France lands one of Europe's largest AI infrastructure programs
SoftBank is committing up to EUR 75bn (US$87bn) to build and run 5 gigawatts of AI data-center capacity across France - the group's biggest AI infrastructure investment in Europe. The plan is aimed at supplying high-performance compute to AI developers, cloud providers, enterprises, and public institutions.
The shape of the build-out
- The first phase is a EUR 45bn (US$52bn) investment delivering 3.1 GW in the Hauts-de-France region, with initial sites planned for Dunkirk and further locations at Bosquel and Bouchain, targeted for delivery in 2031.
- It was unveiled at the 2026 Choose France summit hosted by President Macron, signaling a coordinated industrial-policy push.
- French ministers tied their backing explicitly to digital sovereignty, jobs, and strengthening national infrastructure.
Power and manufacturing are central
Energy availability is the gating factor for hyperscale AI, and the program is built around it. EDF is involved in the Bouchain development, supplying what it frames as competitive, sovereign, low-carbon electricity from a repurposed former industrial site. Meanwhile, SoftBank is forming an industrial partnership with Schneider Electric at the Port of Dunkirk, building a cluster of two facilities - one making enclosures, the other integrating data-center power modules - that pairs SoftBank's robotics and automation with Schneider's energy-tech expertise. The companies argue the cluster will localize and harden European supply chains, enabling faster, more energy-efficient deployment.
The bigger bet
Masayoshi Son framed it as a conviction play: the countries that build the infrastructure for the AI transition will shape the future of industry and society, and France - with its industrial base, talent, and national ambition - is positioned to be a leading European hub. Beyond construction, SoftBank expects the program to create thousands of jobs across engineering, energy systems, robotics, and advanced manufacturing, and plans university and training partnerships to feed the skills pipeline. It's a vivid example of how the AI race is increasingly being run on power and real estate, not just models.
