France has decided that the future of artificial intelligence won’t just be powered by GPUs and grid upgrades; it will be fuelled by fast neutrons. In a move that places the country at the frontier of next-generation energy systems, French startup Stellaria has secured the first commercial reservation for its advanced reactor, Stellarium, and will work with Equinix to deliver 500 MW of clean nuclear power for AI infrastructure.
If the plan succeeds, this would be the first time a nuclear reactor is explicitly designed and deployed to run commercial-scale AI workloads.
A Different Kind of Reactor for a Different Era
Stellaria’s design doesn’t belong to the conventional family of reactors that power grids today. Stellarium is a fast-neutron molten-salt reactor using liquid chloride fuel, a configuration engineered not only to generate energy but also to destroy nuclear waste.
The reactor’s fast neutrons can break down actinide-rich materials such as plutonium, uranium, and MOX fuels, giving it a capability that traditional reactors have never offered at commercial scale.
What’s remarkable is the reactor’s scale: a device designed to potentially consume more waste than it generates, with a footprint of roughly four cubic meters and an additional containment barrier for safety. Even more impressively, Stellarium is engineered to operate for two decades without refueling.
Why AI Needs a New Energy Playbook
Electricity demand tied to AI and high-performance computing is on track to rise 4% by 2027, and even major economies like the United States are struggling to supply enough power for hyperscale facilities. Grid bottlenecks, multi-year queue delays, and surging demand are pushing data-center developers toward unconventional solutions.
Stellaria and Equinix plan to deploy Stellarium units across Equinix’s European campuses, with each reactor delivering around 250 MWe, allowing data centers to operate with their own carbon-free baseload rather than relying solely on constrained national grids. Equinix is also exploring how these reactors could support a new class of more autonomous, self-powered facilities.
Why Nuclear Makes Sense for AI
AI workloads can’t tolerate unstable power. They require long, uninterrupted runs, predictable loads, and zero-downtime environments, a match that nuclear reactors can deliver more reliably than solar, wind, or existing grid resources.
Fast-neutron reactors add further benefits: a waste-consuming fuel cycle, stable carbon-free generation, and the ability to follow load as computing demand fluctuates.
As AI accelerates faster than policymakers can react, energy experts are even floating futuristic scenarios, including AI infrastructure powered from orbit. Until then, terrestrial solutions like Stellarium are filling the widening energy gap.
The Road From Lab to Grid
Stellaria recently closed a €23 million funding round to speed up Stellarium’s progression. The timeline is ambitious:
- 2029: first fission reactions
- 2035: beginning of commercial deployment
France views the project as part of a larger national ambition: reducing fossil-fuel dependence and reclaiming global leadership in advanced nuclear systems. If the commercial rollout succeeds, a single Stellarium could power a city of over 400,000 people, or keep the world’s most computationally demanding AI systems running around the clock.
With fast neutrons, molten salts, and a reactor designed to shrink nuclear waste rather than expand it, Stellaria is positioning France to set the template for AI-ready nuclear power- 500 MW at a time.
If the world is entering an era where energy is the real limit to intelligence, France is signalling that those limits are optional.
