Playing Trumps: France to Allocate Gigawatt to Create ‘Nuclear’ AI Cluster

France is looking to build on its nuclear industry to catch up with AI leaders. The country will set aside a gigawatt for a new AI project that investors and the government estimate will cost tens of billions of dollars, The Wall Street Journal reports. Combined with other projects, mostly financed by Middle Eastern foundations, the country and Europe will see a significant boost in AI computing.

The new “atomic” initiative will see the AI ​​data center receive its first 250 MW by the end of 2026. This will allow it to compete with the American Stargate project, supported by OpenAI, SoftBank and others. The US project starts with a 200 MW campus with the possibility of expanding to 1.2 GW.

According to FluidStack, the company responsible for the new “atomic” AI cluster, construction will begin in the third quarter of 2025. But there are no guarantees yet that the project will proceed as planned and receive enough funds and chips. Own funds and loans in the amount of €10 billion ($10.3 billion) will be used to finance the first stage. It was also announced that negotiations are underway with one of the world’s largest AI developers about using the new facility, which can accommodate about 120 thousand NVIDIA accelerators. The head of OpenAI Sam Altman recently visited France.

By 2028, the number of accelerators in the cluster may grow to 500 thousand, and by 2030 the site may expand to 10 GW. Most of the funds will go to NVIDIA AI accelerators. FluidStack reported that it is in constant contact with NVIDIA, which allegedly promised to deliver the products as soon as they are needed. Earlier, it was also reported that France and the UAE intend to invest up to €50 billion in a gigawatt AI data center. In total, €109 billion in investments in the French AI sector were announced.

Image source: Alice Triquet/unsplash.com

France has 57 reactors at 18 nuclear power plants, generating two-thirds of the country’s energy. In 2024, it generated about 20% more electricity than it used, exporting the surplus. France’s state-owned energy company EDF (Électricité de France) announced last year that it was ready to provide power and land for three gigawatt AI data centers.

If FluidStack is implemented as planned, the balance of power in the AI ​​market could partially shift in favor of France and the EU in general. Some European companies, including France’s Mistral AI, are engaged in cutting-edge developments, but overall Europe is largely lagging behind American and Chinese companies, although it will try to catch up by investing €200 billion in AI development.

The emergence of low-cost but efficient models from China’s DeepSeek has raised serious doubts about the need for large clusters of chips, temporarily crashing the stock market in the AI ​​space. However, players from Blackstone and Brookfield to NVIDIA itself remain optimistic that high-performance models will require even more chips and data centers.