PhysicsX: The F1-Born British AI Startup Rewriting How Hardware Gets Engineered
A London company founded by Formula 1 engineers is teaching AI to understand physics itself — and using it to redesign jet engines, chips, and cars in a fraction of the usual time. Here's who PhysicsX is, what they've built, and why a 3D printing shop is paying attention.
An AI Company for the Physical World
Most AI headlines are about chatbots and images. PhysicsX is chasing something harder: teaching AI to reason about the laws of physics, so engineers can design real hardware at the speed of software.
PhysicsX calls itself a "physical AI" company, and it's a fair description. Instead of generating text, its models predict how a shape will behave under airflow, heat, stress, and load — the kind of analysis that traditionally requires hours of supercomputer simulation. The startup builds an AI-native software stack that shifts heavy numerical physics simulation over to fast AI inference, so design, testing, and optimization that once took weeks can happen in seconds.
Its customers aren't hobbyists. PhysicsX works with leading organizations across aerospace and defense, automotive, semiconductors, materials, and energy — exactly the industries where a half-percent gain in efficiency is worth a fortune, and where being wrong is expensive.
From the Pit Lane to the Frontier of AI
PhysicsX's motorsport DNA is not a marketing gimmick — it's the origin story. The company was co-founded in 2023 by two people who spent careers squeezing physics for performance.
Robin Tuluie
Co-founder and chairman. Formerly Head of R&D at Renault (Alpine) F1 and Mercedes F1, and Vehicle Technology Director at Bentley Motors. A physicist who's spent decades turning simulation into lap time.
Jacomo Corbo
Co-founder and CEO. Former Chief Scientist and co-founder of QuantumBlack (later AI by McKinsey), and Chief Race Strategist at Renault (Alpine) F1. The bridge between hard AI research and industrial deployment.
"The tectonic plates of the global economy are being reshaped by industrial manufacturing... innovation within these fields has never been more urgent."
— Jacomo Corbo, CEO & Co-founder, PhysicsXPhysics Foundation Models
The technical heart of PhysicsX is a pair of model families that work together on its engineering platform:
Large Physics Models (LPM)
Trained on enormous corpuses of simulation data, these predict how a design performs — aerodynamics, structural stress, flow, thermal behavior — at near-simulation accuracy but orders of magnitude faster.
Large Geometry Models (LGM)
Models that understand and manipulate 3D geometry. They generate design variants while baking manufacturing constraints — like minimum wall thickness or draft angles — directly into what they produce.
The Aerospace Flagship: LGM-Aero & Ai.rplane
In late 2024 PhysicsX released LGM-Aero, billed as the first Large Geometry Model for aerospace engineering. It was trained on a corpus of more than 25 million geometries and tens of thousands of computational fluid dynamics (CFD) and finite element analysis (FEA) simulations — built using Siemens Simcenter software and provisioned on AWS. Alongside it came Ai.rplane, a free, public-access demonstrator that lets anyone generate a flying shape and instantly see predicted aerodynamic performance, stability, and structural stress, in under a second, where traditional simulation would take hours.
Why "in seconds" is the whole point
The bottleneck in advanced engineering isn't ideas — it's the time to test them. By collapsing each simulation from hours to a fraction of a second, PhysicsX lets engineers explore thousands of design variations where they'd previously evaluate a handful. The result is access to designs that human intuition and brute-force computing would never have reached.
Traditional Simulation vs. The PhysicsX Approach
| Dimension | Traditional CFD / FEA | PhysicsX AI Inference |
|---|---|---|
| Time per evaluation | Hours on a cluster | Often under a second |
| Design space explored | A handful of variants | Thousands of variants |
| Hardware needed | Large compute clusters | GPU inference, cloud-served |
| Workflow role | Verification near the end | Exploration across the whole lifecycle |
| Output | Point results | Full-field predictions with confidence intervals |
| Best at | High-fidelity ground truth | Rapid, broad optimization |
Crucially, PhysicsX isn't trying to delete traditional simulation — it uses it. High-fidelity solvers generate the training data, and an "active learning" loop triggers fresh high-fidelity runs whenever the AI is uncertain. It's a hybrid: the reach of AI with the reliability of physics.
Funded Like a Foundational Bet
PhysicsX has raised capital quickly and from telling names. It emerged from stealth in late 2023 with a $32M Series A. In June 2025 it raised a $135M Series B led by Atomico, with Temasek, Siemens, and Applied Materials joining. By late 2025, an extension from NVIDIA's venture arm pushed the round past $155M and the valuation toward $1 billion — with NVIDIA signaling it could commit up to $100M across rounds.
