
Prometheus is the industrial AI startup co-led by Jeff Bezos, building an "artificial general engineer" to automate the design and manufacturing of physical products like jet engines, chips and drug compounds.
The company has raised $18.2 billion across two rounds in under eight months, reaching a reported $41 billion valuation with institutional backing from JPMorgan, Goldman Sachs and BlackRock.
Pre-IPO access to this allocation is now available on WLTH.
There's a useful test for whether an AI company's valuation is real or theater: ask what happens if it never ships a product for another two years. Most AI startups fail that test immediately. Prometheus is one of the few that doesn't.
Jeff Bezos hasn't run a company day to day since he left Amazon's CEO chair in 2021. He just started again, as co-CEO of an industrial AI startup that has raised $18.2 billion in eight months and sits at a reported $41 billion valuation. That combination of enormous capital and near-total silence is the clearest signal yet that institutional money has moved past chatbots and into physical AI as the next platform shift.
Bezos co-founded Prometheus in November 2025 with Vik Bajaj, a chemist and physicist who spent time at Google X before co-founding Verily, Alphabet's life sciences division, and later Foresite Labs. Bezos took the co-CEO title himself. It's his first day-to-day operating role since Amazon.
The company is headquartered in San Francisco, with teams in London and Zurich, and had grown to roughly 150 people by mid-2026, several of them researchers pulled out of OpenAI, DeepMind, Meta and Nvidia. Shortly after launch, Prometheus also acquired General Agents, an agentic AI startup, and folded its team in.
What they're building, by their own description, is an "artificial general engineer." Not a chatbot, not a robot. Bezos has called it a modern version of CAD software, then immediately noted that comparison is too simple for what the system is meant to do. Bajaj has been more concrete: a jet engine takes engineering teams a decade or more to design, prototype and manufacture, and that's the kind of end-to-end problem Prometheus wants to compress into an AI-driven workflow. Aerospace, chip manufacturing, automotive production, drug design. Bezos has said the company has nothing to do with robotics, and that Prometheus has no formal ties to Amazon or Blue Origin, beyond floating Blue Origin as a future customer.
Most of what's notable about Prometheus isn't the technology pitch. It's the speed and size of the capital behind it.
Series A closed in November 2025 at $6.2 billion, with Bezos as the largest individual backer. Seven months later, Series B closed at $12 billion, valuing the company at a reported $41 billion. JPMorgan Chase, Goldman Sachs, BlackRock, DST Global and Arch Venture Partners all participated, a roster that reads like a late-stage pre-IPO round, not a Series B.
Bezos has said most of that capital is going toward compute and toward building proprietary training data from scratch. Large language models scaled by training on the existing internet. Prometheus can't do that, because there's no equivalent corpus of physical-world engineering data sitting around waiting to be scraped. The company has to generate it directly, through real experimentation, and that single fact explains the size of these rounds better than anything in the company's marketing.
There are also unconfirmed reports that Bezos and Bajaj are separately pursuing up to $100 billion for a holding company that would acquire legacy industrial businesses and feed their operational data back into Prometheus. One source described it to the Financial Times as a "Berkshire Hathaway-type holding company" for AI-driven industrial transformation. If that materializes, today's $41 billion figure will look like an early marker rather than a ceiling.
A $41 billion valuation with no shipped product would be indefensible by normal venture math if it were backed by typical venture money. It isn't. JPMorgan, Goldman Sachs and BlackRock are not in the business of funding vibes. When that group underwrites a pre-revenue company at this size, the bet isn't on Bezos's name, it's on physical AI, AI systems trained to operate in and learn from the real world rather than purely digital environments, becoming a genuinely separate platform shift from the large language model wave, with its own category-defining winners.
That thesis has company. Fei-Fei Li's World Labs, Yann LeCun's AMI Labs, and Physical Intelligence, which Bezos personally backed in 2024, are all circling versions of the same idea, and Meta, Nvidia and OpenAI all have their own physical AI efforts underway. Prometheus stands out for how much capital has moved to back it before there's a public product to evaluate, which is exactly the kind of early positioning that's historically defined the biggest returns in venture-backed technology: SpaceX, Anthropic and the rest of today's most valuable private AI companies all looked unproven and overcapitalized at this same stage.
Prometheus is underwriting a specific, well-resourced bet: that AI can learn from physical trial and error at real scale, that a team built from the best labs in the world can move faster than the competition while the category is still forming, and that Bezos's operational track record, the one that turned Amazon into the most efficient logistics machine on earth, carries directly into a new industry. The capital behind the company suggests the market is already pricing that bet as likely to pay off.
Because there isn't any to buy, not in any form retail investors can access. Rounds at this size and stage go to existing venture backers, sovereign capital, and a handful of strategic institutions. That door stays closed for years, often right up until a company actually goes public, and by the time Prometheus reaches a listing, most of the value will already have been created and captured privately. This is the same access gap that kept ordinary investors out of early SpaceX and early Anthropic, until platforms built specifically to close that gap existed.
WLTH provides tokenised economic rights, not direct equity, with no shareholder rights in the underlying company, to its allocation in Prometheus. It's structured exposure to potential future value in the company.
Access to this allocation is currently priced at a $47.88 billion valuation.
Access is live now. Additional pre-IPO opportunities across AI, robotics and infrastructure are available here.
Disclaimer: WLTH provides economic exposure via tokenised rights, not direct equity or shareholder rights in the underlying company. This article is for informational purposes only and does not constitute financial or investment advice.