SoftBank Group announced last week that it will invest up to €75 billion, roughly $87 billion at current exchange rates, to build and operate 5 gigawatts of AI data center capacity across France. The announcement was made at President Emmanuel Macron’s annual “Choose France” summit and marks the largest AI infrastructure commitment SoftBank has made anywhere in Europe.

The scale is genuinely difficult to grasp. For context, 5 gigawatts is roughly the combined output of five large nuclear reactors, and SoftBank wants it dedicated entirely to powering AI compute.

What SoftBank Is Actually Building

The investment is structured in two phases. The first phase commits €45 billion to deliver 3.1 gigawatts of capacity in the Hauts-de-France region by 2031. Three specific sites are already identified: Dunkirk (Loon-Plage), Bosquel, and Bouchain. SoftBank is partnering with French state-owned utility EDF on the Bouchain site and with Schneider Electric on a manufacturing cluster at the Port of Dunkirk, where prefabricated data center components will be produced at scale.

The second phase, contingent on how the first plays out, brings the remaining 1.9 gigawatts online for a total of 5 GW.

Why France, Not the US or UK

The answer is nuclear power, and it explains a lot about where global AI infrastructure is heading.

France generates roughly 70 percent of its electricity from nuclear reactors run by EDF, making it the world’s largest net electricity exporter. Industrial power prices there come in at less than half what comparable buyers pay in the UK, and the grid’s low carbon intensity gives hyperscalers a clean story for their sustainability commitments.

The US has land and cheap gas, but permitting new data center campuses increasingly runs into grid interconnection queues stretching four to seven years. France has existing infrastructure, available industrial sites near the coast, and a government that is actively clearing the path for AI investment.

SoftBank’s choice is not just about today’s electricity costs. AI training workloads are intensely energy-constrained, and the companies building for 2028 and beyond are securing power access now.

The Bigger Picture: Europe Is Competing

This announcement lands at a moment when European governments are scrambling to attract AI infrastructure investment before the continent’s window closes. The US and China have pulled far ahead on AI compute capacity, and European policymakers are increasingly worried about becoming dependent on foreign AI infrastructure.

Masayoshi Son, SoftBank’s founder and CEO, has made AI his defining bet for this era of the company. His previous Vision Fund investments were often criticized for valuing growth over fundamentals. This wave is different in character: physical infrastructure, contracted energy, and a clear industrial rationale.

The partnership with Schneider Electric, a French industrial giant, is also meaningful. It means components will be manufactured locally, jobs will be created along the supply chain, and France gets something beyond a landlord arrangement with a foreign capital allocator.

What This Means for Business

For business leaders watching AI deployment costs, this story matters for two reasons.

First, the economics of running AI agents and large language models at scale are directly tied to where compute lives and what it costs to power it. As more AI infrastructure gets built close to cheap, stable nuclear power, the cost curves for enterprise AI workloads will continue to fall. What costs a certain amount to run today will cost less in 2027 and less again in 2029.

Second, announcements at this scale signal where the serious money thinks AI is heading. SoftBank is committing €75 billion because it expects AI demand to grow far beyond current levels. That is not a hedge. That is a bet that AI agents, not AI assistants, will be running a significant portion of enterprise workflows within this decade.

For companies still debating whether to invest in AI automation or upskill their teams to work with AI tools, that signal deserves attention. The infrastructure is being built. The question is whether your organization will be ready to use it when the costs drop and the capabilities arrive.

If you are evaluating where AI fits in your business operations today, a fractional AI advisor can help you map a practical strategy without committing to a full-time executive hire. The infrastructure decisions being made this year will shape what is affordable and accessible in three years.

---

Enterprise DNA works with business leaders on practical AI strategy and deployment through Omni Advisory and helps teams build the data literacy to take advantage of what is coming.