SoftBank’s France bet turns AI into an energy question

What this is about

SoftBank founder Masayoshi Son has, according to an AFP report dated 31 May 2026, announced plans to invest up to €75 billion in AI infrastructure in France. Of that amount, €45 billion is intended to go into data centers in Hauts-de-France by 2031. Industrial partner Schneider Electric is expected to support equipment and design.

This is not a routine data-center story. If the plans hold, it would be one of Europe’s largest AI infrastructure bets. The interesting part is not only the amount, but the reason: Son points to France’s position as an energy producer and exporter as a decisive location advantage. AI therefore becomes a very concrete question of power, cooling, land, supply chains and political permits.

What the investment actually does

The reported plan describes a first expansion stage of around 3.0 to 3.1 gigawatts of data-center capacity. Later, according to the cited figures, capacity could rise to as much as 5.0 gigawatts. France reportedly had roughly 1.5 gigawatts of installed data-center capacity at the end of 2025. SoftBank’s bet would therefore not merely expand the market; it would change its scale.

The first sites are expected to be in northern France, including Dunkirk, Cambrai and Amiens. Schneider Electric is named in the reports as a partner for design, electrical equipment and an industrial production structure at the port of Dunkirk. That matters because AI data centers need more than GPUs. They need transformers, power distribution, cooling, fire protection, grid connections, maintenance teams and a supply chain that can scale for years.

Why it matters

Europe has talked about technological sovereignty for years. In AI, however, that sovereignty is not decided only by models, research groups or regulatory texts. It is also decided by whether European companies can obtain affordable compute capacity in Europe. Without capacity, training and inference workloads move to wherever power, capital and chips are available.

France’s case shows the hard foundation of the AI economy. A country with relatively abundant low-carbon power generation, industrial land and political will can suddenly become an AI infrastructure location. At the same time, new conflicts appear: grid stability, water use, local acceptance, energy prices and the question of whether public infrastructure indirectly strengthens mostly global platforms.

For developers and companies, the practical consequence is clear: if Europe gains more of its own data-center capacity, latency, data residency and regulatory control can improve. But that does not happen automatically. The deciding factors are who can use the capacity, at what price and under which security and sustainability conditions.

In plain language

Think of AI infrastructure like a large bakery. A good recipe matters, but without ovens, power, flour, staff and delivery routes, no bread leaves the building. Europe has many AI recipes: research, startups, regulation. SoftBank’s plan is about the ovens.

When a region suddenly wants to build twice or three times as many ovens, the story is no longer just about bakers. Power lines must be strengthened, suppliers organized and neighbors convinced. That is why this story is bigger than an investment number.

A practical example

A European medical-imaging company wants to train a new model in 2028 and later process 2 million diagnostic requests per day. Today it might need to buy compute across several US cloud regions, legally secure data flows and live with varying latency.

If several gigawatts of AI capacity are actually built in France, the same company could run parts of training and inference closer to European customers. At 2 million requests per day, a few milliseconds per request matter, but so do privacy review paths, audit logs and resilience. The infrastructure does not solve the product problem. It makes it possible for European providers to compete in the right weight class.

Scope and limits

• The announced amount is a plan, not finished infrastructure. Permits, grid connections, construction costs and chip availability can change both scope and timeline.
• More data centers do not automatically mean more European control. If capacity is mainly occupied by a few global platforms, dependence remains.
• Energy remains the bottleneck. Even lower-carbon electricity is not unlimited, and local communities will judge noise, land use and grid load.

The key open question is therefore not only whether SoftBank builds. It is what conditions Europe attaches to this new AI infrastructure so that it truly strengthens research, midsize companies and safe applications.

SEO & GEO keywords

SoftBank, Masayoshi Son, France, AI data centers, Hauts-de-France, Dunkirk, Schneider Electric, AI infrastructure, data centers, European AI sovereignty, energy grid, Choose France