SoftBank plans up to €75 billion investment in French AI centers

SoftBank plans to invest up to €75 billion ($87 billion) to build 5 gigawatts of AI data center capacity in France, including an initial €45 billion phase for 3.1 gigawatts in Hauts-de-France by 2031. The project would expand SoftBank's AI infrastructure footprint in Europe and reinforce France as a major hub for next-generation digital infrastructure, with Schneider Electric set to partner at Dunkirk. The announcement adds to SoftBank's broader global AI buildout, though financing capacity remains a key execution risk.

Analysis

This is less a single project announcement than a geopolitical industrial policy signal: France is trying to become the default non-U.S. landing zone for AI compute, and SoftBank is helping validate that thesis with a scale commitment large enough to force ecosystem reallocation. The near-term winners are not just the obvious infrastructure vendors, but the local power, grid, cooling, and permitting stack—projects at this size tend to pull forward capex for transformers, switchgear, fiber, and industrial real estate long before revenue shows up.

The second-order effect is competitive pressure on European cloud and AI startups: sovereign compute lowers the cost of retaining enterprise workloads in-region and reduces the need to route demand through U.S. hyperscalers. That said, the real bottleneck is not capital but execution time—multi-gigawatt builds face multi-year grid interconnect risk, water constraints, and municipal permitting friction, so the market should discount the full economic impact heavily until 2028-2031 rather than front-running headline optimism.

For SoftBank, the strategic upside is optionality: if these projects become the physical layer of AI demand, it diversifies the company beyond software-linked exposure and gives it leverage over capacity-constrained infrastructure economics. The contrarian view is that the announcement may be more about deal access and political signaling than fully financed deployment; if credit markets tighten or AI capex multiples compress, the funding stack becomes the most vulnerable link, not the technology thesis.

Against Oracle, the move is mildly negative at the margin if it signals AI capacity broadening away from a concentrated U.S. partnership model, but the more likely outcome is incremental demand for anyone able to monetize enterprise AI workloads. The market may be underestimating how much of the value accrues to picks-and-shovels suppliers rather than the project sponsor; the better trade is to own the enabling stack and fade the long-dated execution premium in the promoter.

Key Decisions for Investors

• Long Schneider Electric (SU.PA / SBGSY) on a 6-18 month horizon: best positioned to capture early order flow from power management, cooling, and industrial buildout; risk/reward is favorable because revenue recognition can start before full capacity comes online.
• Add to European grid/infrastructure beneficiaries via a basket trade: long SU.PA + long Siemens (SIE.DE) vs short a broad European software basket over 3-12 months; thesis is capex migrates to physical infrastructure while pure-play AI software valuations remain crowded.
• Avoid chasing Oracle (ORCL) on this headline; if the market starts pricing broad AI compute dispersion as dilution to centralized cloud incumbents, ORCL could underperform on a 1-3 month relative basis versus infrastructure names.
• Express a financing-risk view by buying long-dated puts on SoftBank Group (9984.T) or using a put spread out 6-12 months; the project pipeline is ambitious, but balance-sheet and credit-market constraints are the key reversal risk.
• If European data-center names sell off on permitting or power-supply concerns, buy the dip selectively in 3-6 month windows; the secular theme is intact, but execution delays will create entry points rather than thesis breaks.