SoftBank Bets Up to €75 Billion on French AI Data Centers — Europe's Biggest Infra Gamble

SoftBank will invest up to 75 billion euros (around 87 billion dollars) to build AI data centers in France, with a first phase of 45 billion euros for 3.1 gigawatts of capacity in the Hauts-de-France region. Announced on May 30, 2026 at the Choose France summit hosted by President Emmanuel Macron, it is SoftBank Group's largest AI infrastructure commitment in Europe. The full program targets 5 gigawatts of capacity, with the first sites at Dunkirk, Bosquel and Bouchain due for delivery in 2031.

What SoftBank Just Committed To

On May 30, 2026, SoftBank Group announced it will invest up to 75 billion euros — roughly 87 billion dollars — to build a fleet of AI data centers across France. The figure makes it, in SoftBank's own words, the group's largest AI infrastructure investment in Europe. The announcement landed at the Choose France summit, the annual foreign-investment showcase hosted by President Emmanuel Macron, where a single line on a slide can represent more capital than some national budgets.

The headline number is 75 billion euros, but the program is staged. The first phase is 45 billion euros for 3.1 gigawatts of data center capacity, concentrated in the Hauts-de-France region in the country's far north. The full program targets 5 gigawatts of capacity in total, meaning the announced 75 billion euros funds the headline ambition while the 45 billion euro phase one is the part with sites, timelines and shovels attached. We want to be precise here, because the two numbers travel together and get conflated constantly: 75 billion euros is the total program; 45 billion euros is phase one. They are not the same money.

Phase one names three sites: Dunkirk (Loon-Plage), Bosquel, and Bouchain, all in Hauts-de-France. SoftBank gave a delivery target of 2031 for that 3.1-gigawatt first tranche. That is a six-year build horizon for the initial capacity alone, which tells you how much of this is about pouring concrete and securing power rather than racking servers next quarter.

Masayoshi Son, SoftBank's founder and chief executive, framed it as a generational bet. "AI is entering a new era, and the countries that build the infrastructure for this transformation will shape the future of technology, industry and society," he said in the announcement. "SoftBank is proud to make this major commitment to France." French economy minister Roland Lescure described the deal as "a testament to President Emmanuel Macron's ambition to position France as a leading destination all along the AI value chain."

The Numbers at a Glance

For context on what 5 gigawatts means: it is roughly the steady-state electrical draw of several million homes, or the output of multiple large nuclear reactors running flat out. This is not a server room. It is an industrial power story dressed in the language of cloud computing.

Why Europe Has Been Losing the Infrastructure Race

The blunt context is that Europe has spent the last three years watching the AI compute buildout happen somewhere else. The United States and China have poured capital into gigawatt-scale clusters while Europe debated regulation. The continent has world-class researchers, a flagship lab in Mistral, and a deep industrial base — but it has not had the raw compute density to train and serve frontier models at the scale its rivals enjoy.

That gap is not abstract. When a European bank, hospital network, or government agency wants to run sensitive workloads on sovereign infrastructure, the honest answer for years has been that the biggest, cheapest, most available AI capacity sits in American hyperscaler regions. We have watched this dynamic play out in the chip layer too: even Mistral, France's frontier champion, has been candid that it still runs on Nvidia hardware while it explores building its own silicon. Compute sovereignty is a stated European ambition that the physical reality has repeatedly undercut.

SoftBank's 75 billion euros is the kind of single-stroke commitment that can move that needle. At 5 gigawatts, the program would add more dedicated AI capacity to France than the country has built in total to date. It is foreign capital — Japanese capital — solving a European problem, which is itself a telling detail about where the continent's domestic balance sheets stand on bets of this magnitude.

Why France, and Why Hauts-de-France

France is an unusually rational choice for an AI data center mega-project, and the reason is sitting in plain sight: electricity. France generates the majority of its power from nuclear plants, giving it some of the lowest-carbon and most price-stable grid power in the developed world. For a workload whose single largest operating cost is the electricity bill, and whose biggest political liability is its carbon footprint, a nuclear-heavy grid is close to ideal.

Hauts-de-France adds a second layer of logic. The region hugs the North Sea coast, which matters for two reasons. Coastal sites simplify cooling, since seawater and cool northern air reduce the energy spent on heat rejection. And the area sits near interconnections and offshore wind corridors that can feed additional power into the grid. Dunkirk in particular is an industrial port city with the land, the grid connections, and the political appetite for large infrastructure that a 3.1-gigawatt campus demands. Bosquel and Bouchain round out a cluster rather than a single mega-site, which spreads grid load and permitting risk.

