Google DeepMind partners with CFS to steer SPARC fusion plasma

Google DeepMind and Commonwealth Fusion Systems have opened a research partnership pointing AI tools at SPARC, the compact tokamak CFS is building to be "the first magnetic fusion machine in history to generate net fusion energy." The DeepMind announcement organizes the work into three tracks: plasma simulation, scenario search for net-energy operation, and eventually real-time control inside the machine.

The centerpiece is TORAX, an open-source plasma simulator DeepMind built in JAX. "TORAX is a professional, open-source plasma simulator that saved us countless hours in setting up and running our simulation environments for SPARC," said Devon Battaglia, Senior Manager of Physics Operations at CFS. DeepMind's own pitch is that the simulator, paired with reinforcement learning and its evolutionary-search system AlphaEvolve, will let agents "explore vast numbers of potential operating scenarios in simulation, rapidly identifying the most efficient and robust paths to generating net energy" before SPARC ever fires up.

The relationship is not purely academic. Google agreed earlier this year to buy 200 megawatts of power from CFS's first ARC commercial plant in the early 2030s, and it deepened an existing equity stake in the company. Writing on The Tokamak Times, CFS's Principal Open Innovation Engineer Jaydeep Deshpande said "AI tools carry serious potential to accelerate learning on SPARC and then to push toward making our ARC fusion power plants more efficient."

The collaboration builds on DeepMind's earlier work with the Swiss Plasma Center at EPFL, where reinforcement learning controlled magnets on the smaller TCV research tokamak. SPARC is a much bigger target: CFS has said the reactor will be finished sometime in 2026.