France is trying to turn artificial intelligence into a national infrastructure project. Late Sunday local time, French president Emmanuel Macron announced €109 billion in private investments for the country’s AI ecosystem, a package worth around $112 billion at current exchange rates.
The announcement came as Paris hosts the Artificial Intelligence Action Summit, the third international summit focused on AI after earlier events in Bletchley Park in the U.K. and in Seoul, South Korea. Macron presented the package as a sign that France and Europe intend to accelerate in a global AI race shaped by data centers, foundation models, energy access and startup competition.
A French answer to Stargate
Macron directly compared the €109 billion package with Stargate, the $500 billion investment program led by OpenAI and SoftBank to build multiple AI data centers in the United States. He said the French package represents the same ratio for France as Stargate does for the United States.
The comparison matters because France has 68 million inhabitants, which is 5x fewer people than the U.S. Macron’s argument is not that the French package matches Stargate in total size. It is that, relative to population, France is attempting to put forward an AI investment effort of comparable scale.
Much of the money is expected to go toward new AI-focused data centers. That is why Stargate is the obvious benchmark: both initiatives are tied to the physical infrastructure needed to train and run advanced AI systems.
Several major investment pledges had already emerged before Macron’s announcement. TechCrunch counted up to €83 billion ($85 billion) as of Sunday, based on commitments from foreign and local players.
- €30 billion to €50 billion from the United Arab Emirates and MGX.
- €20 billion from Canadian investment firm Brookfield.
- €10 billion from Bpifrance.
- €3 billion from French telecom company Iliad.
Macron also mentioned Orange and Thales as investors in the program, while some companies had not yet announced their plans.
Why data centers are at the center
The largest AI systems need heavy computing capacity, and that computing capacity is increasingly concentrated in specialized data centers. France’s investment push is therefore not just about software companies or research teams. It is also about where the machines that support AI development will be built and powered.
The French government announced that the country plans to pledge one gigawatt of nuclear power to AI training by the end of 2026. That detail connects the AI package to a broader infrastructure argument: tech companies are looking for locations for power-hungry data centers, ideally powered by carbon-free electricity.
France is positioning itself as one of those locations in Europe. The country’s electricity system is a central part of that pitch. The majority of France’s electricity production comes from nuclear power plants, and France produces more electricity than it uses.
According to Macron, France exported 90TWh of electricity to neighboring countries in 2024. The government now wants to use that headroom to attract foreign investments into AI infrastructure.
That makes the French AI strategy unusually tied to energy policy. The investment package is not only a bet on AI startups or model development. It is also a bet that available electricity, data center capacity and national industrial planning can help France compete for AI projects that might otherwise go elsewhere.
Startups remain part of the pitch
Macron also pointed to French AI startups including Mistral, Wandercraft and Owkin. The mention of Owkin came with an important caveat: the company moved its headquarters to the U.S.
Even so, Macron argued that Europe remains competitive in artificial intelligence startups. He also suggested that DeepSeek creates an opportunity to catch up, because it showed a more frugal approach to building open models after adapting accessible innovations from the latest OpenAI model.
That point goes to a core tension in AI. One path emphasizes scale: larger systems, larger clusters and larger capital commitments. Another path looks for efficiency, reuse and lower-cost approaches. Macron’s comments suggest France wants to participate in both directions, backing infrastructure while also arguing that model development does not have to be defined only by the biggest spenders.
Mistral is central to that startup narrative. Arthur Mensch, co-founder and CEO of Mistral, announced plans to invest billions in an AI cluster in Essonne. He said the cluster would help the Paris-based company train more efficient systems within months.
The source describes Mistral as arguably the only European company working on foundation models that can compete with models from Alibaba, Anthropic, DeepSeek, Meta, OpenAI and others. That makes its data center project a symbolic piece of the broader French plan: infrastructure is being built not only for foreign capital, but also for a domestic company trying to stay in the foundation model race.
The strategic question for Europe
The French announcement arrives in a wider contest over where AI capacity will be built. The U.S. has Stargate. France is now presenting its own investment package as a national response, anchored by private capital, data centers, startups and nuclear power.
The scale is large, but the harder test is execution. Pledges from the United Arab Emirates and MGX, Brookfield, Bpifrance, Iliad and other investors still need to become real projects. Data centers need sites, power and timelines. AI startups need enough computing capacity to keep improving their systems.
Still, the strategy is clear from the facts Macron chose to emphasize. France wants to use its electricity surplus, its nuclear-heavy power mix and its existing AI startup base to become a preferred European location for artificial intelligence infrastructure.
That makes the €109 billion package more than a funding headline. It is France’s attempt to define a role in the next phase of AI: not only as a market for AI products, but as a place where the physical and technical foundations of AI are built.