France has moved from summit announcements to actual construction. The 2025 Choose France Summit generated a record €40.8 billion in investment pledges, with €20.8 billion earmarked specifically for AI, nearly triple the figure from the previous year. This is not a policy plan. It is capital being committed and infrastructure going into the ground.
The centerpiece is Mistral, France's flagship AI company. Founded just three years ago, it has raised over $3.9 billion in total, including an $830 million debt package from seven banks to build a 44-megawatt data center near Paris. Debt financing is a meaningful signal: unlike equity, which bets on a future exit, debt requires the company to have enough revenue today to make loan payments. Seven banks agreed that Mistral does. The company is on track to surpass $1 billion in annual recurring revenue by end of 2026, with more than half of that already coming from European clients.
Mistral is also partnering with Abu Dhabi's sovereign investment fund MGX and France's public investment bank Bpifrance to develop Campus AI, a planned network of AI facilities anchored by a 1.4-gigawatt site in the Paris region, with the first operational phase representing €8.5 billion in investment. The UAE connection is part of a broader France-UAE agreement that envisions up to €50 billion in total investments in French AI infrastructure.
The core business case behind all of this is data control. When a European company runs its operations using AI hosted by an American provider, that data sits under US legal jurisdiction. A US law called the CLOUD Act allows American authorities to access data held by US-owned cloud companies anywhere in the world, regardless of where the servers physically sit. French-hosted infrastructure removes this exposure entirely. For banks, hospitals, insurers, and government agencies, this is not a theoretical concern. It is a procurement requirement.
A 2024 survey found that roughly 72% of European enterprise IT decision-makers already cite data sovereignty as a primary factor in cloud vendor selection. France's managed services market grew 142% year-on-year in the second quarter of 2025, driven largely by this regulatory-compliance demand. Mistral has already secured government contracts in France and Belgium partly because its infrastructure sits entirely within EU jurisdiction.
The gap between Europe and the US, however, remains large. In 2025, the US produced around 40 large AI models, China around 15, and the entire EU only about three. American cloud providers control roughly 70% of European digital services. US private investment in AI exceeded $109 billion last year alone, roughly 24 times UK investment. France is building against that backdrop, not erasing it.
There are also real limits to France's current position. Training the most powerful AI models, at the scale that companies like OpenAI operate, requires roughly 50 times more computing power than what Mistral used to train its current models. France does not yet have, and has not funded, infrastructure at that scale. For the most demanding AI development, French companies still depend on US cloud computing. That dependency partially contradicts the sovereignty argument, and it is worth being honest about.
What France has built is a credible, funded, legally differentiated AI alternative aimed at specific buyers: regulated industries in Europe that need provable data control. Healthcare operators, insurers, financial services firms, and public sector bodies are the natural first customers. For businesses in those sectors across Europe, French AI infrastructure is no longer hypothetical. It is operational, growing, and backed by real money.
For everyone else, the honest assessment is: watch the 200-megawatt target for 2027. If Mistral hits it, Europe will have built one of the continent's largest dedicated AI compute networks in under three years, and the sovereign AI thesis will have moved from political ambition to commercial reality.