DeepMind Partners with Commonwealth Fusion Systems to Use AI for Fusion Energy

Google’s DeepMind has announced a collaboration with Commonwealth Fusion Systems (CFS) to apply artificial intelligence to fusion energy research, according to a company blog post. The partnership focuses on using AI to simulate and control plasma inside fusion reactors, aiming to help CFS achieve stable and efficient fusion power.

The collaboration centers on DeepMind’s TORAX simulator, an open-source tool built in JAX that models plasma behavior inside a tokamak reactor. CFS is using TORAX to run virtual experiments and refine operational strategies for its upcoming SPARC reactor, which is designed to be the first magnetic fusion device to generate more energy than it consumes.

DeepMind’s AI systems will also employ reinforcement learning and evolutionary search methods to identify optimal reactor settings and develop real-time plasma control strategies. These tools are expected to help manage the intense heat produced during fusion and ensure stable operation within reactor limits.

The partnership builds on DeepMind’s previous work with the Swiss Plasma Center, where it demonstrated that deep reinforcement learning can control the magnetic fields of a tokamak. Through this collaboration, DeepMind and CFS aim to accelerate progress toward commercially viable fusion energy.