In mid-2026, reporting indicated a $300M Series C led by Temasek that valued the company at roughly $2.4 billion, with continued backing from chip heavyweights NVIDIA and Applied Materials. The pattern is unmistakable: strategic industrial and semiconductor players, not just generalist VCs, are betting that engineering itself is going AI-native.
Read the investor list, not just the number
Siemens makes the simulation software. NVIDIA makes the chips that run the AI. Applied Materials builds semiconductor equipment. When your strategic investors are the companies whose tools and customers you depend on, it's a vote of confidence in the technology — and a hint at the ecosystem PhysicsX is plugging into.
Where Dreaming3D Sits in This Story
We're a San Diego 3D printing shop, not an aerospace AI lab — so why are we writing about PhysicsX? Because they're proof of a trend that runs straight through our world: hardware is increasingly imagined, designed, and proven in software before anyone touches a machine. Companies like PhysicsX are the front of that pipeline. Physical fabrication — machining, casting, and 3D printing — is the back.
The catch is that AI-optimized designs tend to be weird: organic, lattice-filled, consolidated shapes that conventional manufacturing struggles to make. That's exactly where additive manufacturing shines, and it's a thread we pick up in our companion piece on whether PhysicsX does 3D printing (spoiler: not directly — but the connection is real). For founders and engineers here in San Diego, the takeaway is simple: the design tools are getting radically faster, and you'll still need someone local who can turn a file into a part.
Got an Ambitious Design? Let's Make It Real.
Whether your geometry came from CAD, a generative tool, or a 3D scan, Dreaming3D turns files into physical FDM and resin parts — fast, and local to San Diego.
Start a Print RequestCall/Text: 858-342-6984 · Email: dreaming3dprinting@gmail.com
Web: dreaming3d.net · Instagram: @dreaming3dprinting
PhysicsX FAQ
What does PhysicsX actually do?
PhysicsX builds AI software for engineering. Its models predict how physical designs behave — aerodynamics, stress, heat, flow — and generate optimized geometry, replacing hours of traditional simulation with AI inference that runs in seconds. It's used in aerospace, defense, automotive, semiconductors, materials, and energy.
Who founded PhysicsX and when?
It was co-founded in 2023 by Robin Tuluie (ex-Renault and Mercedes F1 R&D, ex-Bentley) and Jacomo Corbo (ex-QuantumBlack/McKinsey and Renault F1 race strategist). The company is headquartered in London with a New York office.
How much is PhysicsX worth?
After a $32M Series A in 2023 and a $135M Series B in 2025 (later extended past $155M with NVIDIA backing, near a $1B valuation), 2026 reporting indicated a roughly $2.4 billion valuation on a $300M Series C led by Temasek. Figures move quickly, so treat valuations as point-in-time.
What are Large Physics Models and Large Geometry Models?
Large Physics Models (LPMs) predict physical performance from a design. Large Geometry Models (LGMs) understand and generate 3D shapes while respecting manufacturing constraints. Together they let engineers generate a design and instantly estimate how it will perform.
What is Ai.rplane?
Ai.rplane is a free, public-access demonstrator built on PhysicsX's LGM-Aero model. You can generate an aircraft shape and see predicted aerodynamic performance, stability, and structural stress almost instantly — a showcase of how fast AI-based engineering can be compared to traditional simulation.
Does PhysicsX make 3D printers or print parts?
No. PhysicsX is a software company — it designs and optimizes geometry but doesn't manufacture anything. Its outputs still need a production route, which is often additive manufacturing (3D printing) because AI-optimized shapes are so complex. We dig into that connection in our companion article.
How does this relate to a local 3D printing service?
AI design tools generate the geometry; someone still has to make the physical part. Dreaming3D handles that last step locally in San Diego — printing prototypes and functional parts on FDM and resin from whatever file you bring, including complex, optimized designs.
Design in Software. Build With Dreaming3D.
The future of engineering is digital — but parts are still physical. We're your San Diego partner for turning designs into FDM and resin prints, scanning, and modeling.
Visit Dreaming3DCall/Text: 858-342-6984 · Email: dreaming3dprinting@gmail.com
Web: dreaming3d.net · Instagram: @dreaming3dprinting
Alt Headline Options — delete before publishing
1. Who Is PhysicsX? The London AI Startup Teaching Machines Physics
2. Physics Foundation Models: How PhysicsX Is Speeding Up Engineering by Orders of Magnitude
3. From Formula 1 to a $2.4B Valuation: The Rise of PhysicsX