PhysicsX Just Raised Another $300M — Temasek Bets on 'Physics AI' That Cuts Days-Long Simulations to Seconds

AI moved past words and images — now it's computing the laws of physics

Here's the deal: on June 8, London-based AI simulation startup PhysicsX announced an oversubscribed $300 million Series C at a ~$2.4 billion valuation. Singapore sovereign fund Temasek led, with M&G Investments and Intrepid Growth Partners joining new. The returning list is heavy too — Applied Materials, Atomico, General Catalyst, NVIDIA, and Siemens all bet again.

What PhysicsX does is "physics AI" simulation. Put simply, when you design an airplane wing or a chip, you need to compute "how will air resistance behave, how will heat spread" — and PhysicsX uses AI to make that brutally fast. Simulations that used to run for days on a supercomputer, PhysicsX claims to predict in seconds with an AI model.

The numbers back it. Over the past year, recognized revenue doubled, booked revenue tripled, customer count more than doubled, and the team grew past 300, doubling in size. Funds go to global growth, platform upgrades, larger pre-trained physics AI models, plus US expansion and a new Singapore office. It's a round that shows AI burrowing past chat and images into the core of engineering.

The cast — who's who

Start with PhysicsX, which bills itself as the physics AI company for industrials. Its core mission is "speeding up hardware innovation." When you build a physical product — plane, car, chip, energy gear — the biggest bottleneck is simulation. Every design change means recomputing physical behavior, which takes hours to days. PhysicsX's AI models cut that to seconds, letting engineers test far more designs far faster.

Temasek, which led, is Singapore's sovereign fund — a giant managing hundreds of billions and known for long-term bets on future industries. Its leading the round signals more than financial interest; it touches national-level industrial strategy. PhysicsX opening a Singapore office fits that.

The returning lineup tells you what kind of company this is. NVIDIA holds the GPUs that accelerate simulation; Siemens is a powerhouse in industrial automation and digital twins; Applied Materials is a semiconductor-equipment giant. All of them being in is proof PhysicsX is tech used on real industrial floors — not just software VCs, but actual manufacturing and chip heavyweights betting strategically.

What exactly happened

Let's get the facts straight. On June 8, PhysicsX announced a $300 million Series C that was oversubscribed — meaning investors wanted more than the allocation. Valuation: ~$2.4 billion. Temasek led; M&G and Intrepid joined new; Applied Materials, Atomico, General Catalyst, NVIDIA, and Siemens participated as existing backers. Use of funds: global growth, platform expansion, frontier research (larger pre-trained physics AI models), US expansion, and the Singapore office.

The technical crux is "predicting physical behavior in seconds." Traditional simulation directly solves physics equations via methods like FEM (finite element) or CFD (computational fluid dynamics) — precise, but heavy. PhysicsX pre-trains AI on physics data so that when a new design comes in, it predicts the result fast without fully solving the equations — keeping reasonable accuracy while boosting speed by orders of magnitude. When that works, the number of designs an engineer can test per day explodes.

Item	Detail
Announced	June 8, 2026
Raise	$300M Series C (oversubscribed)
Valuation	~$2.4B post-money
Lead	Temasek
New backers	M&G Investments, Intrepid Growth Partners
Returning	Applied Materials, Atomico, General Catalyst, NVIDIA, Siemens
Core tech	AI physics simulation for design (days → seconds)
Growth	Recognized rev 2x, booked rev 3x, customers 2x+, team 300+ (1yr)
Sectors	Aerospace, automotive, semiconductors, energy
Use of funds	Global expansion, platform, pre-trained models, US/Singapore

The picture: PhysicsX isn't at "demo stage" — it's a working company with fast-growing revenue. With heavyweights like Temasek, NVIDIA, and Siemens on top, "AI physics simulation" is moving into the industrial mainstream.

What each side gets

PhysicsX is the most direct beneficiary. $300M secures the compute to build larger pre-trained physics models and to expand into the US and Singapore. The simulation market's customers are high-barrier industries — aerospace, automotive, semiconductors — which means sticky, long-term usage once you're in. Spend the money to widen that base fast and you harden a moat before late entrants arrive.

Temasek and the strategic backers bet on "the infrastructure of future industry." If AI physics simulation becomes standard, every hardware company adopts it — and they're staking out that chokepoint early. For industrial giants like Siemens and Applied, there's strategic synergy beyond financial returns: "how do we wire this into our products and customers?" NVIDIA, again, is the pickaxe seller — faster simulation means more GPU demand.

Industrial engineers gain a lot too. Where a single design change once meant waiting days for results, seconds changes how work happens. From reviewing one design a day to running dozens or hundreds. Ultimately that means better planes, more efficient chips, and safer cars, faster.

History check — does "AI for science and engineering" work?

There's a landmark success in applying AI to scientific simulation: DeepMind's AlphaFold. Predicting protein structures was a problem that took enormous time even on supercomputers, and AlphaFold cracked it fast with AI, transforming biology. PhysicsX is going after the "physics and engineering version" — predicting phenomena like airflow, heat, and stress with AI instead of proteins. The thesis AlphaFold proved — that AI can massively accelerate scientific simulation — is extending into industrial design.

