SoftBank is putting up to €75 billion behind French AI data centers, turning Europe’s compute race into a much more serious capital contest.
SoftBank is no longer talking about artificial intelligence as a portfolio theme. It is trying to build the physical layer underneath it, and France has just become the next big testing ground for Masayoshi Son’s most ambitious bet since the Vision Fund era.
The Japanese group has committed to develop and operate 5 gigawatts of AI data center capacity in France, with a total investment of up to €75 billion. The first phase is expected to put €45 billion into 3.1 gigawatts of capacity in the Hauts-de-France region by 2031, with sites tied to Dunkirk, Le Bosquel and Bouchain. That is not a small corporate expansion. It is industrial policy with a private-sector balance sheet attached.
SoftBank announced the plan on May 30 as part of the Choose France summit hosted by President Emmanuel Macron. According to Reuters, Son also told La Tribune Dimanche that the first €45 billion would be invested over the next five years, while the broader French program would reach €75 billion. That matters because this is exactly the kind of investment France has been trying to attract: high-value infrastructure, energy-intensive operations and a stronger local role in the AI supply chain.
The reason France is suddenly attractive is not mysterious. AI data centers need electricity before they need anything else. France has nuclear power, industrial land, a developed grid and a government willing to treat data centers as strategic infrastructure rather than just another real estate category. For AI builders, that combination is becoming rare.
Son has made the energy point directly, saying France’s position as an energy producer and exporter was decisive. That tells you where the AI race has moved. The bottleneck is no longer just model talent or venture funding. It is power availability, permitting, cooling, chips, grid connections and the ability to bring enormous facilities online without waiting a decade.
The local design of the project is also important. SoftBank plans to work with Schneider Electric on an industrial production cluster at the Port of Dunkirk, including facilities for enclosures and data center power modules. EDF is tied to the Bouchain site. This is not simply a foreign investor renting land and plugging in servers. The pitch is that France can host the infrastructure and help manufacture part of the system around it.
For Europe, that is the real prize. Policymakers have spent years talking about technological sovereignty, often with more ambition than execution. Compute capacity is where that phrase becomes practical. If European startups, enterprises and research institutions cannot access affordable high-performance infrastructure on the continent, they remain dependent on American cloud platforms and Asian hardware supply chains no matter how many AI strategies governments publish.
SoftBank is stretching across three AI fronts
The open question is whether SoftBank can fund all of this at the scale Son is now describing. The company has already made huge commitments around OpenAI and AI infrastructure outside Europe. Bloomberg reported earlier this year that SoftBank had invested about $34.6 billion in OpenAI for an 11% stake as of December, and later reported that the group secured a $40 billion bridge loan tied partly to a $30 billion follow-on investment.
That does not make the French pledge impossible. SoftBank has a long history of using asset sales, financing structures and strategic partners to chase large technology cycles. But it does mean investors should separate the headline number from the construction reality. A pledge of up to €75 billion is a statement of intent. Turning it into working capacity means years of capital deployment, grid work, equipment procurement and local approvals.
This is where Son’s style cuts both ways. His biggest calls can look reckless before they look obvious. Alibaba made SoftBank famous. WeWork made it look undisciplined. Arm has helped restore credibility, especially as AI demand raises the strategic value of chip architecture. The French data center plan now sits inside that same pattern: a huge swing at the infrastructure layer of the next technology cycle.
For entrepreneurs and AI companies, the practical question is less dramatic but more useful. More European compute capacity could lower dependence on a small group of hyperscalers, give regulated industries better local options and make it easier for startups to build services that need serious inference capacity close to customers. It could also push rivals to move faster, from cloud providers to sovereign funds to specialist data center operators.
There is another side to that opportunity. Giant AI facilities will compete for power, water, skilled workers and public patience. Communities will ask what they get in return beyond construction jobs and political announcements. Governments will need to show that AI infrastructure can support local industry, not just serve as a warehouse for global model companies.
SoftBank’s France plan is current, large and credible enough to change the conversation, but it is still early enough that execution matters more than the headline. Watch the first phase in Hauts-de-France. If the Dunkirk, Le Bosquel and Bouchain sites move from announcement to construction on schedule, Europe’s AI infrastructure race will have a new benchmark. If they slip, the market will be reminded that in AI, money is only the start of the build.
Also read: A new switching device points to cooler and faster AI chips • Defense officials now see AI risk as a front-line security threat • GitHub is making Copilot costs harder for founders to ignore
