Jeff Bezos sat down with CNBC’s David Faber at Prometheus’s San Francisco headquarters and, for the first time, explained publicly what he is building with his money. The company he co-leads, Prometheus, announced it had raised $12 billion at a $41 billion valuation.
The round, a Series B, is one of the largest ever raised by an AI startup, and it is Bezos’s first operational role in a technology company since he stepped down as Amazon’s chief executive in 2021. The investors include JPMorgan Chase, Goldman Sachs, BlackRock, DST Global and Arch Venture Partners, along with Bezos himself. The valuation came in above the roughly $38 billion reported in April, a 7.9% markup at the final close.
Prometheus launched in November with a $6.2 billion initial raise, the Series A, in which Bezos was the largest backer. The new round brings total funding to more than $18 billion, a figure that puts the company in the same financial league as the frontier AI labs it says it is not competing with.
The company’s stated ambition is different from OpenAI’s or Anthropic’s. Prometheus is building what it calls an “artificial general engineer”: AI systems that can design and manufacture complex physical things, from jet engines to drug molecules. Where the leading AI labs sell models that reason over text, Prometheus wants models that reason over materials, physics and manufacturing processes, then drive the tools that make the objects real.
Bezos’s partner in the venture is Vik Bajaj, a former co-founder of Verily, Google’s life sciences unit, who serves as co-chief executive alongside Bezos. The company employs about 150 people, a deliberately small operation for a startup valued at $41 billion. In the CNBC interview, the two executives described the bet in industrial terms: if AI can compress the time it takes to design, test and produce a physical product, it changes the economics of manufacturing the way Amazon’s logistics changed retail.
The round is also a statement about where AI capital is flowing. Venture investors have poured money into “physical AI” in recent months, betting that the sector is more defensible than pure software because the physical world creates constraints, patents and supplier relationships that code alone does not. A model that learns to design a turbine blade is harder to copy than a chatbot, the argument goes, because the expertise is embedded in years of engineering data and factory processes.
The size of the round matters for its own sake. A $12 billion raise at a $41 billion valuation tells the market that investors believe the appetite for AI infrastructure has no ceiling, and that the next phase of the boom belongs not to text generation but to embodied, physical systems. It also signals that Bezos, who built Amazon into a logistics and cloud computing colossus, intends to be a principal in the AI buildout rather than a passive observer.
The company has disclosed little about its technology or its customers. Prometheus says it is working with manufacturers in aerospace, energy and pharmaceuticals, and that its systems are being tested on design problems that would take human engineers months to solve. The executives said the company’s goal is not to replace engineers but to give them tools that automate the most expensive parts of the design-to-production loop.
The valuation also raises the bar for Prometheus’s competitors. Physical AI startups that raised at far lower valuations now face a yardstick set by a company with Bezos’s name attached, and the funding gap between the frontier AI labs and the physical AI sector has narrowed to the point where both are raising at similar scales. Investors said the round will force a reckoning over which physical AI approaches can actually deliver: the field is crowded with simulation companies, robotics firms and materials startups, and capital is beginning to concentrate in the few with the resources to build at industrial scale.
The timing is deliberate. The AI industry has spent three years proving that models can reason over language; the next frontier is proving they can reason over matter. Prometheus’s pitch to manufacturers is that the design-to-production loop, which today runs on human specialists, spreadsheets and years of iteration, can be compressed by systems that generate designs, simulate their behavior and drive the machines that build them. Jet engine components, pharmaceutical molecules and alloy formulations are the targets the company names, and each is a market where a small improvement in design speed translates into billions of dollars.
For the industry, the round raises the stakes. The frontier AI companies have spent billions on compute to improve reasoning; Prometheus is spending billions to connect reasoning to the physical world. Whether the strategy pays off will take years to judge, but the money is already moving. Bezos, who rewrote the rules of e-commerce and cloud computing, is now betting that he can do the same for the factory floor, and he has $18 billion of his own and others’ capital to prove it.
