Prometheus raised $12 billion at a ~$41 billion valuation led by Jeff Bezos, JPMorgan, Goldman Sachs, BlackRock, DST Global and Arch Venture Partners. That's the short answer. The longer answer is more interesting.
Because this isn't another chatbot raise. Prometheus is betting that the next frontier of AI isn't writing text or code — it's engineering the physical world, from jet engines to drug molecules. And the man who built the world's largest logistics machine just stepped off the sidelines to run it himself. Here's what the round means, and why every megafund suddenly wants exposure to “physical AI.”
Prometheus $12B raise: round terms and lead investors
Prometheus raised $12 billion at a roughly $41 billion post-money valuation, announced June 11, 2026. The round was anchored by Jeff Bezos as a personal investor and co-led with a syndicate of JPMorgan, Goldman Sachs, BlackRock, DST Global and Arch Venture Partners. Combined with the company's $6.2 billion launch round in November 2025, total capital raised now exceeds $18 billion — across a team of roughly 150 people.
Round size
$12B
Announced June 11, 2026
Post-money valuation
~$41B
Up from launch in Nov 2025
Launch round
$6.2B
November 2025
Total raised
$18B+
Across two rounds
Employees
~150
~$120M raised per head
Valuation per employee
~$273M
Among the highest ever
The headline that should make every founder blink: ~$273M of valuation per employee. Most pre-revenue companies measure capital efficiency in dollars per head; Prometheus measures it in hundreds of millions.
Who is behind Prometheus — Bezos and Vik Bajaj
Prometheus was co-founded by Jeff Bezos and Vik Bajaj, a Stanford-affiliated scientist who previously co-founded Alphabet's Verily life-sciences lab and spent years inside Google X. The notable shift in this round is Bezos's role. He started as the lead financial backer when the company launched in late 2025 — now he's moved into an operational co-CEO seat. In his own words: “I became so impressed by what was happening and the potential that I decided I couldn't sit on the sidelines and I needed to jump in with both feet.”
That matters because Bezos hasn't taken an operating role at a company since stepping down as Amazon CEO in 2021. He has plenty of capital deployed across AI — but Prometheus is the first where he's back in the building. Pair a logistics-and-manufacturing obsessive with a deep-science co-founder, and you get a company explicitly aimed at the messy intersection of atoms and algorithms rather than the comparatively clean world of software.
What Prometheus actually builds: the “artificial general engineer”
Prometheus describes its goal as an “artificial general engineer” — AI that can design and help manufacture complex physical products end to end. The targets span aerospace, automotive, computing hardware, manufacturing and drug discovery. The core technical bet is “physical AI”: instead of training models purely on text and images scraped from the internet, Prometheus trains on real-world experimental data and robotics interactions so the systems learn the actual laws of physics rather than statistical patterns of language.
| Dimension | LLM / chatbot AI | Physical AI (Prometheus) |
|---|---|---|
| Training data | Text, code, images from the web | Experimental + robotics data |
| What it learns | Language & pattern statistics | Laws of physics & materials |
| Output | Words, code, pixels | Engineered designs & processes |
| End market | Software, search, content | Aerospace, autos, chips, drugs |
| Verification | Human review of text | Physical testing & manufacturing |
If it works, the payoff is enormous: collapsing the design-to-manufacturing cycle for jet engines, semiconductors or therapeutics from years to months. If it doesn't, $18 billion will have funded one of the most expensive research labs in history. There is no in-between at this valuation — which is exactly why a pre-revenue company is being priced like a mid-cap public corporation. You can see how AI valuations are being set across the sector on our AI Valuations tracker.
Why physical AI became the new megaround magnet
The Prometheus round lands in a market that has quietly normalized the nine-figure check. Investors backed roughly 250 financings of $100M+ in the first half of 2026, and the median U.S. late-stage round hit exactly $100 million — a number that would have been a blockbuster five years ago. But multi-billion-dollar private rounds remain rare, reserved for a tiny club: Anthropic, OpenAI, xAI, and now Prometheus.
What's changed is the thesis. The frontier-model labs have largely conquered text and code, and capital is hunting for the next trillion-dollar surface area. “Physical AI” — robotics, manufacturing, materials, drug design — is the answer venture has landed on. Bezos is triple-exposed: beyond Prometheus, he's backed robotics-foundation-model companies Physical Intelligence and Generalist AI. The pattern is unmistakable.
Software is crowded
Frontier LLMs are converging; differentiation and margins are getting harder to defend in pure software AI.
Atoms are a bigger market
Aerospace, autos, semiconductors and pharma are multi-trillion-dollar industries barely touched by modern AI.
Data moats are real
Experimental and robotics data can't be scraped from the web — whoever generates it first owns a durable edge.
Capital is patient again
Sovereigns, banks and crossover funds want AI exposure and will underwrite long, pre-revenue R&D bets.
The investor syndicate: why banks, not just VCs, wrote the checks
The most telling detail in the cap table isn't Bezos — it's the banks. JPMorgan, Goldman Sachs and BlackRock are balance-sheet and asset-management giants, not traditional venture firms. Their presence signals that frontier AI has outgrown the classic VC model: the checks are now too large for any single fund, and the institutions that manage trillions in client capital want direct exposure rather than a sliver via a fund-of-funds.
DST Global (Yuri Milner's crossover firm) and Arch Venture Partners (a deep-science specialist that has minted biotech giants) round out the syndicate with the two skill sets the deal actually requires: writing enormous late-stage checks and underwriting hard science. It's the same blurring of lines we've tracked at Anthropic and xAI — when rounds get this big, the buyer list stops looking like Sand Hill Road and starts looking like Wall Street.
The risks hiding inside a $41B pre-revenue valuation
A $41 billion mark on a company with no disclosed revenue and ~150 employees prices in near-flawless execution. The honest case for caution: physical AI is unproven at scale — training on robotics and experimental data is far slower and more expensive than scraping the web, and verification means physically building and testing things, not just reading output. The end markets (aerospace, pharma, semiconductors) are slow-moving, heavily regulated, and notoriously hard for outsiders to penetrate.
There's also concentration risk: $18B riding on a single thesis with one operating founder who has many other commitments. None of this means the bet is wrong — Bezos has a track record of being early and right on infrastructure-heavy ideas everyone else thought were too capital-intensive. But at this valuation, the margin for error is thin, and the burn against ~150 people implies the real spend is on compute and labs, not headcount. Watch whether the next milestone is a shipped product or just a bigger round.
$18 billion, ~150 people, zero shipped products, and the most accomplished operator of his generation back in the building.
Prometheus is the purest bet yet that the next AI trillion is in atoms, not text.
Track frontier-AI valuations and the megaround pipeline on the AI Valuations and Unicorn Tracker dashboards at Value Add VC. Originally published in the Trace Cohen newsletter.