A Lausanne startup founded by two former Google DeepMind researchers has raised $3.7 million (CHF 3 million) to build the fusion control systems that reactor developers say nobody is currently selling them. Fusionality, which emerged from Venture Kick and EPFL’s Swiss Plasma Center, closed the pre-seed round from Zurich-based Founderful and London’s Playfair Capital.
The company’s co-founders, Federico Felici and Jonas Buchli, heard the same complaint repeatedly from fusion developers. ‘We would ideally buy many of the components to make our control system, but there isn’t really anybody providing them,’ Felici said, ‘and especially there isn’t anybody who speaks the language of fusion.’
Why Fusion Control Systems Need a Shared Foundation
Running a fusion reactor requires split-second decisions about plasma temperature, shape, and fuel density. The physics involved is extraordinarily complex, and most fusion startups currently build their control software from scratch. Felici puts the waste in stark terms: ‘80% of each company’s control system is really exactly the same.’
Fusionality’s proposition is to build that shared 80% once, then let clients customise the remainder for their specific reactor design. According to Fusion Energy Base, the architecture is designed so that each subsequent deployment costs meaningfully less than the first, as components are reused across devices with minimal modification. Tech Funding News reports that more than 30 private fusion companies currently need the plasma software Fusionality is developing.
For now the company is concentrating on magnetic confinement, the approach that uses powerful electromagnets to hold plasma dense and hot enough to sustain a reaction. Commonwealth Fusion Systems, Realta Fusion, Proxima Fusion, and Type One Energy are among the startups working in that space. Felici said Fusionality will eventually extend its systems to other fusion approaches.
Felici is cautious about overstating what artificial intelligence can do today. ‘I wouldn’t be somebody who advocates for AI to take the role of controlling the entire fusion reactor,’ he said. AI is more likely to ‘complement or enhance or optimize’ parts of the system in the near term, he added, though he expects it to ‘play an important role in future control systems.’
From a London Hackathon to a Lausanne Startup
The partnership between Felici and Buchli stretches back further than their DeepMind years. According to the EPFL newsroom, Felici first met DeepMind scientists at a hackathon at the company’s London headquarters in 2018, a meeting that seeded what became a formal research collaboration between EPFL’s Swiss Plasma Center and DeepMind.
That collaboration produced one of the more cited results in applied fusion science. The team trained a deep reinforcement learning agent to control EPFL’s TCV tokamak, a donut-shaped experimental device, and published the results in Nature in 2022. Felici, Buchli, and three colleagues are listed as equal contributing authors on that paper, and several of them, including both founders, filed a provisional patent on the work.
The two founders bring complementary academic backgrounds to the venture. Felici earned his PhD in plasma physics at EPFL, then held a faculty position at TU Eindhoven before returning to lead control activities at EPFL’s Swiss Plasma Center, and subsequently spent 2.5 years at Google DeepMind before founding Fusionality, according to Startup Ticker. Buchli, who serves as CTO, held a faculty position at ETH Zurich before joining Google DeepMind as a senior research scientist, according to the Greater Geneva Bern Area agency.
‘We spent quite a lot of our time solving exactly the problems that our customers need to solve,’ Felici said of their research years. ‘It felt like the right time.’
Founderful, founded in 2017 and headquartered in Zurich, focuses on pre-seed investments in Switzerland across energy, software, and materials. It has previously backed EPFL spin-out Isospec Analytics. Playfair has also invested in UK fusion-isotope startup Astral Systems, giving both investors prior exposure to the fusion supply chain.
Fusionality’s team of seven will use the pre-seed funding to develop what Felici described as a ‘tightly selected number of technologies,’ declining to name them. He compared the approach to assembling Lego blocks: ‘We start with a few of these blocks and then we build out more,’ he said. ‘We have the ambition to serve the really wide range of technology that you need to operate a fusion reactor.’
With more than 30 potential customers and no direct competitors in commercial fusion control software, the company’s first real test will be whether it can convert that open field into signed contracts before the reactor developers decide, once again, to build it themselves.
