France is dramatically expanding its bid to build homegrown AI computing power. On June 1—on the sidelines of the Choose France 2026 investment summit—Bpifrance, French AI company Mistral and Abu Dhabi-backed investor MGX signed an agreement to scale the Campus AI program to 3 gigawatts of computing capacity across France.
The move reframes Campus AI from a single site into a national network of “AI Gigafactories,” with the consortium describing the selection of a second location as imminent. The first campus, in Fouju in the Seine-et-Marne department east of Paris, is still under construction and is slated to open in 2028.
Even with that first site not yet online, the partners are pressing ahead and formally doubling investment. In a LinkedIn post, French President Emmanuel Macron put a number on the ambition: €7.5 billion (about $8.1 billion), a figure not included in Bpifrance’s official statement but reported by Le Figaro.
A national network of AI “gigafactories,” not a single data center
Campus AI brings together four shareholders: Bpifrance, France’s public investment bank; Mistral, the developer behind the Le Chat assistant and the Mistral Large family of models; MGX, an Emirati investment fund focused on AI and advanced technologies; and Nvidia, the dominant supplier of accelerated computing hardware.
Nvidia’s presence in the ownership structure is a signal in itself. Access to GPU chips has become the central constraint for any company trying to train and run large-scale generative AI systems.
Mistral locks in the most contested resource in AI: compute
The most consequential element of the agreement isn’t just the size of the planned network—it’s the priority access Mistral receives to all computing capacity deployed across the sites. As a shareholder in the joint venture, Mistral isn’t simply renting GPUs on the spot market; it is securing guaranteed supply at a time when global demand for compute is surging and access to Nvidia-backed infrastructure can take months, even years.
At Fouju, Mistral has already secured 96 megawatts, with a contractual ramp-up to 200 MW. That capacity is positioned as central to the competitiveness of its foundation models—training, fine-tuning and, especially, commercial-scale inference. Without that secured resource, Mistral’s growth would depend on U.S. cloud providers, raising sovereignty concerns for a French—and more broadly European—AI player.
The near-term timeline is also tightening. Starting in the third quarter of 2026, Mistral is set to gain an additional 10 MW for inference through a data center in Les Ulis, in the Essonne department south of Paris, according to ZDNet France. The company has framed infrastructure as a core pillar of its strategy, alongside model quality, at its AI Now Summit.
The comparison with major U.S. AI labs is stark. OpenAI, Google DeepMind and Anthropic operate at a different order of magnitude, with direct access to hyperscalers. Mistral is pursuing a different model: a sovereign, European infrastructure built through a joint venture rather than bought off the shelf. The upside is control and continuity of access; the risk is scale—200 MW at Fouju is meaningful for a European player, but modest compared with gigafactories planned in the United States.
Choose France 2026 spotlights AI infrastructure as industrial policy
Campus AI was not the only major digital-infrastructure announcement tied to Choose France 2026. The summit drew an unusually large set of commitments in the sector. Japan’s SoftBank announced €45 billion (about $48.6 billion) in investment in French data centers by 2031, with a longer-term target of €75 billion (about $81.0 billion). Canadian asset manager Brookfield added €10 billion (about $10.8 billion) more to reach up to €30 billion (about $32.4 billion) in total AI infrastructure investment in France, according to Le Figaro.
Macron welcomed what he described as France becoming “the leading European destination for decarbonized AI infrastructure.” The claim is political, but it rests on an industrial argument emphasized by the sources: France’s electricity mix, dominated by nuclear power, can offer a lower carbon footprint than much of Europe—an advantage for energy-hungry AI data centers courting customers with ESG commitments.
Campus AI fits that pitch, while also aiming beyond France. The stated goal of building a pan-European network of “AI Gigafactories” will hinge on the next site, which all cited sources describe as close to being selected. It will need the same fundamentals: available land, access to very high-power grid connections, and proximity to operators able to absorb the capacity. For now, the Paris region anchors the buildout—Seine-et-Marne for the first campus and Essonne for the Les Ulis inference data center.
The real test arrives with Fouju’s planned 2028 opening. Until then, the consortium will need to show that a financing model mixing a French public investment bank, an Emirati sovereign-backed fund and private industrial players can hold up against competitors with far larger balance sheets. If Macron’s €7.5 billion (about $8.1 billion) figure is confirmed in official documents, it would rank among the largest private commitments ever made to digital infrastructure in France.
