The simplest way to understand SoftBank’s decision is this: AI infrastructure is no longer only a technology business. It is becoming an energy business. SoftBank is not choosing France because France suddenly became the center of global software innovation. It is choosing France because France has what AI data centers desperately need: large-scale electricity, nuclear power, industrial land, grid operators, energy companies, and a government willing to treat AI infrastructure as a national industrial project.

SoftBank Group has announced a plan to develop up to 5 gigawatts of AI data center capacity in France. The total investment could reach €75 billion. The first phase alone is expected to involve about €45 billion and aims to deliver 3.1 gigawatts of AI data center capacity in the Hauts-de-France region by 2031.

The first sites include Dunkirk, specifically Loon-Plage, as well as Bosquel and Bouchain. These are not random locations. They sit in northern France, where ports, industrial land, old power infrastructure, transmission access, and manufacturing networks already exist.

This is why the project matters. AI data centers are no longer small server rooms. They are becoming industrial-scale power consumers. At 5 gigawatts, the planned capacity is comparable to the output scale of several nuclear reactors. In the AI era, the question is no longer only “Who has the best model?” It is also “Who has enough electricity to run the model?”

AI used to look like a software race. Now it is becoming a race for electricity, land, cooling, transmission lines, and permits.

Why France?

At first glance, France may not look like the obvious place for a massive AI data center project. People associate France with wine, luxury goods, tourism, aerospace, nuclear power, and manufacturing. They do not usually think of France as the first destination for AI infrastructure.

But from SoftBank’s point of view, France has one decisive advantage: electricity.

France is one of Europe’s most important electricity producers and exporters. Its power system is built around a large nuclear fleet. Nuclear power gives France a relatively stable, large-scale, low-carbon electricity base. When renewable power is added, most of France’s electricity generation comes from low-carbon sources.

This matters because AI data centers consume enormous amounts of electricity. The cost of power directly affects operating costs. The carbon profile of power affects environmental and regulatory pressure. The reliability of power affects whether expensive AI servers can run continuously.

For SoftBank, France is attractive because it can offer a combination that many countries cannot easily provide: large electricity supply, low-carbon generation, national energy planning, industrial policy, and public-sector coordination.

France’s AI advantage is not only talent. It is nuclear electricity.

The real AI bottleneck is power

The AI industry used to talk mostly about GPUs, foundation models, training data, and cloud platforms. Those are still important. But the infrastructure bottleneck has shifted.

Advanced AI data centers need high-performance chips, but chips are only useful if they can be powered, cooled, connected, and housed. A site with land but no grid connection is not enough. A site with power but no cooling plan is not enough. A site with cheap electricity but slow permitting is not enough.

This is why countries with strong energy systems are becoming more attractive. AI data centers require long-term power contracts, substations, transmission upgrades, backup systems, water or air-cooling solutions, and regulatory approval.

In that sense, AI infrastructure is starting to resemble heavy industry. The competitive edge is not only code. It is industrial execution.

SoftBank’s move reflects this change. The company is not simply leasing office space for servers. It is trying to build a European AI infrastructure hub around energy availability and industrial coordination.

Why Dunkirk and northern France matter

Dunkirk is a strategically logical location. It is a port and industrial region with existing energy and heavy-industry infrastructure. It has space for large-scale facilities. It is connected to European logistics networks. It also sits in a region that France wants to revitalize through advanced manufacturing and energy-transition investment.

The project is expected to involve Schneider Electric, which is strong in data center power equipment and modular electrical systems. EDF, France’s state-backed electricity giant, is also part of the broader energy logic around the project.

This is important because a data center of this size cannot be built by a technology investor alone. It needs electrical equipment suppliers, energy companies, grid operators, construction firms, cooling-system providers, regulators, and local governments.

In the United States, these pieces are often fragmented across private utilities, state regulators, landowners, local politics, and private developers. In France, the state can coordinate more directly. That does not mean France is always faster. But for a project of this size, state coordination can be an advantage.

A 5 GW AI project is not just a tech project. It is a grid project, an industrial-land project, and a national energy-planning project.

