Brussels: The European Commission has opened its most ambitious industrial bet on artificial intelligence to date, inviting industry consortia to build up to seven AI gigafactories across the Union. Officials frame the call as the moment Europe stops renting the compute that trains frontier models and starts owning it.
The numbers explain the urgency. Up to 10 billion euro in EU and national money will anchor the programme, and the Commission expects that pledge to pull in at least 20 billion euro of private capital. The combined figure tops 30 billion euro, one of the largest coordinated infrastructure commitments the bloc has attempted outside energy and transport.
What the gigafactories are meant to do
Each site will bundle advanced AI processors, cloud software stacks, high-speed connectivity and energy-efficient data centres into a single facility. The design lets start-ups, universities, small manufacturers and public bodies rent capacity for training, inference and fine-tuning that few of them could afford alone. The EuroHPC Joint Undertaking, which already runs Europe’s public supercomputers, will manage the evaluation and award the contracts.
The initiative sits inside the wider InvestAI strategy, which aims to mobilise 200 billion euro for artificial intelligence and give the bloc what officials call sovereign compute. The gigafactories represent the physical layer of that plan, the racks and cooling systems that turn ambition into usable capacity.
The timeline is tight but concrete. The call closes on 12 November 2026, award decisions should land by early 2027, and construction is meant to begin the same year. That schedule reflects how quickly the compute gap has widened as American and Chinese labs pour capital into ever larger clusters.
The questions the money cannot answer yet
Scale alone will not settle Europe’s AI position. Three problems shadow the announcement. Energy comes first, because gigawatt-class data centres demand power that several national grids already struggle to supply, and the promise of efficient design does not remove the raw load. Chips come second, since the most capable processors still ship from a handful of non-European suppliers, and a factory full of imported silicon delivers capacity without sovereignty.
Demand is the third worry. Building compute assumes European firms will use it at a rate that justifies the outlay. The continent’s AI start-ups remain smaller and fewer than their American rivals, so the gigafactories could either seed a wave of home-grown model builders or sit half-idle while the best European talent keeps renting capacity abroad.
Location politics will complicate matters further. Seven sites across twenty-seven member states guarantee a scramble, and governments will read each award as a verdict on their digital ambitions. The Commission insists that industrial merit, not geography, will pick the winners, yet cohesion pressures rarely stay out of decisions this large.
For all those caveats, the strategic logic holds. Compute has become the raw material of the AI economy, and a bloc that imports almost all of it cedes control over price, access and priorities. By committing public money on this scale, Brussels signals that it treats AI infrastructure the way it once treated railways and power lines, as a shared foundation too important to leave to the market alone.
The gigafactories will not close the gap with the United States overnight, and honest officials admit as much. What the call does is give European industry a credible reason to build at home rather than abroad, and it puts real money behind a phrase that has echoed through Brussels for two years. Whether the AI gigafactories become engines of European autonomy or expensive monuments to it will depend on what runs inside them once the concrete sets.
