SoftBank’s €75 Billion France Bet Signals Japan’s Hedge in Global AI Race
€45 billion first-phase commitment for 3.1 GW of nuclear-powered data centers positions France as Europe's AI infrastructure hub while Japanese capital diversifies beyond US-China tech rivalry.
SoftBank Group announced a €45 billion first-phase investment to build 3.1 GW of AI data center capacity in northern France by 2031, marking Europe’s largest single AI infrastructure commitment and positioning France as a neutral ground in the US-China tech rivalry.
The investment, part of a total €75 billion program targeting 5 GW across France, leverages the country’s nuclear-dominated electricity grid to deliver low-carbon baseload power at scale, according to SoftBank Group. Three facilities in Dunkirk, Bosquel, and Bouchain will anchor the Hauts-de-France region’s transformation into a European AI hub, with partnerships confirmed between SoftBank, state-owned utility EDF, and industrial automation firm Schneider Electric.
France’s strategic advantage rests on its nuclear fleet, which supplies 60-75% of daily electricity generation and enabled the country to export 90 TWh of decarbonised power in 2024—the highest level in over two decades, per DataCenterDynamics. This surplus positions France against grid-constrained rivals: Ireland capped data center construction in Dublin in 2022 due to power shortages, while the Netherlands faces similar transmission bottlenecks.
Geopolitical Hedging Drives Capital Allocation
SoftBank CEO Masayoshi Son framed the investment as a response to accelerating US and Chinese AI buildouts, stating Europe and Japan “have to also go fast, not to be left out,” in remarks to CNBC. Son characterised the AI Infrastructure wave as “probably 50x bigger than dot-com” during a June 1 briefing, signalling his belief that the current cycle dwarfs prior technology investment eras.
The timing reflects Japanese strategic calculus. With US-China tech decoupling constraining cross-border capital flows and export controls tightening on advanced semiconductors, SoftBank’s European pivot creates a third infrastructure option outside the bipolar competition. The commitment also follows Son’s April 2026 meetings with President Emmanuel Macron in Tokyo, where an initial €100 billion figure was floated before being moderated to €75 billion, Bloomberg reported.
“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.”
— Masayoshi Son, Chairman and CEO, SoftBank Group
SoftBank shares rose 14% on May 31 following the announcement, extending year-to-date gains to 70%. The stock surge reflects investor confidence that Son’s infrastructure thesis—positioning SoftBank as an ‘AI landlord’ rather than a pure venture capital player—can capture value from surging compute demand projected to reach 945 TWh globally by 2030, up from 415 TWh in 2024, according to International Energy Agency estimates cited by Nuclear Business Platform.
France’s Digital Sovereignty Play
Macron has positioned the SoftBank deal as validation of France’s EU digital autonomy strategy, leveraging nuclear capacity to avoid the renewable intermittency challenges facing German and Spanish data center markets. “Thanks to our nuclear plants, we have the ability to open Data Centers, to build computing capacity, to be at the heart of the artificial intelligence challenge,” Macron stated, per Global Banking & Finance.
The investment arrives as the EU AI Act enters full enforcement on August 2, 2026, with high-risk system regulations already active since February 2, per AI CERTs News. France’s regulatory streamlining for strategic AI projects contrasts with permitting gridlock elsewhere: the US has 2,600+ GW of generation and storage projects awaiting grid connection approvals, while Ireland and the Netherlands have imposed de facto data center moratoria.
A November 2025 Accenture study found 62% of European organisations seeking sovereign AI solutions, with adoption rates reaching 72% in Germany, Ireland, and Denmark. The push reflects concerns over US hyperscaler dominance and Chinese technology transfer risks, creating political momentum for European-domiciled infrastructure even when operated by non-EU capital.
Project Financing Model Raises Execution Risk
SoftBank will rely primarily on project financing and long-term customer offtake agreements rather than balance sheet capital, according to RCR Wireless. This model—common in energy infrastructure but untested at €75 billion scale for AI data centers—requires securing anchor tenants willing to commit multi-year capacity contracts before construction financing closes.
The Port of Dunkirk industrial cluster, which will host the first facility at Loon-Plage, includes plans for a robotised manufacturing plant to support ongoing operations. The project claims to generate 15,600 jobs, though that figure includes construction, operations, and indirect supply chain roles across multiple sites through 2031.
| Country | Primary Constraint | Status |
|---|---|---|
| France | Permitting speed | Streamlined for strategic projects |
| Ireland | Grid capacity (Dublin) | Construction cap since 2022 |
| Netherlands | Transmission limits | Regional moratoria active |
| Germany | Renewable intermittency | Requires storage co-investment |
Hedi Ollivier, Director of Development for EMEA at Colt Technology Services, noted that “the power grid is very strong and in terms of network, we’re very well connected” in France, adding that “there’s plenty of land available” compared to constrained sites in Ireland and the Netherlands, per DataCenterDynamics.
Competitive Dynamics with US Hyperscalers
The France buildout positions SoftBank to compete for European AI workloads currently dominated by US hyperscalers—Microsoft, Google, Amazon—whose combined European data center investments exceeded $30 billion in 2024-2025. However, those firms benefit from vertical integration: they control cloud platforms, AI models, and customer relationships, while SoftBank must secure third-party tenants to fill capacity.
Son’s comment that SoftBank is “expanding a lot in the U.S.” and can “make France the center of Europe” suggests a hub strategy: using French nuclear power as a cost advantage to attract workloads that would otherwise deploy in Virginia or Ireland, then leveraging geographic proximity to serve EU markets under data sovereignty requirements.
- Japanese capital is diversifying AI infrastructure exposure beyond US-China rivalry, treating Europe as neutral ground
- France’s nuclear fleet provides rare baseload power advantage as global data center electricity demand approaches Japan’s total annual consumption
- Project financing model at €75 billion scale remains unproven—execution depends on securing anchor tenant commitments before construction starts
- EU AI Act compliance creates regulatory moat that may attract workloads from firms seeking to avoid fragmented national frameworks
What to Watch
Customer contract announcements in Q3-Q4 2026 will determine whether SoftBank can secure the anchor tenants needed to close project financing for the Dunkirk facility. Delays beyond early 2027 would signal weaker demand than Son projects or pricing mismatches with hyperscaler alternatives.
EDF’s ability to deliver dedicated nuclear capacity without grid conflicts will face its first test as permitting advances. Any delays in reactor output commitments would force reliance on France’s export surplus, reducing the cost advantage over renewable-dependent competitors.
Macron’s government must maintain regulatory streamlining as construction begins. Reverting to standard permitting timelines—particularly for environmental approvals—would erase France’s speed advantage and validate concerns that European infrastructure projects cannot match US or Chinese execution velocity.
The EU AI Act’s August 2 enforcement deadline will clarify compliance costs for high-risk AI systems. If regulatory burdens prove materially higher than anticipated, the sovereign infrastructure thesis weakens, as firms may prefer US jurisdictions with lighter-touch AI governance despite data residency trade-offs.