Summary

• EuroHPC has launched a call for consortia to build and operate AI Gigafactories in the EU.
• Public funding will be used as an anchor customer to help unlock private investment.
• The programme links AI sovereignty, cloud capacity, datacentre infrastructure, energy use, and industrial policy.

The EuroHPC Joint Undertaking has opened a call for tenders to select consortia that will build and operate future AI Gigafactories in the European Union, turning part of Europe’s AI sovereignty agenda into a procurement exercise.

The facilities are intended to combine large numbers of advanced AI processors with the software, cloud technology stacks, storage, high bandwidth connectivity, and energy efficient datacentre infrastructure needed to train, fine-tune, and run large AI models. The call is open to consortia or special purpose vehicles that bring together companies, public bodies, investors, and other partners.

EuroHPC and participating states plan to jointly procure compute access time from the selected facilities. That public demand is designed to work as an anchor customer, reducing the investment risk for projects that require heavy upfront capital before utilisation is guaranteed. The submission deadline is 12 November 2026, with successful projects expected to be selected in early 2027 and begin operations within 18 months.

The procurement is expected to support up to seven AI Gigafactories, each equipped with a very large number of state of the art AI processors and located in at least seven member states. Some facilities may operate across borders, allowing infrastructure to be distributed rather than tied to a single physical site.

Although the policy language is about technological leadership and strategic autonomy, the operational question is more direct. European companies, researchers, startups, public bodies, and industrial groups need access to AI compute that is affordable, reliable, and governed under rules they can understand. Without that, high-end AI development remains concentrated around whichever global cloud and chip supply arrangements can be secured fastest.

The programme builds on the EU’s earlier AI Factories initiative, but it moves into a heavier infrastructure tier. AI Factories are intended to provide support services around Europe’s high performance computing network. AI Gigafactories are being framed as large industrial AI facilities that can offer frontier model training, fine-tuning, and inference services at a scale closer to the facilities being built by hyperscalers and AI labs.

Capital intensity will test the model. Chips, power, cooling, land, connectivity, and specialist operations teams are all constrained resources, while Europe’s energy and planning systems already struggle to absorb datacentre demand in several mature markets. The best proposals will therefore need to show more than hardware ambition; they will need credible energy access, resilient operations, secure cloud interfaces, and a route for smaller users to consume compute without building infrastructure teams of their own.

The use of public procurement also changes the market signal. Rather than waiting for private capital to solve a strategic dependency, the EU is using committed demand to make projects investable. If the model works, European organisations will gain a more credible route to high-end AI capacity. If it moves too slowly, the region’s AI policy will keep running behind the infrastructure choices already being made elsewhere.