From numerical simulation to inference: how PhysicsX is rewriting the industrial engineering stack
Key takeaways from the Scaling Europe interview with PhysicsX co-founder and CEO Jacomo CorboDealroom has a profile for this one. Try Dealroom → (July 2026):
The core pitch. PhysicsX is building an AI-native engineering simulation stack that replaces legacy numerical simulation with inference — AI models that predict physics and chemistry, running 10,000 to close to a million times faster and trained not just on simulation data but on real-world measurements from wet labs, wind tunnels, dynos and even jet engines in flight.
Where it is deployed. The technology spans aerospace and defence, semiconductors (equipment makers and increasingly the fabs), automotive and processing-intensive industries such as lithium, copper, steel and aluminium — for example controlling casting processes in a foundry to mitigate defects, or managing aluminium smelting to cut electricity use and improve yield.
Demand outstripping supply. Corbo describes an inflection in demand across industrial sectors, including the data-centre supply chain (gas turbines, chilling units, step-down transformers), leaving the company supply-side limited — customers now wait close to two quarters for a next deployment.
Scaling laws for physics. Over the past 18 months PhysicsX has established data and model-size scaling laws for its pre-trained models: larger models trained on more data generalise better, giving the conviction to spend on building larger, more capable large physics models.
Recent raise. The company recently closed a $300M Series C at a $2.4bn valuation — over double its Series B valuation in under 12 months — led by Temasek with participation from Nvidia, General CatalystDealroom has a profile for this one. Try Dealroom →, Atomico and Siemens.
Compute partnership. PhysicsX is working with Deutsche Telekom and Nvidia on an industrial AI cloud — used both to train its largest pre-trained models and to mobilise training compute closer to customers who run production workloads on Azure and AWS.
Building from Europe. Around 190 of its colleagues sit in London (about 250 across the continent), roughly 80 in New York, with growth into Singapore and the Bay Area. Corbo argues Europe's engineering and physics talent depth is a tremendous advantage, and that building there forced global thinking from inception — though US and Asian adoption still outpaces Europe's.
The future engineer. Corbo expects simulation, systems and process engineers to move up the abstraction ladder, operating end-to-end with fewer handoffs — "the 10x engineer, but in industrials."
Read more: PhysicsX
