafactories

The European Commission will spend €10 billion to construct seven AI gigafactories across member states, with an additional €20 billion expected from private investors. Applications open now and close November 12, 2026. Successful projects will be announced in 2027 and must become operational within 18 months of signing.

The facilities will house advanced AI processors, high-speed connectivity, and energy-efficient data centres. Their purpose: give European startups, SMEs, and larger enterprises access to infrastructure for training, inference, and fine-tuning frontier AI models. The Commission frames this as a strategic move to close the compute gap with the United States and China.

What exactly are these gigafactories?

The term "gigafactory" here refers to large-scale AI compute centres, not the battery plants Tesla popularised. Each facility will combine processors, software stacks, cloud infrastructure, and high-bandwidth networking into a single resource pool. The Commission says these seven new sites will join 19 existing AI gigafactories in Europe.

Henna Virkkunen, EVP for Tech Sovereignty, Security and Democracy at the European Commission, called the scale of compute power "a strategic necessity for Europe as AI development accelerates."

The facilities must comply with EU rules on data protection, security, and ethics. That constraint is deliberate. Brussels wants European companies to train models on European infrastructure under European law, rather than depending on hyperscalers headquartered in jurisdictions with different regulatory frameworks.

How the bidding works

The Commission has opened a call for tenders. Consortiums or special-purpose vehicles that bring together companies, public entities, investors, and other partners may apply. A single project can sit in one member state on one site, spread across several locations in one country, or span multiple member states through distributed computing arrangements.

All applications go through a competitive valuation. Winners get announced sometime in 2027. From contract signature, each facility has 18 months to reach operational status.

What this means for AI builders in Europe

The immediate implication is access. Training a frontier model today requires tens of millions of dollars in compute. Most European startups cannot afford that, so they rent capacity from US hyperscalers or simply build smaller models. If these gigafactories deliver subsidised or priority access to startups and SMEs, as the Commission claims, European AI companies could train larger models without relocating or restructuring around US cloud providers.

The harder question is execution. Eighteen months from contract to operational is aggressive for large-scale data centre builds. Power, permitting, and chip supply chains have all slowed similar projects elsewhere. The Commission has not disclosed what happens if a winning consortium misses the deadline.

The sovereignty bet

Brussels has been explicit: this is about technological sovereignty. The EU does not want its AI sector dependent on American or Chinese compute. That goal carries trade-offs. Mandating EU-rules compliance may slow some projects. Restricting access to non-EU partners could limit collaboration. And if the gigafactories cost more to operate than hyperscaler alternatives, usage may lag.

Still, for companies that need to keep training data under GDPR jurisdiction, or that face customer pressure to avoid US cloud providers, these facilities could solve a real problem. The open question is whether seven new sites, added to 19 existing ones, actually close the compute gap, or just narrow it.

Source: TahawulTech.com / Daniel Shepherd