Fusionality Raises $3.7M to Build Control Systems for Fusion Reactors
Lausanne-based startup Fusionality has raised a $3.7 million (CHF 3 million) pre-seed round to build control systems and simulation environments for fusion reactor developers. The round was backed by Founderful and Playfair. TechCrunch
The company was founded in 2026 by Federico Felici and Jonas Buchli. Felici serves as CEO and Buchli as CTO. TechCrunch
Fusionality is initially focusing on magnetic confinement fusion, the approach that uses powerful magnetic fields to contain superheated plasma inside a reactor. The company plans to build a suite of control systems and simulation environments that fusion startups can adapt to their specific reactor designs. TechCrunch
Felici and Buchli met while teaching AI to control an experimental tokamak — a doughnut-shaped fusion device — at EPFL in Switzerland. Felici worked at EPFL, while Buchli worked at Google DeepMind. Felici later moved to Google DeepMind, where he developed simulations and machine learning interfaces for fusion devices. TechCrunch
Felici said fusion companies told him there is no one providing control-system components that 'speak the language of fusion.' TechCrunch
According to the company's website, fusion organizations face a common bottleneck: too few control-system experts are available globally to build in-house diagnostics, control, and data teams. Fusion devices require these systems to measure plasma behavior, interpret it quickly, and control the plasma through actuators in real time. Fusionality
The website also states that Fusionality's technology provides real-time diagnostics for fusion device operations. It describes the company as providing measurement, control, and data systems for the fusion era, with a mission to help fusion teams analyze, operate, and control their devices with confidence. Fusionality
The broader context here is the convergence of two trends this publication has tracked closely over the past decade: machine-learning-driven control for complex physical systems has matured, and private capital has surged into fusion energy. At Google DeepMind, Buchli and Felici worked at the intersection of both. Their plan to productize control systems and simulation environments targets a genuine gap in the fusion supply chain. Magnetic confinement fusion depends on real-time plasma control with sub-millisecond latency requirements, meaning the system must react in less than a thousandth of a second. A shortage of specialists who can build custom diagnostics and actuator loops for novel reactor geometries is a concrete, widely acknowledged bottleneck. If Fusionality can deliver adaptable control-system components, it could compress the iteration cycle for reactor startups that would otherwise need to assemble bespoke teams.
Worth flagging: the gap between a well-funded pre-seed round and a deployed, production-grade control system for a fusion device is substantial. Fusionality is entering a domain where control-system failures carry physical risk, not just software downtime. The validation cycle for components used in magnetic confinement fusion will be measured in years, not quarters. The company's ability to build trust with reactor developers and demonstrate that its simulation environments can faithfully model plasma behavior under real-time constraints will determine whether it moves from an ambitious concept to an indispensable layer in the fusion stack. The $3.7 million pre-seed gives the team runway to build, but the path to commercial adoption runs through integration with operating experimental reactors.
In this author's view, the verticalization pattern Fusionality is pursuing mirrors what happened in the autonomous vehicle industry roughly a decade ago. General-purpose ML frameworks existed, but the companies that built specialized simulation, sensor-fusion, and control-stack tooling for self-driving cars created durable value because they understood the physics and latency requirements of the domain. Fusionality is making a comparable bet: that fusion-specific control-system tooling, built by people who have worked hands-on with tokamaks and ML interfaces, will be more valuable to reactor developers than off-the-shelf industrial control software.
The optimism here is warranted but measured. Fusion energy has absorbed decades of investment and engineering effort without delivering grid-scale power. The current wave of private fusion startups has shortened the timeline rhetoric, but plasma confinement at net energy gain — producing more energy than the reaction consumes — remains an unsolved engineering problem. Fusionality is not promising to solve that problem. It is promising to build the control and simulation infrastructure that other companies need to solve it faster. That is a narrower claim, and a more credible one, than another startup promising commercial fusion by the end of the decade. If the company delivers on its initial plans for magnetic confinement fusion tooling, it will give reactor developers one fewer reason to reinvent infrastructure and one more reason to focus their scarce talent on the physics problems that remain.