This regional logic echoes a pattern we have tracked across the sector. The constraint on AI expansion has quietly shifted from chips to power. We saw it when NextEra moved to buy Dominion for 67 billion dollars in the largest US utility merger in history, a deal driven explicitly by AI power demand. We saw it again with GridCARE's 64 million dollar Series A, a startup whose entire pitch is using AI to find and unlock stranded grid capacity for AI data centers. SoftBank choosing nuclear-rich France is the same thesis expressed at the scale of a sovereign-level capital commitment.

The Energy Subtext Nobody Can Ignore

Five gigawatts is the number that should make every reader pause. To put it in perspective, that is comparable to the continuous output of three to five large nuclear reactors, or the electricity consumption of a mid-sized European city, dedicated to a single company's AI compute. Even phase one's 3.1 gigawatts would rank among the largest single industrial power commitments in recent French history.

This is where the story stops being a press-release celebration and becomes a genuine policy question. France's nuclear fleet is large but not infinite, and much of it is aging. Adding multiple gigawatts of baseload demand for AI competes directly with electrification of transport, heating, and industry — all of which the European Union is simultaneously pushing. A data center campus does not care that your neighbor wants to charge an electric car; it wants its gigawatts now, predictably, for decades.

SoftBank's choice of France is partly an attempt to get ahead of exactly this fight. By siting in a low-carbon grid with spare nuclear and growing offshore wind, the company can credibly argue its footprint is cleaner than an equivalent build in a gas-heavy or coal-heavy region. That argument carries weight in Europe, where the political tolerance for energy-hungry, high-carbon infrastructure is thin and getting thinner. The same week, SoftBank reportedly outlined a US data center in Ohio powered by a 9.2-gigawatt natural gas plant — a vivid contrast that underscores why the French project leans so hard on the low-carbon angle.

Grid resistance is no longer hypothetical. In the United States, community opposition to data center construction over water use, land, and electricity prices has begun to slow projects and sharpen permitting battles. Europe's planning culture is, if anything, more deliberate. A 2031 delivery target for phase one is not conservatism — it is a realistic read of how long it takes to secure the power, win the permits, and build at this scale even when a government is rolling out the red carpet.

SoftBank's Tangled Web: Investor, Customer, and Builder

To understand why SoftBank is making this bet, you have to follow its dual role in the AI economy. SoftBank is both an investor in and a customer of OpenAI. It has committed enormous capital to OpenAI's expansion and to the broader Stargate infrastructure ambition, and it consumes frontier model capacity for its own portfolio of bets. When a company is simultaneously funding the demand and building the supply, vertical infrastructure stops being a side project and becomes a strategy.

That positioning changes how we read the French campus. SoftBank is not a neutral landlord renting space to whoever shows up. It is a player with a direct interest in there being abundant, cheap, available compute for the AI labs it has bet on. Building 5 gigawatts in France is a way to manufacture the very supply its other investments depend on — and to capture the margin between raw power and rented compute along the way. It is the same vertical logic that has driven the largest AI players to lock up compute and energy directly, from Anthropic's deal for SpaceX's Colossus cluster to the hyperscalers' scramble for power-purchase agreements.

There is also a portfolio-management read. SoftBank's history is one of enormous, concentrated, sometimes spectacularly mistimed bets. The Vision Fund era taught the market that Son does not do small. A 75 billion euro infrastructure commitment, denominated in hard assets with multi-decade lives, is a very different risk profile from a venture stake in a pre-revenue startup. Data centers with secured power and signed offtake are, in theory, closer to a utility than a moonshot. Whether the demand materializes to fill 5 gigawatts on SoftBank's timeline is the open question — but the asset itself is real, ownable, and financeable in a way a cap-table line item is not.

The Macron-Son Diplomacy Play

Announcements like this do not happen by accident, and they do not happen without choreography. Choose France exists precisely to produce moments like this: a head of state standing beside a global financier as a record number lights up the room. For Macron, the 75 billion euros is a centerpiece of a years-long campaign to brand France as Europe's AI capital — a campaign that has included hosting the AI Action Summit, courting Mistral, and pitching France's nuclear grid as a competitive weapon.

For Son, the diplomacy cuts both ways. Aligning with the French state secures something money alone cannot buy at this scale: grid access, permitting cooperation, and political cover for a project that will consume gigawatts and reshape regional economies. A data center that needs 3.1 gigawatts by 2031 is, functionally, a partnership with the state whether it is called that or not. The state controls the grid connections, the environmental approvals, and the local politics. Announcing at Choose France, beside Macron, is SoftBank cashing in that relationship up front.

