France’s public investment bank Bpifrance, French AI startup Mistral AI, and Abu Dhabi sovereign tech fund MGX say they will expand “Campus AI” into a nationwide network of high-performance computing sites, targeting 3 gigawatts of AI-dedicated capacity.
The partners are pitching the effort as more than another data center buildout. Campus AI is described as a mesh of high-performance computing infrastructure meant to meet rising demand from French companies and research labs for AI training and deployment.
At 3 GW, the backers say the project would rank among Europe’s most ambitious AI infrastructure initiatives in raw computing power—an industrial-scale push framed as part of France’s broader drive for “digital sovereignty,” or reducing reliance on non-European platforms for critical technology.
Campus AI: from a Paris pilot to a nationwide buildout
Until now, Campus AI operated as a pilot site. The national expansion changes the project’s purpose, the partners argue: it is no longer a demonstration but a foundational layer for French digital sovereignty.
The trio has leaned into the phrase “AI factories,” signaling an industrial approach—producing compute capacity at scale across multiple geographic locations rather than concentrating everything in a single facility.
The initiative lands in a crowded global race. The article points to U.S. hyperscalers—Microsoft, Google and Amazon—as well as Gulf countries investing heavily in similar infrastructure. Bpifrance, Mistral and MGX say their differentiator is governance and ownership that does not run through major U.S. platforms, with European and French actors involved in the capital structure.
For Mistral, founded in 2023, the stakes are also practical. Training and running large language models requires massive compute resources. Preferential access to a national infrastructure could reduce dependence on U.S. cloud providers and strengthen Mistral’s pitch to European customers focused on data sovereignty.
Why MGX’s role matters—and the governance questions it raises
MGX’s presence is a central part of the announcement. The Abu Dhabi fund has already invested in multiple AI infrastructure projects in Europe and Asia, and its move into Campus AI fits a strategy of geographically diversifying technology assets, according to the article.
For France, the involvement of foreign capital in what the partners describe as sensitive digital infrastructure raises governance questions that Bpifrance and the French government have not yet detailed publicly. Similar foreign investments in strategic digital assets have repeatedly sparked debate in France and across the European Union.
The partners have not yet specified the exact ownership structure, where the first sites will be located, or the timeline for ramping up to the 3 GW target. The announcement—dated June 26, 2026—sets out the alliance without laying out operational details.
What 3 gigawatts of AI compute means in practical terms
Three gigawatts is presented as a reference point tied to the electricity demand used to gauge the footprint of an AI data center fleet. By comparison, the article notes that major Microsoft or Google projects in Europe are typically measured in several hundred megawatts per site.
Reaching 3 GW would therefore imply either a constellation of several dozen sites or a smaller number of very large installations spread across France.
The scale reflects the technical reality of modern AI: training a large language model can tie up thousands of GPUs for weeks, and “inference”—running the model to answer user queries—also requires permanent, robust infrastructure.
For French companies that want to integrate AI without relying exclusively on AWS, Azure, or Google Cloud, the partners argue that a credible national option could shift commercial leverage. That is the market Campus AI aims to capture, with the next milestones being the actual rollout, the step-by-step ramp toward 3 GW, and the first regional sites coming online.
