SoftBank's €75 Billion Bet on AI Data Centers in France: What It Means
SoftBank Group announced plans to invest €75 billion (about $87 billion) to build AI data center capacity in France—the company's largest infrastructure investment in Europe. The announcement came at a summit hosted by President Emmanuel Macron in 2026.
The investment will be rolled out in phases. In the first phase, SoftBank will spend €45 billion to deliver 3.1 gigawatts (GW) of computing power by 2031 across three locations in the Hauts-de-France region: Dunkirk, Bosquel, and Bouchain. These facilities will handle both AI training (the initial learning process for AI models) and inference (running trained models to make predictions). Eventually, the full commitment will reach 5 GW.
What Does 5 GW Actually Mean?
To put this in perspective: a typical large data center operates between 100 and 300 megawatts (MW). SoftBank's 5 GW commitment is equivalent to roughly 15 to 50 conventional data centers—though AI-optimized facilities pack more computing power into each facility, so the comparison isn't perfect. The sheer scale is significant enough to reshape how AI computing resources are distributed across Europe.
SoftBank is partnering with a company called Sesterce to develop a 1 GW facility in Bosquel. The Dunkirk site has a dual purpose: it will host data center operations and also include advanced manufacturing capabilities (in partnership with Schneider Electric) to build some of the equipment locally rather than importing everything from elsewhere.
Why France specifically. The Hauts-de-France region offers several advantages: it sits near submarine cables that carry international data traffic, has existing power infrastructure, and offers a regulatory environment suited to large industrial projects. France also generates significant electricity from nuclear power, which matters for high-energy-consuming operations like AI data centers.
The Money and the Timeline
An €75 billion commitment is an enormous sum—it approaches SoftBank's total market value. The company has not fully disclosed how it will finance this, but CEO Masayoshi Son has historically used debt and asset sales to fund major projects. Word is that Son initially discussed figures as high as $100 billion with French officials, so the final €75 billion likely represents a negotiated figure.
The timeline is ambitious but drawn out: 3.1 GW by 2031, with the remaining 1.9 GW to follow in later phases. SoftBank has not specified when those later phases will be complete.
Why Now, Why This Big
This is SoftBank's first major European data center commitment, which marks a shift for a company that has traditionally focused on the United States, Japan, and Asia. Son highlighted President Macron's personal support as a factor in the decision. In his announcement, Son framed the investment as a race for infrastructure: "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."
That framing sits at the heart of SoftBank's broader strategy. The company has positioned itself as a major financier of AI development through its Vision Fund vehicles and direct investments—including stakes in Arm Holdings, which designs chips used in AI training and inference.
The broader pattern worth noting is that this kind of massive, forward-looking infrastructure commitment has precedent. When Amazon and Microsoft were building cloud data centers, they built capacity years ahead of when companies actually migrated their workloads to the cloud. They bet correctly that businesses would eventually move. SoftBank appears to be making a similar long-term wager: that AI computing demand will grow for years to come, justifying the construction of facilities now rather than waiting for demand to materialize first.
Technical Realities and Local Advantages
AI data centers differ from conventional cloud data centers in meaningful ways. AI training requires high-speed connections between the many computing chips working on the same problem simultaneously. AI inference—where trained models make predictions—prioritizes fast response times to users. The facilities SoftBank builds will need specialized design for these requirements.
France's regulatory environment offers particular benefits. European data protection rules require that some data be processed locally rather than sent overseas, which encourages companies to use French facilities. The country's nuclear power plants also provide relatively stable electricity costs—a major factor for any operation that runs continuously at high power.
The partnership with Schneider Electric positions SoftBank to tap into EU policies that encourage domestic technology manufacturing, which may include subsidies or tax breaks for equipment built in Europe rather than imported.
What This Signals to the Rest of the Industry
A commitment this large puts pressure on other big tech companies to accelerate their own European AI data center plans. Microsoft, Google, and Amazon have announced European expansions, but none have committed to a single country at anything close to the 5 GW scale.
The investment also validates France's effort to position itself as a European hub for AI infrastructure, competing with established data center centers in Ireland, the Netherlands, and Scandinavia.
For the broader AI industry, this signals that the current bottleneck—difficulty getting access to specialized computing power to train large language models—may begin to ease as private money pours into purpose-built facilities. More infrastructure capacity could accelerate AI development timelines.
The deeper reading here is about how SoftBank sees AI's future. A €75 billion commitment that won't deliver most of its capacity until the early 2030s indicates SoftBank believes we are very early in an AI adoption cycle—a multi-decade transformation, not a near-term bubble. The company is betting that demand for AI computing will continue growing at rates that justify massive capital commitments today, even though the facilities won't be fully utilized until years from now. That confidence, from an investor of SoftBank's scale and experience, carries weight in how the rest of the industry thinks about the next decade.
