A group of former Google DeepMind researchers has launched a new AI-focused clean energy startup, raising an undisclosed initial seed round to build machine learning foundation models specifically tailored for nuclear fusion reactor design and real-time plasma control.
The startup leverages reinforcement learning and high-fidelity physics simulations to optimize magnetic confinement in tokamaks and stellarators, addressing core instability challenges that currently limit sustained net-energy output. By automating complex plasma turbulence predictions and physical hardware optimization, the company aims to significantly compress the engineering timelines required to bring commercial fusion energy to the electric grid, aligning with hyperscale cloud providers' soaring demand for zero-emission, baseload power.
