LONDON: PhysicsX, a UK-based AI engineering company, has closed a $300 million Series C at a $2.4 billion valuation, the company said in a June 8 announcement. Singapore’s state investor Temasek led the oversubscribed round, with M&G Investments and Intrepid Growth Partners joining as new backers. Existing investors NVIDIA, Siemens, Applied Materials, Atomico, General Catalyst, NGP, July Fund, and Radius also participated.
The raise lands roughly one year after the company’s $135 million Series B in June 2025. That step-up implies a valuation increase of close to 18 times in twelve months, putting PhysicsX in a small cluster of industrial AI startups crossing the multi-billion-dollar mark in 2026, alongside infrastructure raises such as Supabase’s $500 million Series F at $10.5 billion.
Why industrial AI is pulling in strategic capital
PhysicsX positions its platform as a replacement for conventional physics simulation. Traditional solvers used in aerospace, semiconductor, and automotive engineering can take hours or days to model how a part will behave under stress, heat, or fluid flow. The company says its AI models return comparable results in seconds, letting engineering teams test orders of magnitude more design variants in the same window.
The core technology is what PhysicsX calls Large Physics Models, or LPMs. The name is a deliberate analogy to large language models, but the training data is simulation results and physical equations, not text. Depending on the problem, the models draw on transformer architectures, geometric deep learning, or neural operators that approximate solutions to partial differential equations. The output is a model that has internalized how a class of physical systems behaves, then generalizes to design variants it has not seen.
The investor list reads as a strategic map of who needs faster physics simulation. NVIDIA supplies the GPUs that train and run the models, and its industrial AI thesis was laid out in Jensen Huang’s Computex 2026 keynote. Siemens runs one of the largest industrial software businesses in the world and announced a commercial partnership with PhysicsX earlier this year. Applied Materials sells the equipment that makes the chips PhysicsX customers design. Temasek’s lead position signals long-horizon conviction from a fund that backs few private software companies at this scale, and its 2025 entry into the cap table gave PhysicsX a path into Asian industrial customers that European startups rarely access.
For founders watching from outside the category, the takeaway is concrete. Industrial workflows that were modeled with finite-element solvers for the last forty years are now in play. The buyers are large, the cycles are long, and the data is proprietary. Those are exactly the conditions that produce defensible AI companies rather than thin wrappers on a public model.
What does PhysicsX’s $2.4 billion valuation mean for industrial AI startups?
The valuation establishes a public reference price for vertical AI applied to engineering workflows. With NVIDIA, Siemens, and Applied Materials on the cap table, PhysicsX now sets expectations for any startup pitching AI-native simulation, materials discovery, or hardware design tooling. Founders in adjacent categories can credibly point to an 18x mark-up in twelve months.
That signal matters because most AI capital in 2025 and early 2026 chased general-purpose models and infrastructure. The PhysicsX round is part of a shift toward applied, domain-specific AI where the moat is proprietary training data. In this case, that means decades of physics simulations and partnerships with industrial operators who own the data. Bezos-backed Flourish AI and other applied-AI startups have raised on a similar logic in recent months.
The company reports doubled year-over-year recognized revenue, tripled booked revenue, and a customer count that more than doubled over the past year. Headcount sits above 300, roughly twice its size twelve months ago. Customers span aerospace and defense, semiconductors, automotive, industrial machinery, energy, and materials. That mix mirrors where the supply chain is most exposed to manufacturing cycles. Shaving days off a design loop in those industries translates directly into delivered units.
The valuation also reframes how late-stage investors price applied AI. A $2.4 billion mark on a four-year-old company selling into traditional industrials is the kind of number reserved for consumer SaaS five years ago. The shift reflects how quickly buyers are willing to commit budget once a model demonstrably collapses a workflow from days to seconds.
What to watch next
PhysicsX said the new capital will fund global expansion, deeper platform development, and frontier research into larger pre-trained LPMs. The research path mirrors the scaling trajectory of language models: bigger pre-training corpora, more parameters, broader generalization across physical domains. Whether physics data scales the same way text data has is the open technical question the next round of research will test.
Two near-term markers matter for founders watching the category. First, whether Siemens or another strategic backer pushes PhysicsX’s platform into a standard design workflow at a Fortune 500 manufacturer, which would mark the first time AI-native simulation displaces a legacy CAE vendor at scale. Second, whether competing LPM efforts emerge. IEEE Spectrum has documented growing academic interest in physics-trained models, which usually precedes a wave of well-funded startups. For founders working on vertical AI, the playbook on display is straightforward: pick a slow legacy workflow with proprietary data, replace it with a model that runs in seconds, and the strategic capital will find you.
