Monday, October 12, 2026

Chips & Hardware

Ex-DeepMind researchers raise $3.7M to build shared fusion-reactor controls

Fusionality, founded by Google DeepMind alumni Federico Felici and Jonas Buchli, has raised a $3.7 million (CHF 3 million) pre-seed round to build control systems shared across fusion-power startups.

Two men seated in yellow armchairs look at a laptop screen together in a modern office.
Photo: Fusionality

Federico Felici and Jonas Buchli kept hearing the same thing from fusion power companies: they’d ideally buy the components for their control systems rather than build them from scratch, but almost nobody sells them, and “there isn’t anybody who speaks the language of fusion,” Felici told TechCrunch. The two had spent years developing ways to control experimental fusion devices, and decided the timing was right to build that missing supply themselves.

Felici and Buchli founded Lausanne-based Fusionality this year and quickly raised a $3.7 million (CHF 3 million) pre-seed round from Founderful and Playfair; Felici serves as CEO and Buchli as CTO. The company is part of an emerging supply chain serving the fusion power industry — some startups build components that turn fusion breakthroughs into grid electricity, while others are established manufacturers making high-precision parts. Few companies, though, specialize in the hardware and software needed to control fusion reactors themselves. “Across the industry, many companies make their own control systems from scratch,” Felici said, “but actually, 80% of each company’s control system is really exactly the same.”

Fusion power plants generate electricity by fusing atoms in superheated plasma, which requires split-second decisions to maintain the right temperature, shape, and fuel level; the control systems that do this are complex, but much of the underlying physics is consistent across reactor designs. Fusionality plans to build a suite of control systems and simulation environments that startups can fine-tune to their specific reactor. Felici and Buchli met while teaching AI to control an experimental tokamak, with Felici at Switzerland’s EPFL and Buchli at Google DeepMind; Felici later moved to DeepMind himself to develop simulations and machine-learning interfaces for fusion devices. Felici said AI will play an important role in future control systems but isn’t yet ready to handle the whole challenge, and is more likely today to complement, enhance, or optimize parts of the system rather than run a reactor outright.

For now, Fusionality is focused on magnetic confinement, the approach used by startups like Commonwealth Fusion Systems, Realta Fusion, Proxima Fusion, and Type One Energy, with plans to serve other fusion approaches eventually. The seven-person team will use the pre-seed funding to develop what Felici called “a tightly selected number of technologies,” which he declined to specify, likening them to Lego blocks the company will keep adding to.

Why it matters

If fusion startups really do share 80% of their control-system needs, a common supplier could cut costly duplicated engineering across the industry and speed how fast reactor designs reach the grid.