Jacomo Corbo was in the middle of a computer science PhD at Harvard when he stepped out of it. The reason was not a lab. It was a Formula 1 team.
That detour matters more than it looks. On the grid, engineers live and die by simulation: how air bends over a wing, how a tyre heats. Every run costs hours of computing, and the season does not wait. Somewhere in that world Corbo met Robin Tuluie, a veteran who had led research and development at two F1 teams, Renault's Alpine and Mercedes, and before that run vehicle technology at Bentley.
Engineers in aerospace, defence and cars were still waiting on slow solvers from the likes of ANSYS and Dassault Systemes.
Corbo's own path then turned in two directions at once. He became Chief Scientist at QuantumBlack, the data consultancy, and he also worked as Chief Race Strategist at Renault's F1 team. Numbers by day, split-second calls by weekend. Sources do not say what he felt in those years, and nothing here pretends to.
What the two men saw was a gap. Engineers in aerospace, defence and cars were still waiting on slow solvers from the likes of ANSYS and Dassault Systemes. Machine learning could learn the physics from past simulations and answer in seconds. The idea was simple to say and hard to sell, because the customers build things that must not fail.
The company took shape in London. Sources disagree on the exact date: Dealroom lists 2019, while some press says 2023. In November 2023 it raised a Series A, and then the quiet part ended. By June 2025 it had closed a EUR 117.3 million Series B led by Atomico, with Temasek, Siemens and Applied Materials joining. The firm said revenue had more than quadrupled since the Series A, and headcount passed 150.
Nvidia's venture arm added $20 million that November, bringing the Series B past $155 million at a valuation near $1 billion. Then, in June 2026, secondary reports said a $300 million round led by Temasek had pushed it to $2.4 billion. Treat that last figure with care; the coverage is thin.
There is a price to being early in industrial AI. Contracts are slow, trust is earned model by model, and the incumbents are not standing still. The founders have said little publicly about the lows, and the record on their private lives is thin.
What the story does show is where the idea came from: not from a pitch deck but from the pit wall, where simulation either saves a race or wastes a night. For a first-time founder, the lesson is plain. The unfair advantage is often a problem you watched up close for years before anyone paid you to solve it.