Another success thread is "AI surrogate models" — approximating heavy physics simulations with lightweight AI. The technique is already established in academia and aero/auto R&D. PhysicsX is one of the firms lifting it from lab-grade to commercial product, with fast revenue growth as proof. So the tech itself is validated; now it's a fight over "how broadly and how accurately you commercialize."

But the shadow is clear. AI simulation's biggest weakness is "trust in accuracy." In fields like aerospace and semiconductors, where a tiny error is catastrophic, engineers can't simply trust "what the AI predicted." You can't bet a wing's structural integrity on an AI approximation alone. So critical stages often get re-verified with traditional simulation. History has plenty of "fast but inaccurate" approximations that never earned trust and stayed in the lab. PhysicsX's real test isn't speed — it's proving trustworthy accuracy on the industrial floor.

Competitor counter-plays — how do the others respond?

First, the traditional simulation incumbents. Ansys, Dassault Systèmes, and Siemens have ruled industrial simulation for decades. Their counter-play is "embed our own AI" — bolting AI acceleration onto existing software to say "you don't need a startup, we get faster too." Siemens investing in PhysicsX is two-sided — part hedge, part collaboration, part insurance.

Big Tech's AI-for-science teams are a variable. NVIDIA pushes physics-simulation platforms like Earth-2, and Microsoft and Google spend heavily on "AI for Science." If these giants release general-purpose physics AI models, specialists like PhysicsX must differentiate on deeper domain expertise — "general from Big Tech, but the deep know-how of aerospace and chips is ours."

The open-source trend could be a counter too. As more physics AI models get open-sourced, companies may say "we'll deploy it ourselves rather than pay for a commercial solution." PhysicsX's defense is that it sells more than a model — industrial data, accuracy validation, and integration services bundled together. Even with an open model, validating and operating it to aircraft-certification standards is another job entirely. Creating value there is the survival strategy.

There's also a data-moat dimension worth naming. The reason PhysicsX can claim seconds-not-days isn't a clever model alone — it's the proprietary physics data it has accumulated training on real aerospace, automotive, and semiconductor problems. A rival can copy an architecture, but it can't easily copy years of paid customer engagements that quietly generated labeled physical-behavior data. That's the same flywheel you see in robotics (whoever collects the most physical data wins) showing up in simulation. Each new customer feeds the models, the better models win the next customer, and the gap compounds. If PhysicsX uses this $300M to widen that data lead while incumbents are still bolting AI onto legacy tools, the moat could harden faster than the competitive response — which is exactly the bet Temasek, NVIDIA, and Siemens just made.

So what actually changes — by audience

For manufacturing and hardware firms, the key is that "design speed" becomes a core competitive edge. As AI physics simulation spreads, the company that tests far more designs in the same time ships better products. In industries where design is competitiveness — auto, aero, semiconductors — how fast you adopt this tool could become a product gap years later.

For AI and deep-tech founders, this signals that "AI for engineering" is a genuinely lucrative market. Big Tech owns chatbots and image AI, but domains requiring deep expertise — like industrial simulation — leave room for startups. PhysicsX's fast revenue growth and $2.4B valuation show the upside. Just weigh it against the high wall of "accuracy trust."

For investors, read this as "another face of physical AI." If robots (Generalist AI) are about "giving AI a body," PhysicsX is about "computing the laws of physics with AI." Both are branches of the same big move — AI expanding past digital into the physical world. But simulation has a clearer revenue path than robots — there are already paying industrial customers. Keep that different risk/return profile in mind.

In sum, PhysicsX's $300M signals that "AI is becoming a core tool of engineering." The AI that wrote text and drew images is now computing the air resistance of a wing and the heat of a chip. It won't be dramatic overnight, but it'll slowly change how the cars, planes, and chips we use get built.

FAQ

Q. What exactly is good about AI physics simulation? A. When designing planes, cars, or chips, it cuts the work of computing physical behavior (air resistance, heat, stress) from days to seconds. That lets engineers test far more designs per day, building better products faster.

Q. So do traditional tools (like Ansys) become unnecessary? A. Not soon. AI simulation is fast, but in fields like aerospace where tiny errors are fatal, critical stages often get re-verified the traditional way. The two will likely be complementary for a while — "explore fast with AI, validate finally with traditional methods."

Q. Isn't a $2.4B valuation too expensive? A. Opinions can differ. But PhysicsX isn't at demo stage — it's a company with fast-growing revenue (booked revenue tripled), so the valuation isn't purely on a promise. Still, it's a private-market estimate and highly volatile.

Q. NVIDIA invested here too? Why? A. Faster simulation means more demand for the GPUs that run it. Whether physical AI or physics simulation, it all runs on NVIDIA chips — so it bets on promising names with the "pickaxe seller" strategy.

References

• PhysicsX Announces $300M Series C to Accelerate Physics AI for Industrial Engineering — PhysicsX Newsroom
• PhysicsX reels in $300M to speed up hardware design with AI — SiliconANGLE
• Temasek leads UK AI firm PhysicsX $300M Series C funding — TNGlobal
• PhysicsX From London Raises $300M, Valuation Surges to $2.4 Billion — Trending Topics
• NVIDIA-Backed PhysicsX Raises $300 Million At $2.4 Billion Valuation — Benzinga

This article is not investment advice. The raise, valuation, and growth metrics reflect company announcements and press reports and may differ from reality; valuation in particular is a private-market estimate and highly volatile. Make decisions at your own discretion and risk.