- SoftBank announced a commitment to develop and operate up to 5 GW of AI data center capacity in France.
- The first phase is described as an initial €45B investment to deliver 3.1 GW in Hauts-de-France by 2031.
- Sites named by SoftBank are Dunkirk/Loon-Plage, Bosquel and Bouchain; the plan still depends on buildout, power, permits and execution.
SoftBank Group announced on May 31 that it plans to build a large AI data center footprint in France, framing the commitment as part of the 2026 Choose France summit hosted by President Emmanuel Macron.
The core number is big, but the status matters. SoftBank is not saying the capacity already exists. It is announcing a plan to develop and operate up to 5 GW of AI data center capacity in France, representing investment of up to €75B, with the first phase focused on 3.1 GW in Hauts-de-France by 2031.
What happened
SoftBank's official release says the company has made a commitment to develop and operate AI data centers in France. It calls the first phase an initial €45B investment and says the projects will support demand from AI companies, cloud providers, enterprises, public institutions and research organizations.
TechCrunch reported the same headline numbers and noted the context: SoftBank is both an investor in and customer of OpenAI, while large data center projects are drawing more scrutiny over energy use, grid strain and local environmental concerns.
What the plan includes
The first phase includes data centers in Dunkirk, also identified as Loon-Plage, Bosquel and Bouchain. SoftBank says the Hauts-de-France phase is intended to deliver 3.1 GW of AI data center capacity by 2031. The broader French program is described as up to 5 GW.
The announcement also includes a manufacturing and supply-chain element. SoftBank says it will partner with Schneider Electric in Dunkirk on a large industrial production cluster, including a SoftBank-operated facility to manufacture enclosures and a Schneider Electric facility to integrate data center power modules.
Image: Bouchain power station cooling tower in northern France - Grobert, Wikimedia Commons, CC BY-SA 3.0.
What is confirmed
Confirmed: SoftBank publicly announced the plan; the stated top-line investment is up to €75B; the stated first phase is €45B; and the first phase target is 3.1 GW in Hauts-de-France by 2031.
Also confirmed from SoftBank's release: the named locations include Dunkirk/Loon-Plage, Bosquel and Bouchain, and the Dunkirk industrial plan involves Schneider Electric. DCD reported the announcement in the data center industry context as a planned buildout of up to 5 GW.
What is not confirmed
The announcement does not mean 5 GW of capacity is built, energized or already serving AI workloads. It also does not settle the final project schedule, the full financing path, local approvals, grid-connection timing, environmental review or the ultimate customers for the capacity.
That distinction matters because AI infrastructure claims are easy to overstate. A gigawatt-scale data center program requires land, transmission access, substations, cooling systems, servers, networking, long procurement chains and local political support. The press release is the beginning of the source trail, not the end of the buildout.
Why France matters
France is trying to position itself as a European AI infrastructure hub. The SoftBank release points to grid infrastructure, industrial land, engineering talent and national AI ambition as reasons for the investment. The French government has also been actively courting AI and data center announcements through Choose France.
Hauts-de-France gives the plan an industrial angle as well as a compute angle. Dunkirk and the surrounding region already sit in a heavy-industry and energy corridor, while Bouchain is tied to existing power infrastructure. That does not eliminate grid constraints, but it explains why the region is being framed as strategically useful.
Why it matters
If the first phase is delivered, it would add a major European AI compute platform at a time when AI companies are racing to secure power and data center capacity. It would also deepen SoftBank's role in the AI infrastructure stack beyond capital allocation and into physical compute operations.
The harder question is whether the promised scale can arrive on the announced timeline. For readers, the useful way to track this is not by repeating the €75B figure alone. Watch permits, grid agreements, site-level construction starts, equipment orders, named customers and whether the first 3.1 GW stays on track toward 2031.
What to watch next
Watch for local filings in Dunkirk/Loon-Plage, Bosquel and Bouchain; EDF and RTE grid-connection details; Schneider Electric's manufacturing timeline; and any specific anchor customers. Also watch environmental and local-government responses, because large AI data center projects increasingly face questions over water, land, electricity demand and household utility costs.
For background on the facilities behind announcements like this, read NoDechev's explainer: What Are AI Data Centers?
NoDechev rating: confirmed announcement, planned infrastructure. SoftBank's commitment and first-phase targets are sourced; delivery, final capacity and operational status remain future milestones.
Also Read
AI data center announcements are really about compute, power, cooling and permitting.
Read the AI data centers explainer →

Image: Computer server rack - ronK / Wikimedia Commons, CC BY-SA 2.5.