A Number That Reframes The Conversation
SoftBank has committed to develop and operate up to 5 gigawatts of AI data center capacity in France, with investment reaching as much as 75 billion euros, or roughly 87.5 billion dollars. Announced around the Choose France summit, the plan is not a single facility but a program, and its scale changes the frame of the discussion. We are no longer talking about a data center as a building. At 5 gigawatts, we are talking about a compute footprint whose power draw rivals a meaningful slice of a national grid, planned and financed as a multi-year industrial undertaking rather than a real estate project.
The ambition sits inside a familiar SoftBank register. Founder Masayoshi Son framed it in civilizational terms, saying that "AI is entering a new era, and the countries that build the infrastructure for this transformation will shape the future." That is the kind of statement that invites skepticism given SoftBank's history of enormous headline commitments, and enterprises should read the number as a ceiling and an aspiration rather than a guaranteed spend. But even a fraction of 5 gigawatts delivered on European soil would materially change the supply of high-density AI compute available to the region's businesses.
Phase One Is Concrete, The Rest Is Ambition
Beneath the eye-catching total sits a first phase that is specific enough to take seriously. SoftBank has earmarked 45 billion euros to deliver 3.1 gigawatts of AI data center capacity in the Hauts-de-France region, spread across three named sites: Dunkirk at Loon-Plage, Bosquel, and Bouchain, with a target completion of 2031. Naming the sites and attaching a phased budget is what separates this from the vaporware that has plagued the sector, where gigawatt promises evaporate the moment power and permitting realities intrude. The details suggest planning has advanced beyond the press release.
That said, the timeline deserves emphasis for anyone tempted to treat this as near-term capacity. Completion by 2031 means the bulk of these gigawatts will not serve a customer workload for years, and the gap between announcement and energized racks is where most mega-projects lose momentum. Enterprises evaluating European compute options over the next 18 to 24 months should not plan around SoftBank's France buildout at all. The relevance is strategic and long-range: it signals where a major operator intends to place capacity, and it puts pressure on hyperscalers to match sovereign-friendly commitments in the same market.
Power Is The Product, And France Has It
The most important detail is not the money, it is the electricity. State-owned utility EDF is providing SoftBank a former power plant site in Bouchain for development, tying the project directly into France's nuclear-heavy generation base. This is the crux of why France, specifically, is attracting this investment. The binding constraint on AI infrastructure in 2026 is no longer access to chips or capital, it is access to large, reliable, low-carbon power on a predictable timeline. France's nuclear fleet gives it something scarce: abundant baseload electricity that is both plentiful and comparatively clean.
We have argued before that the AI buildout is quietly becoming an energy story wearing a technology costume, and this deal makes the point explicit. Locating gigawatt-scale compute next to a decommissioned power plant is a way of solving the interconnection problem before it starts, inheriting grid infrastructure and generation proximity that would take years to build from scratch. For countries competing to host AI capacity, the lesson is blunt. The winners will be the ones that can offer developers power, not tax breaks, and France is playing the one card that hyperscaler capital cannot simply manufacture elsewhere.
Building The Supply Chain, Not Just The Buildings
SoftBank is also trying to localize the supply chain rather than import it. The company will partner with Schneider Electric to develop an industrial production cluster at the Port of Dunkirk, including a SoftBank-operated facility to manufacture data center enclosures and a Schneider Electric site to integrate power modules. This is a notable departure from the standard model, where an operator buys globally sourced equipment and assembles it on site. Manufacturing enclosures and power systems near the data centers themselves is an attempt to compress lead times and capture more of the economic value locally.
Schneider Electric chief executive Olivier Blum framed the technical stakes precisely, saying that "the challenge of AI is to deliver both speed and energy efficiency at scale." That is the quiet engineering reality behind the political theater. High-density AI racks are punishing to power and cool, and the components that manage electricity distribution are as much a bottleneck as the accelerators inside the servers. By building a power-equipment cluster alongside the compute, SoftBank is acknowledging that the hard part of a gigawatt data center is not the silicon, it is everything that keeps the silicon fed and cool without wasting energy the grid cannot spare.
What It Signals For Enterprise Cloud Strategy
For European enterprises and their technology leaders, the strategic signal outweighs the immediate utility. A commitment of this size, backed by a national utility and a domestic industrial partner, is a statement that sovereign AI compute is becoming a category of infrastructure that governments will help underwrite. French minister Roland Lescure called the decision "testament to President Emmanuel Macron's ambition to position France as a leading destination," and that political framing matters. It suggests future capacity in the region may come with data-residency and governance characteristics that appeal to regulated industries wary of dependence on non-European providers.
The practical takeaway is to watch the power map, not the press releases. Where large, clean, dependable electricity is available, AI capacity will follow, and the organizations that plan their multi-year cloud and data strategies around energy geography will be better positioned than those chasing the newest chip. SoftBank's France bet may or may not fully materialize at 5 gigawatts. Either way, it confirms the direction of travel: AI infrastructure is consolidating around a handful of power-rich regions, and sovereignty, energy and compute are becoming a single strategic question that enterprises can no longer treat separately.
