PhysicsX Raises $300M to Scale Physics AI

PhysicsX announced an oversubscribed $300 million Series C that values the London-based engineering AI startup at approximately $2.4 billion, according to the company and reporting by Sifted and Tech.eu - roughly double the just-under-$1 billion valuation from its 2025 Series B. Temasek led the round, with new institutional backers M&G Investments and Intrepid Growth Partners and increased stakes from existing investors including Applied Materials, NVIDIA, Atomico, General Catalyst, and Siemens. Reporting says the company has grown from about 150 to 350 employees over the past year.

Per the company and coverage in The Next Web and Tech.eu, PhysicsX develops an AI-native simulation stack that combines learned physical inference with conventional numerical methods. It uses neural-network models it calls Large Physics Models to generate physical predictions far faster than traditional CAE (computer-aided engineering) solvers, targeting verticals including aerospace, automotive, semiconductor manufacturing, energy, and materials.

The company says the Series C will expand its platform and AI research, accelerate US growth, and fund a new office in Singapore, Temasek's home market.

Applying large learned models to physical systems follows a broader pattern in which machine learning augments or replaces parts of expensive numerical pipelines. Teams pursuing physics AI typically invest in training-data curation, hybrid modeling that blends learning with conservation laws, and validation against high-fidelity solvers and physical testbeds. For practitioners, that implies continued demand for high-dimensional simulation-data infrastructure, physics-informed verification tooling, and MLOps practices that integrate learned surrogates with conventional solvers.

The valuation and investor mix are widely read as evidence of investor appetite for industrial AI that can shorten hardware development cycles. The presence of strategic investors such as Applied Materials and Siemens points to interest from chipmaking and industrial-automation incumbents, not just software buyers.

PhysicsX frames its value proposition around the physics bottleneck in engineering, arguing that progress on hard problems in the physical economy depends on how fast engineers can work through underlying physics. Co-founder and CEO Jacomo Corbo told Sifted the company will remain headquartered in London, which he described as a "wonderful place" to build a global business.

The Temasek-led, oversubscribed round materially extends PhysicsX's runway and broadens its strategic investor base, underscoring accelerating investment and competition at the intersection of machine learning and high-fidelity engineering simulation. Independent benchmarks and customer pilots will be the clearest tests of learned-surrogate fidelity versus established solvers.