France has surplus electricity, but that does not solve everything

France has recently been a major net exporter of electricity. That gives the country an advantage. But “having electricity” and “delivering electricity to an AI data center” are not the same thing.

A data center needs power at a specific location, at a specific voltage, through a reliable grid connection, with backup capacity and cooling infrastructure. It also needs enough transmission capacity so that a giant new power consumer does not destabilize the local system.

So the challenge is not only whether France produces enough electricity overall. The challenge is whether the grid can deliver enough electricity to the exact sites where SoftBank wants to build.

This is why grid operators and power infrastructure companies matter. AI infrastructure requires transformers, substations, switchgear, high-voltage connections, cooling systems, backup generation, and long-term maintenance.

In other words, SoftBank is not just buying power. It is buying into an entire energy and industrial ecosystem.

Why low-carbon electricity matters for AI companies

AI companies face a difficult contradiction. They promise a more intelligent and efficient future, but their data centers consume huge amounts of electricity. That creates political and environmental pressure.

If AI data centers are powered by coal or gas-heavy grids, their carbon footprint becomes harder to defend. If they are powered by low-carbon electricity, the political case becomes easier.

This is where France is strong. Its nuclear-heavy system gives AI investors a way to claim relatively low-carbon computing capacity. That is valuable in Europe, where climate policy, corporate sustainability targets, and public scrutiny are more intense than in many other regions.

This also helps France’s own industrial strategy. By combining nuclear power with AI infrastructure, France can argue that it is not only hosting data centers. It is building a low-carbon digital industrial base.

In the AI era, cheap electricity is valuable. Cheap low-carbon electricity is even more valuable.

Macron’s industrial strategy is part of the story

The project also fits President Emmanuel Macron’s broader strategy. France wants to attract foreign investment, strengthen European technological sovereignty, and turn its energy base into an advantage for next-generation industries.

The Choose France summit has become a platform for announcing major investment commitments. SoftBank’s plan is one of the most important because it directly connects France’s energy policy with Europe’s AI ambitions.

Europe has worried for years that it depends too heavily on U.S. cloud providers, U.S. AI platforms, U.S. chips, and U.S. digital infrastructure. Building large-scale AI data center capacity in France helps Europe claim that it can host part of the AI stack on its own soil.

Of course, this does not make Europe fully independent. The GPUs and advanced chips may still come from companies such as Nvidia. Many AI models and customers may still be American. Financing may depend on global capital.

But location still matters. If the computing infrastructure is built in Europe, it creates jobs, grid investment, industrial demand, local tax revenue, and strategic leverage.

What SoftBank gains from the project

For SoftBank, this project is part of a much larger AI strategy. The company has already made major investments in OpenAI and has been involved in large AI infrastructure plans such as Stargate in the United States.

SoftBank’s basic thesis is that AI demand will keep expanding and that the world will need massive computing capacity. If that thesis is correct, owning or controlling data center capacity becomes strategically valuable.

The logic is similar to owning railways during an industrial boom or owning telecom towers during a mobile boom. The most visible value may come from AI models and applications, but the infrastructure underneath can also become a major profit pool.

SoftBank also gains geographic diversification. If the company only builds AI infrastructure in the United States, it is exposed to U.S. power-market bottlenecks, U.S. permitting, U.S. political risk, and U.S. customer concentration. France gives it a European base.

That matters because AI demand will not be limited to Silicon Valley. European governments, companies, research institutions, defense contractors, banks, pharmaceutical firms, and industrial groups will all need computing capacity.

Why this could reshape Europe’s data center map

Europe’s data center market has traditionally been concentrated around places such as Frankfurt, London, Amsterdam, Paris, and Dublin. But AI is changing the map.

Older data center hubs were built around connectivity, finance, internet exchanges, cloud demand, and proximity to customers. AI data centers care more about power scale. Training workloads do not always need to sit next to financial exchanges or dense urban centers. They need cheap, reliable, massive electricity.

This means industrial regions with strong power infrastructure can become more important. Former steel, port, power, or manufacturing regions may gain new value as AI infrastructure sites.

