The European Union’s tender framework for AI gigafactories now allows bidders to propose facilities that can grow in stages rather than committing immediately to the largest scale originally envisaged. The shift reflects mounting concerns over electricity supply, grid connections, financing gaps and the ability of governments and private investors to deliver massive AI computing hubs on a compressed timetable.
The bloc’s original ambition centred on up to five AI gigafactories backed by a €20 billion public-private financing facility. These centres were designed to house more than 100,000 advanced AI processors each, with the computing power needed to train frontier models and support strategic industrial applications. The updated approach keeps that broad objective intact but gives consortia more room to start with lower power capacity and expand as demand, funding and permits mature.
The change marks a pragmatic recalibration rather than a formal retreat from Europe’s AI sovereignty agenda. Officials still want to reduce dependence on US cloud providers, strengthen domestic compute capacity and support start-ups, universities and industry groups that lack access to large-scale AI infrastructure. Yet the new tender shape acknowledges that Europe’s energy and planning systems are not yet aligned with the size and speed of the build-out seen in the US, where hyperscalers have committed tens of billions of dollars to AI campuses.
The Commission’s AI Continent plan sets a target of mobilising €200 billion for AI investment and tripling the bloc’s data centre capacity within five to seven years. The gigafactory programme sits alongside 19 AI factories already selected or operating across Europe, which are built around supercomputing capacity under the EuroHPC framework and intended to provide access for smaller companies, researchers and public-sector users.
The smaller-scale tender option also responds to a sharp distinction between political ambition and market execution. Last year, 76 expressions of interest were submitted across 16 member states and 60 potential sites, suggesting strong early appetite. Since then, the field has been shaped by harder questions: who pays for the chips, how much power can be guaranteed, whether national governments can commit co-financing, and how quickly the sites can secure environmental, grid and land approvals.
Power availability has become one of the most difficult constraints. AI data centres require dense, reliable electricity supplies and advanced cooling systems, while several European markets are already struggling with grid congestion and high industrial energy prices. Operators also face scrutiny over water use, renewable energy procurement and whether public funding should support infrastructure that may benefit a small group of large technology users.
The new model lets bidders choose between capital-support and off-take structures, with public authorities receiving a share of computing capacity over a five-year period. It also requires proposals to spell out the phasing of investment, public support requested, maximum capacity and financial ceilings. That structure is designed to limit the risk of overpromising while giving governments clearer control over what public money buys.
Several countries are still moving aggressively. Spain has approved €719 million for an AI gigafactory project and hopes to align it with EU financing. France has been positioning itself as a major AI infrastructure base, helped by nuclear power, state support and private-sector plans that include multibillion-euro data centre and research campuses. Germany’s telecoms and industrial groups are also exploring large AI data centre projects tied to European funding.
The competitive backdrop remains unforgiving. Europe has strong research institutions, open-source communities and industrial users in sectors such as healthcare, manufacturing, automotive, finance and climate modelling. Its weakness lies in access to large and affordable compute. Three non-European hyperscalers control more than 70 per cent of the region’s cloud market, while the share held by European providers has fallen from 29 per cent in 2017 to around 15 per cent.