Summary

• SoftBank plans up to €75bn of AI data centre investment in France, starting with 3.1GW of capacity in Hauts-de-France.
• The project links AI infrastructure, low-carbon energy, industrial redevelopment, and Europe’s attempt to reduce dependence on overseas compute.
• France is turning power availability and regional industrial capacity into a national AI infrastructure pitch.

SoftBank Group plans to develop and operate 5GW of AI data centre capacity in France, placing the north of the country at the centre of one of Europe’s largest proposed artificial intelligence infrastructure buildouts.

The Japanese technology investor said the programme could represent investment of up to €75bn. Its first phase involves an initial €45bn commitment to deliver 3.1GW of AI data centre capacity in the Hauts-de-France region, as part of the 2026 Choose France summit hosted by President Emmanuel Macron.

Behind the headline figure is a more physical contest for AI capacity. Countries are competing on land, power, permitting, grid access, cooling, construction capability, and political certainty, while model developers and cloud customers look for infrastructure that can support large-scale training and inference. France is trying to turn its electricity base, including its nuclear-heavy generation mix, into an advantage as demand for compute rises across Europe.

SoftBank said the project is intended to support sovereign AI capacity and provide infrastructure for large-scale model training and inference. Schneider Electric is expected to supply modular solutions, while EDF is making a former power plant site available for conversion into data centre infrastructure.

The plan shows how AI capital expenditure is moving beyond software platforms and model development into heavy industrial systems. The next phase of competition is being built around substations, grid connections, industrial parks, and cooling systems, as much as around research labs and foundation models.

For France, the project gives the government a flagship foreign investment win as European policymakers try to narrow the AI infrastructure gap with the US and China. SoftBank, meanwhile, extends a global AI strategy that already spans chips, model companies, and infrastructure, while gaining a much larger physical presence in Europe’s AI supply chain.

The economics remain demanding. A 5GW AI data centre programme would require enormous power availability, long construction timelines, and sustained demand from model developers, cloud customers, enterprises, and governments. The sector is also facing sharper scrutiny over energy use, water demand, land consumption, and pressure on local grids, especially where AI facilities compete with housing, industry, and public services for infrastructure capacity.

The French project sits inside a wider European debate over digital sovereignty. Policymakers want more control over the infrastructure that processes public-sector, industrial, defence, and enterprise data, but sovereign infrastructure cannot be built through policy declarations alone. It needs capital, supply chains, grid capacity, specialist construction skills, and customers willing to pay for local or European-hosted compute.

That creates an uncomfortable but practical tension. Europe wants more autonomy, yet much of the capital and technology needed to build at hyperscale still comes from outside Europe. SoftBank’s plan strengthens France’s physical AI infrastructure while relying on a major Japanese investor to deliver it.

The project will now be judged by delivery rather than scale of ambition. If the first 3.1GW phase comes online by the planned 2031 horizon, France will have a stronger claim to being a European AI infrastructure hub. If permitting, grid constraints, local opposition, or market demand slow the buildout, the announcement will join a growing pile of AI infrastructure promises waiting for hard evidence.