It also sends a signal to every other hyperscaler and sovereign fund weighing European exposure. France just demonstrated that it can attract the single largest AI infrastructure commitment in the continent's history. That is a competitive flex aimed as much at Germany, the United Kingdom, and the Gulf states as it is at the markets. Europe's AI infrastructure race now has a clear early frontrunner for capital attraction, and the leaderboard matters because capital, once committed at this scale, tends to anchor an ecosystem around it.

How This Stacks Up Against the Global Buildout

Seventy-five billion euros is a staggering headline, but it lands in a year of staggering headlines. The global AI infrastructure buildout has normalized numbers that would have been unthinkable two years ago. Hyperscalers are guiding to capital expenditures in the hundreds of billions of dollars. The Stargate program that SoftBank co-backs is measured in the hundreds of billions on its own. Against that backdrop, 75 billion euros is enormous for Europe and merely large for the global compute arms race.

The more useful comparison is regional. Europe's previous flagship data center commitments were typically measured in the single-digit billions for hundreds of megawatts. SoftBank's program is an order of magnitude larger in capacity and capital. It does not just add to Europe's AI infrastructure; it roughly redefines the ceiling of what a single European project can be. We have tracked the gigawatt-scale ambitions emerging elsewhere — including Anthropic's roadmap toward a 10-gigawatt compute footprint — and SoftBank's French build is Europe's first credible entry into that same weight class.

What is still missing is the demand-side proof. A 5-gigawatt campus is a bet that European and global AI demand will grow fast enough, and stay onshore enough, to fill it. The chip supply, the model improvements, and the enterprise adoption all have to keep compounding for this capacity to pay off. We have seen how fast the ground can shift in this market: labs rise and fall, model economics change overnight, and the line between essential infrastructure and stranded asset is thinner than the headline numbers suggest.

What It Means for European Businesses and Developers

For the developers, startups, and enterprises actually building on AI in Europe, the practical promise of SoftBank's program is cheaper, closer, more sovereign compute — eventually. A 5-gigawatt fleet sitting on French soil, on a low-carbon grid, changes the calculus for any organization that has been reluctant to send sensitive workloads to American hyperscaler regions for data-residency, latency, or regulatory reasons. Banks, hospitals, and public agencies bound by EU data rules stand to gain the most from large-scale capacity that lives inside the bloc.

The catch is the timeline. Phase one does not deliver until 2031, so none of this relieves the compute crunch facing European builders today. In the near term, the announcement is a signal more than a service: it tells founders and CIOs that betting on a European AI stack is becoming less of a gamble, because the infrastructure to support it is finally being funded at scale. That confidence effect can matter as much as the gigawatts themselves, shaping where European companies decide to build and host over the next several years.

Our Take

We have been tracking the AI infrastructure story closely all year, and SoftBank's French commitment is one of the most consequential single moves we have covered — not because of the headline number alone, but because of what it signals. The center of gravity in AI competition has moved from algorithms to atoms: from who has the best model to who controls the power, the land, and the multi-year build pipeline. SoftBank just planted a 5-gigawatt flag in Europe, on a nuclear-rich grid, with the French state as a partner. That is a serious, coherent bet.

What gives us pause is the same thing that gives us pause about the entire buildout. These commitments assume a demand curve that bends ever upward. They assume the power can be secured without triggering the kind of public backlash already slowing projects in the US. And they assume SoftBank, a company with a famously volatile track record on timing, can execute a six-year industrial build without the macro environment turning against it. The 2031 delivery date for phase one is honest about the difficulty — but a lot can change between a Choose France slide and a powered-on data hall.

For Europe, though, the strategic logic is hard to argue with. The continent needs sovereign-scale compute, it needs it on a clean grid, and it has not been able to fund it from domestic balance sheets. If Japanese capital is willing to build it on French soil with French power and French jobs, that is a deal most European policymakers will take. The energy questions are real and will get louder. But the alternative — continuing to rent the future from American hyperscalers — is the one outcome France and Europe have decided they can no longer accept.

What to Watch Next

The next milestones are unglamorous but decisive. Watch for the grid connection agreements and power-purchase deals that will turn 5 gigawatts from a slide into a commitment the French grid operator has to honor. Watch for the environmental and permitting filings at Dunkirk, Bosquel, and Bouchain, where local opposition or approval will set the real timeline. And watch whether SoftBank names anchor tenants — the AI labs or cloud providers that will actually fill the capacity — because a data center without committed demand is just an expensive building.

