Jeff Bezos' Prometheus Raises $12B to Build an 'Artificial General Engineer' for the Physical World

When Jeff Bezos left Amazon’s CEO seat in 2021, he said he planned to use his time on “new products and early initiatives.” Five years later, that initiative has a name, a $41 billion valuation, and a mission statement that would have been science fiction a decade ago: build an artificial general engineer.

Prometheus, the industrial AI startup where Bezos serves as co-CEO, closed a $12 billion Series B financing round on June 11, 2026, according to people familiar with the deal. The raise comes less than a year after the company’s launch, when it debuted with $6.2 billion in initial backing — itself a record for a company at that stage. Combined, Prometheus has now raised more than $18 billion before generating a single line of publicly disclosed revenue.

What Prometheus Actually Does

Bezos was unusually candid about the company’s ambitions at a public event following the close. “Prometheus is the bulk of my time,” he said — a statement that speaks volumes from someone who also chairs Blue Origin, holds a $100 billion-plus investment portfolio, and purchased a major media property.

The company is developing AI software designed to automate the design and manufacturing of complex physical systems. “Artificial general engineer” is the term Prometheus uses internally — not to be confused with artificial general intelligence, but rather the physical-world equivalent: software capable of doing what an elite team of mechanical, chemical, and materials engineers does, across domains ranging from jet engines to pharmaceutical compound synthesis.

Bezos’s co-CEO, Vik Bajaj, brings a complementary background. A Stanford School of Medicine professor, Bajaj previously co-founded Verily — Alphabet’s life sciences research division — where he worked on combining biological data and machine learning for healthcare applications. The pairing of physical-world manufacturing (Bezos’s Amazon supply chain expertise) with life sciences AI (Bajaj’s domain) is not accidental.

The company currently employs approximately 150 people across offices in San Francisco, London, and Zurich. For a company at $41 billion valuation, the headcount is strikingly lean — a deliberate signal that Prometheus is betting heavily on AI doing the heavy lifting, rather than building a traditional engineering workforce.

The Investor Syndicate

The Series B attracted a roster of investors that reads less like a traditional venture capital deal and more like a coalition of institutions making infrastructure bets. JPMorgan Chase, Goldman Sachs, and BlackRock joined the round alongside technology-focused investors DST Global and Arch Venture Partners.

The presence of the three major financial institutions is worth parsing. JPMorgan, which has reclassified AI from experimental R&D to core infrastructure and carries a $19.8 billion technology budget in 2026, is not making a purely financial investment. Goldman’s participation signals that the investment banking community views physical AI as a category, not just a company. BlackRock’s involvement echoes its broader infrastructure-as-asset-class strategy.

Capital will be deployed substantially into computational infrastructure. Designing and simulating physical systems — running materials science calculations, testing aerodynamic models, validating pharmaceutical molecule interactions — requires enormous amounts of specialized compute. Prometheus has been deliberately quiet about its underlying model architecture, but the infrastructure spend suggests frontier-scale training is involved.

Bezos on Labor and Productivity

Asked about the impact of Prometheus’s technology on employment — a question that trails every physical AI announcement — Bezos offered a reframing. Rather than predicting job losses, he argued that AI-driven productivity gains would produce “labor scarcity,” a condition in which demand for workers exceeds supply as the economy expands faster than the workforce can grow.

“Significant productivity in the economy is going to raise the standard of living,” Bezos said, sketching a scenario where AI-augmented engineering makes households wealthier rather than engineers obsolete. He envisions a future where households transition from requiring dual incomes to single incomes, with AI-generated productivity making up the difference.

It is a characteristically optimistic framing from someone who has spent decades defending automation decisions at Amazon. Whether it proves correct depends partly on how fast physical AI capabilities actually mature — a timeline Prometheus itself has an interest in accelerating.

Physical AI vs. Software AI

The distinction between physical AI and the generative AI most consumers interact with is significant. Software AI — producing text, code, images, or reasoning chains — operates in domains with relatively unlimited data and low marginal cost of errors. Physical AI operates under fundamentally different constraints: jet engines have to fly, pharmaceutical compounds have to be synthesized, materials have to hold under load.

This creates a natural moat. Proprietary data from physical manufacturing processes is scarce, expensive to generate, and closely held by incumbent industrial companies. Any company that cracks physical AI at scale will have done so in a domain where competitors cannot simply scrape the internet for training data.

It also explains the valuation. Physical AI represents a greenfield opportunity that pure software AI cannot easily address — the global manufacturing sector generates roughly $14 trillion in annual output, and even a single-digit percentage improvement in engineering efficiency would be worth hundreds of billions of dollars.

Where Prometheus Fits in the AI Landscape

2026’s AI landscape has increasingly stratified into layers: foundation model providers (Anthropic, OpenAI, Google), application layer companies, and the emerging category of domain-specific AI labs targeting sectors where specialized training and proprietary data create compounding advantages.

Prometheus belongs to this third category — alongside other well-funded vertical players in healthcare AI, legal AI, and financial AI — but with a higher degree of technical ambition than most. The term “artificial general engineer” signals that Prometheus is not building a CAD assistant or a materials database. It is building something intended to replace — or radically augment — the entire engineering design and manufacturing workflow.

Whether that ambition translates into a viable product on any near-term timeline remains to be demonstrated. With $18 billion raised, 150 employees, and a co-CEO who describes it as consuming most of his waking hours, Prometheus has the resources and the mandate to try. In an industry that has stopped being surprised by billion-dollar bets, a $41 billion industrial AI company built in under a year is still remarkable.