PhysicsX, the London-based startup that has built an AI platform for hardware designers to run simulations of their work in progress, has raised a large Seres C round of funding to speed up its own expansion. It has raised $300 million, at a valuation of $2.4 billion.
PhysicsX is headquartered in London and emerged from stealth in 2023, co-founded by two Italian physicists. Robin Tuluie and Jacomo Corbo worked on aerodynamics for high-end Formula One race cars in Italy and the UK before spotting an opportunity to build an AI platform to improve the prototyping and simulation process for others.
The startup may have its roots in Europe in racing cars, but its latest round speaks to how the startup now attracts more global interest. Temasek, the investment fund wholly-owned by the Singapore government’s ministry of finance, is leading the round, with participation also from UK’s M&G Investments, Canada’s Intrepid Growth Partners, US semiconductor company Applied Materials, Atomico, General Catalyst, July Fund, NGP, Nvidia, deep tech fund Radius, and Siemens.
The resilience play here is two-fold. PhysicsX, despite the international investors, remains European and points to how startups in the region are tackling industrial-scale problems in the region. It’s also throwing a line to manufacturers specifically, by aiming to make their work more efficient, in theory giving customers the ability to create more cutting-edge products faster.
PhysicsX now says its customers include organisations in working in defence, alternative energy, aviation, data centers, and semiconductors. (Leonardo, one of its partners and customers, paired up with PhysicsX’s Corbo on stage at Resilience Conference last autumn: you can watch that here.)
Like Siemens and Applied Materials, Nvidia is a strategic investor. The AI chip giant has been leveraging its vast market presence to diversify its business for the long run (resilience of another kind). Not only is PhysicsX optimised to run on Nvidia chips but it’s a buyer of them, and the pair partner with third parties like Deutsche Telekom in ‘go to market’ solutions. Nvidia first invested in PhysicsX last year, a $20 million extension of its Series B that valued PhysicsX at just under $1 billion.
The rapid growth of PhysicsX’s valuation underscores the company’s progress. Over the last year, PhysicsX claims to have doubled its number of customers, “doubled year-over-year recognised revenue [and] tripled booked revenue.” (It does disclose specific numbers for any of these metrics.) It also now has 300 employees, which — combined with the likely large amount of computer processing power that it needs to buy and use — likely means it has a lot of costs on its books.
It also points to the huge interest that artificial intelligence companies are commanding right now. Anthropic at the start of June confidentially filed for a public listing in the US at a valuation of nearly $1 trillion. OpenAI is likely to follow suit at a similarly eye-watering figure.
It’s not just technology companies that seem to think that the use of AI in how they operate is inevitable. And that expectation, the investment that’s being laid out, and sometimes even actual usage are all leading to a lot more interest in AI companies that are building inroads into specific uses and areas.
That is where PhysicsX comes into the picture. The company started from a specific pain point that the founders saw in their own work: it costs too much and takes too long to build and test physical prototypes from scratch.
They saw an opportunity to use AI to accelerate that process, essentially applying the concept of training an AI on a giant trove of information to work like humans, in its case an AI to build and run simulations in place of humans building incremental actual prototypes.
Instead of the Large Language Models that you get from companies like OpenAI that let users produce writing, images, work plans and coding from regular-language prompts, PhysicsX has created what it calls a “large physics model” that speaks in the language of objects and materials, prompted by words and calculations that its users submit.
“What PhysicsX buys you is the ability to be able to predict the physics [of a system] with very, very high accuracy and fidelity, doing it, anywhere from 10,000 to a million times faster [than traditional prototyping],” Corbo told me in an interview in 2023.
We’ve asked the company to clarify a few points. It’s not clear how it’s charging users for its services, nor how customers actually engage with PhysicsX to create simulations: natural language? blueprints fed into a slot? pages of calculations? all or none of the above?
And we are also curious about how, longer term, PhysicsX sees the product developing. Today, the startup may be doing a lot more than high-end car design, but designers are still in the driving seat — pun intended — deciding what to test out before using PhysicsX to do the simulating.
Longer term, will this lead to “vibe designing”, where the platform might do the hard work of figuring out what to build next and then building it?
In any case, it’s a very notable round, and puts the startup very much in the spotlight as so-called “foundational AI” companies look for more commercial and higher-margin use cases for their own platforms to prop up their enormous valuations. It will be interesting to watch what verticals they move into and what they look to acquire to do that.
