Former DeepMind Engineers Launch Startup to Streamline Fusion Reactor Control

Federico Felici and Jonas Buchli are among the fortunate entrepreneurs whose startup idea arrived organically.

Time and again, they encountered the same sentiment from fusion companies: “We would ideally source many components for our control systems, but no one provides them—especially vendors who understand the language of fusion,” Felici told TechCrunch.

For years, the co-founders dedicated themselves to developing innovative methods for controlling experimental fusion devices. “We spent considerable time solving precisely the problems our customers face,” Felici noted. “It felt like the right moment to launch.”

To capitalize on this opportunity, the duo founded Lausanne-based Fusionality this year, swiftly securing a $3.7 million (CHF 3 million) pre-seed round from Founderful and Playfair. Felici acts as CEO, while Buchli serves as CTO.

Fusionality represents a growing segment of the supply chain emerging to support the rapidly expanding fusion power industry. While some startups, such as Kyoto Fusioneering, focus on components that translate engineering breakthroughs into grid electricity, others are established manufacturers capable of producing high-precision parts.

However, few companies specialize in the hardware and software necessary to operate the reactors themselves. “Across the industry, many firms build their control systems from scratch,” Felici pointed out. “Yet, roughly 80% of each company’s control system is fundamentally identical.”

Fusion power plants generate electricity by capturing the energy released when atomic nuclei fuse, a reaction most likely to occur in a superheated plasma. However, plasmas are highly volatile, requiring split-second adjustments to maintain the proper temperature, shape, and fuel levels.

The control systems required to maintain these conditions are complex to develop and deploy, yet the underlying physics remains consistent across various reactor designs. Fusionality intends to leverage this commonality, building a suite of control systems and simulation environments that startups can subsequently customize for their specific designs.

Felici and Buchli initially crossed paths while teaching AI to control an experimental tokamak—a donut-shaped device used to study fusion power. Felici was based at EPFL in Switzerland, which operates one such tokamak, while Buchli worked at Google DeepMind. Felici later transitioned to Google DeepMind, where he developed simulations and machine learning interfaces for fusion devices.

Felici believes AI will “play an important role in future control systems,” though it is not yet ready to tackle the entire challenge. “I would not advocate for AI to control the entire fusion reactor,” he stated. Instead, AI is more likely to “complement, enhance, or optimize” specific components of the system.

Currently, Fusionality is concentrating on magnetic confinement, a fusion approach that utilizes powerful electromagnets to maintain fuel density and heat for power generation. Commonwealth Fusion Systems, Realta Fusion, Proxima Fusion, and Type One Energy are among the startups developing such magnetic confinement devices. Felici noted that the company will eventually expand to support other fusion approaches.

The pre-seed funding will assist Fusionality’s current team of seven in developing a “tightly selected number of technologies,” Felici said, though he declined to specify which ones. He compared them to Lego blocks. “We start with a few of these blocks and then we build out more,” he explained. “We have the ambition to serve the wide range of technology required to operate a fusion reactor.”

Source link