EU AI gigafactories plan seeks up to €30 billion for compute buildout

The European Commission has opened bidding for EU AI gigafactories, a plan to build as many as seven large AI computing facilities across Europe with a funding package of roughly €30 billion. The Commission said up to €10 billion from EU and national sources is expected to attract at least €20 billion in private investment, a public-private structure aimed at expanding Europe’s ability to train and run large AI models.

The facilities are intended to serve startups, established companies, research institutions and public agencies that need access to high-end AI infrastructure, according to the Commission. Eighteen EU member states are participating, including Germany and France. Applications are due Nov. 12, 2026, and construction of the first sites is scheduled to start in 2027.

The Commission also said it has signed letters of intent with AMD, Nvidia and Qualcomm to secure access to hardware. It did not disclose binding supply terms, chip volumes, site locations, power arrangements or the expected compute capacity of the proposed facilities.

What are EU AI gigafactories?

AI gigafactories are large-scale computing sites designed to provide the infrastructure used to train and operate advanced AI models. In practical terms, they are meant to pool expensive chips, data center capacity and supporting systems so that European users are not wholly dependent on overseas hyperscalers for frontier-scale compute.

The program is part of the EU’s broader “AI Continent” strategy, according to the Commission. That strategy is meant to increase regional AI capacity at a time when model development is being shaped less by small lab teams and more by access to capital-intensive infrastructure.

How does the EU AI gigafactories plan compare with U.S. spending?

The scale gap remains the main industry context. The Decoder has reported that major U.S. technology companies plan to spend more than $600 billion on data centers this year. Compared with that figure, Europe’s roughly €30 billion package is about one-twentieth the size on a simple headline comparison.

That does not make the EU program irrelevant, but it limits what it can plausibly change by itself. Seven facilities could improve access for European companies and researchers that cannot buy hyperscaler-scale capacity directly. The plan is less likely to match the capital intensity of U.S. AI infrastructure spending if the current race for larger training runs and more inference capacity continues.

For European startups, the practical question is whether the facilities will offer timely, usable capacity rather than a symbolic alternative to U.S. cloud providers. The Commission has not yet detailed pricing, allocation rules, technical specifications or how commercial users will be prioritized against research and government demand.

The bidding process now shifts the plan from strategy language to procurement. The outcome will depend on whether member states and private investors can turn the announced funding structure into sites with enough chips, power and network capacity to matter for companies building or deploying large AI systems in Europe.

This story draws on original reporting from The Decoder.