Prometheus, the physical AI startup co-founded by Jeff Bezos and former Verily co-founder Vik Bajaj, raised $12 billion at a $41 billion valuation, one of the largest private rounds ever closed by an AI company. The round was led by Bezos personally, with JPMorgan Chase, Goldman Sachs, and BlackRock joining. It is the second raise for Prometheus in under a year, following a $6.2 billion initial close late in 2025 — meaning the company has pulled in more than $18 billion of equity capital since launch.
Prometheus is building what it calls an 'artificial general engineer' — software meant to automate the design and manufacturing of complex physical systems, from jet engines to drug compounds. The pitch is to replace large portions of mechanical, chemical, and systems engineering work with AI that can iterate on hardware the way large language models iterate on text. The company has 150 employees across San Francisco, London, and Zurich, and is keeping the specifics of what it has shipped internally under wraps.
Bezos told CNBC that a large share of the $12 billion will go toward compute. That is consistent with the broader physical-AI thesis: training models that reason over geometry, materials, and physics requires more simulation cycles than language pretraining, and the simulation environments themselves are expensive to run. For comparison on what frontier compute now costs at scale, Google is paying SpaceX $920 million per month for compute capacity — a number that sets the floor for what a serious physical-AI program needs to commit to.
Key facts
- 01Prometheus raised $12B at a $41B valuation, with Bezos, JPMorgan Chase, Goldman Sachs, and BlackRock backing the round.
- 02The round follows an initial $6.2B raise late in 2025, bringing total funding past $18B in under a year.
- 03The startup has 150 employees across San Francisco, London, and Zurich, with a large share of capital earmarked for compute.
- 04Prometheus is building an 'artificial general engineer' to automate design and manufacturing of jet engines, drug compounds, and other hardware.
- 05Bezos forecasts 'labor scarcity' from AI productivity gains, breaking with peers predicting widespread job losses.
The $41 billion mark puts Prometheus in the upper tier of private AI valuations, alongside the frontier labs, despite the company having no public product. Investors are betting that physical AI is structurally more defensible than pure software because hardware introduces data, manufacturing, and regulatory moats that code alone cannot. Capital has rotated noticeably into the category over the past year, with money flowing to defense-hardware, robotics, and applied-science startups that would have struggled to raise at these marks two years ago.
Bezos used the raise to stake out a contrarian position on AI and jobs. Rather than mass displacement, he argued for what he called 'labor scarcity' — a world in which AI-driven productivity gains push demand for human workers above supply. His framing puts him at odds with a number of AI executives who have publicly forecast significant white-collar job losses over the next several years.
The position is striking given Bezos's day job. He remains executive chairman and the largest individual shareholder of Amazon, which employs more than 1.5 million people worldwide. Under CEO Andy Jassy, Amazon has laid off tens of thousands of workers over the past year while accelerating internal automation, including warehouse robotics and AI-driven middle-office work. The labor-scarcity thesis and Amazon's headcount cuts are not strictly contradictory — Bezos is talking about the long-run macroeconomy, Jassy is managing a quarterly P&L — but the gap between the two stories is wide enough to notice.
Prometheus is also notable for who is not in the cap table. The round is dominated by Bezos and large financial institutions rather than the venture firms that have driven most frontier-AI rounds. That structure mirrors how some of the most capital-intensive AI bets are now being financed: directly by ultra-high-net-worth principals and balance-sheet investors, rather than through traditional Sand Hill funds that lack the check size to keep pace.
The risks are real and largely unproven. No one has yet demonstrated that a general-purpose model can design a working jet engine or a novel drug compound end-to-end. Physical AI also runs into harder constraints than language models — incorrect outputs in hardware design can produce parts that fail in the field, with consequences that token-level hallucinations do not carry. Prometheus has not disclosed benchmarks, partners, or any shipped systems, so the $41 billion valuation rests almost entirely on the founders, the thesis, and the compute budget the round buys.
What the Prometheus round signals for the AI market is that the capital wall around physical AI has fallen. Until recently, the heaviest checks went to model labs whose output was tokens; now investors are willing to underwrite $12 billion rounds for companies whose output will be parts, molecules, and machines. If Prometheus delivers even partial automation of mechanical or chemical engineering, the competitive pressure lands not on OpenAI or Anthropic but on the incumbent industrial design and CAD vendors — a much larger and more fragmented market, and a much slower one to defend itself.
Working on something we should cover, or seeing a story we missed? Send leads, documents, or feedback to hello@aichatdaily.com. For sensitive tips, see our secure tips page for Signal and PGP options.
Spotted an error? Email hello@aichatdaily.com with the URL and the issue, or read our full corrections policy.
