Google DeepMind alumni launch fusion-control startup Fusionality

Former Google DeepMind employees Federico Felici and Jonas Buchli have founded Lausanne-based Fusionality to build control systems and simulation environments for fusion-power startups. The company quickly raised a $3.7 million (CHF 3 million) pre-seed round from Founderful and Playfair.

Fusion companies often build their control systems from scratch, but Felici said 80% of each system is essentially the same. Fusionality plans to develop reusable systems that startups can adapt to their reactor designs.

Fusion power generates electricity from the energy released when two atoms fuse in superheated plasma. Maintaining the plasma’s temperature, shape, and fuel level requires rapid decisions, while much of the underlying physics remains consistent across reactor designs.

Felici and Buchli met while teaching AI to control an experimental tokamak. Felici worked at EPFL in Switzerland, while Buchli worked at Google DeepMind. Felici later developed fusion-device simulations and machine-learning interfaces at Google DeepMind.

Felici said AI will play an important role in future control systems, but is not ready to control an entire fusion reactor. Fusionality is initially focusing on magnetic confinement, with examples in the field including Commonwealth Fusion Systems, Realta Fusion, Proxima Fusion, and Type One Energy. The team currently has seven members.