US firm advances with Google to fine tune nuclear fusion reactor plasma

US-based nuclear fusion company Commonwealth Fusion Systems (CFS) has partnered with Google’s DeepMind to leverage artificial intelligence (AI) in optimizing the plasma control of its upcoming SPARC fusion reactor. The collaboration utilizes DeepMind’s open-source Torax software, released in 2024, to simulate and model the superhot plasma inside SPARC, aiming to improve operational efficiency and accelerate the development of commercial fusion power plants, known as ARC. By applying reinforcement learning—an AI technique previously used by DeepMind in other fusion research and famously in AlphaGo—the project seeks to identify optimal configurations for fueling rates, radio-frequency heating, and magnet currents while maintaining safe operational limits. This partnership builds on an existing relationship, with Google already investing in CFS and committing to purchase 200 megawatts of power from the first ARC plant expected in the early 2030s. The AI-driven approach could be used both for pre-operation planning and real-time control, including managing heat exhaust in critical reactor regions.