This Is What A $2.4 Billion AI Bet On Jet Engines Looks Like

Most conversations about AI in 2026 are still happening in the same register: chatbots, content, code generation, the occasional enterprise productivity tool. Meanwhile, a quieter and arguably more consequential application is building momentum in the background.

AI is beginning to transform how physical things get designed – the turbines, aircraft, semiconductor chips and data centre cooling systems that the modern economy actually runs on – and the money is starting to follow.

London-based PhysicsX just raised $300 million in a Series C led by Singaporean sovereign wealth fund Temasek, taking its valuation to $2.4 billion – that’s more than double its valuation from less than a year ago. Nvidia, Siemens, Applied Materials and General Catalyst all increased their stakes in the round.

The company now employs 350 people, up from 150 a year ago, generates close to $50 million in annual revenue and is targeting more than double that in 2027. According to The Next Web, its CEO told the Financial Times they’re already moderating rollout to existing customers because demand is outpacing capacity.

The numbers are striking – the reason they’re possible is even more so.

The Anatomy Of AI Engineering

Traditional engineering design relies on physics simulation – running complex mathematical models to predict how a component will behave under real-world conditions. How will this wing profile respond to turbulence? How will this chip package handle heat? How will this turbine blade perform under stress?

These simulations are essential to safe, optimised design, and they’re extraordinarily slow. A single simulation run can take hours, a full design iteration cycle can take months. Engineers run thousands of simulations before a component reaches production.

PhysicsX has built what it calls Large Physics Models – analogous to large language models, but trained on physical equations rather than text. The company trained its models on more than 25 million component geometries, representing tens of billions of mesh elements drawn from computational fluid dynamics and finite element analysis simulations. The result is an AI that can infer the aerodynamic performance, flight stability and structural stress of a component in under a second, rather than several hours. Its Ai.rplane demonstrator – described as the world’s first Large Physics Model for flight – cut aircraft design cycles from months to days.

The application runs well beyond aviation – PhysicsX is targeting aerospace and defence, semiconductors, automotive, energy and renewables, materials manufacturing and data centres. Semiconductors are expected to become its largest segment in 2026, driven partly by the AI infrastructure build-out: every new data centre requires chip packaging, cooling systems and power infrastructure that all need the kind of iterative engineering simulation PhysicsX accelerates.

The proliferation of AI directly drives demand for the tools PhysicsX builds.

Why This Corner Of AI Deserves More Attention

The global industrial AI market was valued at $43.6 billion in 2024 and is projected to reach $153.9 billion by 2030, a compound annual growth rate of around 23%. AI in manufacturing specifically is growing faster, from $9.5 billion in 2025 to a projected $114.5 billion by 2033. These aren’t niche numbers – applying machine learning to the design of physical things is a considerably larger economic activity than generating text or images.

Physical AI – the broader category of AI systems that interact with and operate in the real world – is projected to represent a market of around €430 billion by 2030, according to analysis by Strategy& and PwC. The difference between digital AI (generating content, answering questions) and physical AI (designing turbines, navigating warehouses, simulating stress fractures) is one that investors are actively making, and the capital allocation is starting to reflect it.

PhysicsX is ranked second in Sifted’s AI 100 list of Europe’s most promising AI startups, behind ElevenLabs. Its founders bring an unusual combination of credentials: Jacomo Corbo was chief scientist at McKinsey’s QuantumBlack AI division, and Robin Tuluie led R&D at Renault Alpine in Formula 1 before becoming Bentley’s director of vehicle technology.

Formula 1 engineering – which runs simulations on everything from aerodynamics to tyre compounds to suspension geometry under extreme time pressure – is arguably the best possible training ground for what PhysicsX is building. The same mentality that shaves tenths of a second off a lap time by running ten thousand simulations overnight is now being applied to the design of jet engines and semiconductor packaging.

PhysicsX is staying in London despite expanding into the US and Singapore – that’s an intentional move, and it carries a lot of weight for the whole UK deep tech narrative. A British AI company designing the components inside next-generation aircraft and chips, valued at $2.4 billion and backed by sovereign wealth, is exactly the kind of company the UK’s science and technology ambitions are supposed to produce.

The funding round is the signal – the real story is what PhysicsX’s growth rate says about where the AI economy is actually heading.