Infrastructure

Campus AI Eyes Up to 3 GW Nationwide as Second French Site Looms

· 6 min read · Infrastructure

Campus AI Eyes Up to 3 GW Nationwide as Second French Site Looms

French and trade-press coverage in September 2026 says Campus AI—the joint venture linking MGX, Bpifrance, Mistral AI, and Nvidia—is lifting its national capacity ambition toward as much as 3 gigawatts, roughly doubling beyond the roughly 1.4 GW planned for the Fouju campus in Seine-et-Marne. President Emmanuel Macron publicly framed a €7.5 billion MGX extension as part of making France a leading European destination for lower-carbon AI infrastructure while a second site shortlist advances.

Filed under Infrastructure and dated September 15, 2026, this AI4France briefing treats the scale-up talk as industrial-policy news—not energised megawatts. Fouju’s prefectural authorisation and phased build remain the near-term proof point, with Mistral described as an early customer for tens to hundreds of megawatts once halls come online later in the decade.

Why it matters: French sovereign-AI plans need predictable power, local consent, and exportable industrial value—not just headline gigawatts. A second campus can diversify geography—but only if grid connection, water stewardship, and community benefits are published with the same urgency as investment tweets.

What it means in practice

Campus AI Eyes Up to 3 GW Nationwide as Second French Site Looms — contextual photo

French regional and enterprise leaders should inventory how Fouju phase-one timelines relate to Campus IA’s first sovereign-compute bet; confirm whether second-site criteria include renewable matching and heat reuse; assign an owner for environmental-court risk; and prefer offtake contracts tied to measured availability. Cross-check model-side demand via Mistral’s mega-round and TotalEnergies’ subsurface AI programme.

Caveats come first. Capacity ambitions are not interconnection agreements; €7.5 billion figures mix partners and phases; and local opposition around Fouju shows permitting risk remains live. AI4France therefore presents the 3 GW narrative as directional infrastructure context until site selection and energisation milestones are documented.

What to watch next: named second-site shortlist; published RTE connection calendars; and first Mistral training or inference workloads on Campus AI halls. Readers can continue on the AI4France homepage, or browse the Newsroom for additional briefings.

Bottom line: treat this update as orientation, not instruction. French AI campus ambitions are expanding and still contested on the ground. Organizations that benefit most will verify power and water math, keep municipalities informed, and refuse to confuse a capacity target with online compute.

← Back to AI4France