DeepMind alumni launch Fusionality to standardize fusion reactor control
The Lausanne-based startup raised $3.7 million to replace bespoke engineering with modular operations technology for fusion power.
Former Google DeepMind researchers have launched a new venture aimed at accelerating the arrival of fusion energy by standardizing the complex control systems required to manage reactors. The startup, Fusionality, seeks to eliminate the industry's reliance on custom-built infrastructure in favor of a modular supply chain.
Based in Lausanne, Switzerland, Fusionality has raised $3.7 million (CHF 3 million) in pre-seed funding from investors Founderful and Playfair. The company is led by CEO Federico Felici and CTO Jonas Buchli, both of whom previously worked at Google DeepMind focusing on machine learning interfaces and simulations for fusion devices. Currently employing seven people, the startup provides simulation environments and control systems specifically for magnetic confinement fusion. Its client base already includes several prominent industry players, such as Commonwealth Fusion Systems, Realta Fusion, Proxima Fusion, and Type One Energy.
The shift toward fusion infrastructure
The fusion energy sector has experienced a massive influx of private capital, with over $10 billion invested over the last five years, fueling the growth of more than 60 private companies. While the majority of these firms focus on the fundamental physics of the reactor, a secondary supply chain is now emerging. This layer of the industry provides the engineering-grade infrastructure—including data layers, plasma control, and diagnostics—that is essential for transitioning from experimental prototypes to commercial power plants capable of feeding the electrical grid.
Reducing the engineering burden
Maintaining the stability of superheated plasma requires split-second decision-making and extreme precision. However, the current industry trend involves most companies building their own control systems from scratch. According to CEO Federico Felici, approximately 80% of these systems are essentially identical across different companies. By providing reusable, standardized components for this common infrastructure, Fusionality aims to reduce the engineering overhead for reactor developers, allowing them to focus on their unique physics challenges rather than reinventing basic operations technology.
The role of AI in reactor control
While the founders bring deep expertise in machine learning, Felici notes that AI is not intended to replace human-engineered control entirely. He stated that he does not advocate for AI to take over the role of controlling the entire reactor, suggesting instead that AI is more likely to complement, enhance, or optimize specific parts of the system.
What's next for the grid
As Fusionality scales its modular approach, the industry will be watching whether this standardization can meaningfully shorten the timeline for commercial fusion. The primary challenge remains the transition from laboratory success to a reliable power source, a process that depends heavily on the reliability of the underlying control systems Fusionality is now attempting to commoditize.