France set to pledge one gigawatt of nuclear power for AI

France's President Emmanuel Macron (C), flanked by France's Prime Minister Gabriel Attal (2nd L), France's Minister for Economy and Finances Bruno Le Maire (L), France's Minister for Interior and Overseas Gerald Darmanin (R), and Elysee's general secretary Alexis Kohler (2nd R), chairs a security and defence council at the Elysee presidential palace in Paris on May 16, 2024, after three nights of clashes in France's riot-struck Pacific territory of New Caledonia protest following a reform changing voting rolls that representatives of the indigenous Kanak population say will dilute their vote.

---

• A gigawatt nuclear energy data center could open soon in France
• “Power is the number one consideration” for AI
• Models need more energy for training than ever

---

French President Emmanuel Macron has announced €109 billion in private investments to improve French AI infrastructure.

Key to the investment is one gigawatt of nuclear energy power, which will be destined for the high-performance computers and data centers needed to power emerging AI tools.

Announced as part of the AI Action Summit in Paris, the power will add to France’s already extensive nuclear infrastructure, which consists of 57 reactors across 18 different plants.

France will use nuclear energy to power AI

Around a third of the country’s entire energy consumption comes from nuclear energy, and its infrastructure is so extensive that it exported surplus energy to other countries last year. Using nuclear energy to power its AI data centers seems like the natural progression.

Macron added: “Plug, baby, plug” – a reference to US President Donald Trump’s “Drill, baby, drill” plan to expand oil drilling operations across the States.

Josh Parker, Nvidia’s senior director of corporate sustainability, said at the Summit: “Power is the number one consideration for getting access to AI and bringing our systems online” (via WSJ).

The first wave of the project will consist of 250 megawatts of AI-destined nuclear energy by the end of 2026. FluidStack, the company behind the project, aims to start work in the third quarter of 2024. It’s expecting to get most of its high-performance chips from Nvidia – currently the world’s second-most valuable company with a $3.27 trillion valuation.

Energy demands are only rising, too. Separate WSJ reporting reveals that today’s leading AI models used 30 megawatts of energy for training. That could rise to five gigawatts by the end of the decade.

The latest investments add France to the race of countries and companies bidding big on nuclear and other green energy sources to power future data centers.

You might also like

• We’ve listed the best AI writers you can use today
• Access high-performance chips with the best cloud computing services
• Amazon says it is now the top corporate purchaser of renewable energy in Europe