SoftBank to Invest €75 Billion in 5 GW French AI Data Centers
Massive infrastructure commitment targets 3.1 GW in initial phase
SoftBank Group has announced a landmark €75 billion investment to develop 5 gigawatts (GW) of AI data center capacity across France. The project represents one of the largest single commitments to European AI infrastructure, highlighting the massive scale of power and capital required to sustain the next generation of artificial intelligence.
Key details
The investment will be deployed in phases, with an initial €45 billion allocated to deliver 3.1 GW of capacity in the Hauts-de-France region by 2031. Key sites have been identified in Dunkirk (Loon-Plage), Bosquel, and Bouchain. SoftBank plans to collaborate with Schneider Electric to anchor advanced data center manufacturing in Dunkirk, while working with SB Energy and other strategic partners on facility development.
The 5 GW total capacity is equivalent to the output of several nuclear reactors, underscoring the extreme energy intensity of hyperscale AI "factories." SoftBank Chairman Masayoshi Son emphasized that building this level of infrastructure is essential for countries seeking to lead in the AI transformation.
Why this matters
This project signals a shift toward utility-scale AI infrastructure where compute capacity is measured in gigawatts rather than megawatts. The sheer scale of the €75 billion investment reflects the skyrocketing costs of AI infrastructure, including high-performance GPUs, specialized cooling systems, and the underlying energy grid upgrades necessary to support such concentrated power demand.
Context
The announcement follows a global trend of "sovereign AI" where nations and major corporations race to secure domestic compute hubs. It also aligns with previous gigawatt-scale deals, such as the Google-Anthropic negotiations, as the industry moves beyond traditional data center footprints to meet the training and inference demands of trillion-parameter models.
Risks and open questions
The primary challenge lies in the energy grid's ability to support an additional 5 GW of concentrated load without compromising regional stability or net-zero goals. While the Hauts-de-France region offers industrial advantages, the environmental impact of cooling these massive facilities and the long-term water usage requirements for such high-density compute remain critical areas of concern for local regulators.
What happens next
The first phase of the project is expected to begin development immediately, with the initial 3.1 GW of capacity targeted for completion by 2031. Observers will be watching for specific energy procurement strategies, including potential behind-the-meter renewable projects or nuclear power agreements, to satisfy the campus's immense electricity requirements.
Source: SoftBank Group Published on AI Usage Global, author: AUG Bot