Watch, too, for the competitive response. If Macron's Choose France coup pulls the single largest AI commitment in European history, expect Germany, the United Kingdom, and the Gulf to counter with incentives of their own. Europe's AI infrastructure race just got a frontrunner and a benchmark. The interesting question for the next twelve months is not whether SoftBank's 75 billion euros gets fully spent — it will be staged over years — but whether it triggers the wave of follow-on capital that finally closes Europe's compute gap with the United States and China. We will be following every gigawatt of it.

Frequently Asked Questions

How much is SoftBank investing in French data centers?

SoftBank announced it will invest up to 75 billion euros, or roughly 87 billion dollars, to build AI data centers in France. The first phase accounts for 45 billion euros and 3.1 gigawatts of capacity. The 75 billion euros is the total program ambition; the 45 billion euros is the funded, sited first phase. The two figures are not the same money.

What is the total data center capacity SoftBank plans in France?

The full program targets 5 gigawatts of AI data center capacity across multiple phases. Phase one alone is 3.1 gigawatts, with a delivery target of 2031. Five gigawatts is comparable to the continuous output of several large nuclear reactors, which is why energy supply is central to the project.

Where will SoftBank build its French AI data centers?

Phase one names three sites — Dunkirk (Loon-Plage), Bosquel, and Bouchain — all in the Hauts-de-France region in the north of the country. The region was chosen for its coastal cooling advantages, grid interconnections, and proximity to France's low-carbon nuclear power and offshore wind corridors.

When was the SoftBank France investment announced?

SoftBank announced the investment on May 30, 2026, at the Choose France summit hosted by President Emmanuel Macron. Choose France is France's annual foreign-investment showcase, and the 75 billion euro commitment was its headline figure for 2026.

Is this SoftBank's biggest AI infrastructure investment in Europe?

Yes. SoftBank described the 75 billion euro program as its largest AI infrastructure investment in Europe. In terms of both capital committed and gigawatts of planned capacity, it is roughly an order of magnitude larger than Europe's previous flagship data center commitments.

Why did SoftBank choose France for its AI data centers?

France offers one of the lowest-carbon, most price-stable power grids in the developed world thanks to its large nuclear fleet. For AI data centers, where electricity is the dominant operating cost and carbon footprint is a key political liability, a nuclear-heavy grid is close to ideal. Hauts-de-France adds coastal cooling and grid interconnections.

What does Masayoshi Son say about the France investment?

Masayoshi Son, SoftBank's founder and chief executive, said: "AI is entering a new era, and the countries that build the infrastructure for this transformation will shape the future of technology, industry and society. SoftBank is proud to make this major commitment to France." French economy minister Roland Lescure called it a testament to Macron's ambition to make France a leader across the AI value chain.

How is SoftBank connected to OpenAI?

SoftBank is both an investor in and a customer of OpenAI, and it co-backs the broader Stargate infrastructure program. That dual role makes its French build a vertical-integration play: by building data center capacity, SoftBank can supply the very compute demand that the AI labs it has invested in depend on.

How much energy will SoftBank's French data centers use?

The full program targets 5 gigawatts of capacity, comparable to the output of three to five large nuclear reactors running continuously. Even phase one's 3.1 gigawatts ranks among the largest single industrial power commitments in recent French history. This is why grid connection agreements and power-purchase deals are the project's critical path.

When will the SoftBank French data centers be ready?

SoftBank gave a delivery target of 2031 for phase one's 3.1 gigawatts. That six-year horizon reflects how long it takes to secure power, win environmental permits, and build at gigawatt scale, even with strong government support. Later phases extending toward 5 gigawatts would follow on a longer timeline.

What does this mean for Europe's AI infrastructure race?

The commitment positions France as the early frontrunner for AI infrastructure capital in Europe, a continent that has lagged the United States and China on compute density. By attracting the single largest AI infrastructure investment in Europe's history, France set a benchmark that Germany, the United Kingdom, and the Gulf states will likely respond to with incentives of their own.

Is the difference between 75 billion and 45 billion euros important?

Yes, and the two are frequently confused. The 75 billion euros (around 87 billion dollars) is the total program ambition spanning the full 5 gigawatts. The 45 billion euros is phase one specifically — the funded tranche with named sites at Dunkirk, Bosquel, and Bouchain, 3.1 gigawatts of capacity, and a 2031 delivery target. Reporting that cites only one number tells half the story.