Northern France fits that pattern. It has industrial land and energy infrastructure. It is also close enough to major European markets. If the SoftBank project succeeds, it could encourage more AI infrastructure investment in similar regions.

AI may give old industrial regions a new role: not as steel towns, but as power-rich computing hubs.

The project also carries risks

The biggest risk is execution. A €75 billion, 5 GW plan is enormous. Projects of this size require financing, construction capacity, grid upgrades, equipment supply, permits, land-use coordination, and long-term customers.

The second risk is demand. AI computing demand is growing quickly, but the industry is also debating whether current investment plans are too aggressive. If AI monetization does not grow fast enough, some data center projects could face lower-than-expected utilization.

The third risk is electricity politics. Even if France has strong electricity production, local communities may question whether massive data centers should consume power that could otherwise support households, factories, electrification, hydrogen, rail, or other industrial uses.

The fourth risk is grid congestion. A country can export electricity in aggregate and still face local bottlenecks. If grid upgrades are delayed, data center timelines can slip.

The fifth risk is financing structure. Large data center projects often rely on debt, project finance, customer commitments, and long-term power contracts. If interest rates rise or AI infrastructure valuations weaken, financing terms can become harder.

So the project is impressive, but not risk-free. The announcement is only the first step. The real test will be whether capacity is actually delivered on schedule and filled by paying customers.

What this means for the global AI race

SoftBank’s France plan shows that the global AI race is becoming more physical. The early phase was about algorithms. Then the race moved to GPUs. Now it is moving to data centers, electricity, cooling, power equipment, and national energy policy.

This changes which countries can compete. A country does not need to have the world’s top AI model to become important. It can become important if it has low-carbon power, grid capacity, industrial land, political coordination, and capital access.

That is why France is relevant. France may not dominate consumer AI applications. But it can become a major European AI infrastructure base.

The same logic applies elsewhere. The United States has scale and technology. The Gulf states have capital and energy. Japan has industrial partners and strategic capital. France has nuclear power and European location.

AI infrastructure is becoming a geopolitical asset. Data centers are no longer just buildings full of servers. They are strategic industrial facilities.

The next AI superpower may not be the country with the most apps. It may be the country that can power the most computing.

What to watch next

The first thing to watch is financing. The headline number is up to €75 billion, but large infrastructure plans depend on how equity, debt, project finance, and customer contracts are structured.

The second is grid connection. The most important milestone is not only land acquisition. It is whether transmission capacity, substations, and high-voltage connections can be delivered on schedule.

The third is chip supply. A giant AI data center needs GPUs or specialized accelerators. Even with electricity and land, the project must secure advanced chips.

The fourth is customer demand. Europe needs AI computing capacity, but SoftBank must still attract hyperscalers, AI model companies, enterprises, or government-backed workloads.

The fifth is French nuclear policy. France’s ability to host large-scale AI infrastructure depends on maintaining and expanding reliable low-carbon electricity. Reactor availability, grid investment, and new nuclear plans will all matter.

The sixth is local acceptance. Data centers create jobs and investment, but they also consume power and land. Local communities may support industrial revival, but only if benefits are visible.

Conclusion: AI has become an electricity game

SoftBank’s French AI data center plan is one of the clearest signs that the AI race is entering a new stage. The question is no longer only who builds the smartest model. It is who can build the physical infrastructure needed to run those models at scale.

France is attractive because it has nuclear-heavy, low-carbon electricity, a large power system, strong industrial regions, public-sector coordination, and a government eager to turn energy strength into AI advantage.

SoftBank is betting that Europe will need enormous AI computing capacity and that France can provide one of the few credible locations for that scale. If the project succeeds, northern France could become one of Europe’s most important AI infrastructure hubs.

But the challenge is just as large as the ambition. Building 5 GW of AI data center capacity requires more than money. It requires electricity, grid upgrades, chips, cooling, customers, financing, permits, and political support.

The simplest way to read SoftBank’s France investment is this: AI infrastructure is becoming heavy industry, and France’s nuclear power is becoming a strategic digital asset.