Jeff Bezos's $12B 'Artificial General Engineer'

TL;DR — In June 2026, Jeff Bezos's stealth startup Prometheus (often called "Project Prometheus") raised $12 billion at a roughly $41 billion valuation to build what it calls an "artificial general engineer" — AI meant to design and manufacture complex physical things like jet engines and medicines. Bezos co-founded it with scientist Vik Bajaj and serves as co-CEO, his first operating role since leaving Amazon. Separately, he anchored a $500 million round for brain-inspired AI startup Flourish with a personal check of around $100 million. Together the moves mark Bezos's most aggressive push yet into AI's physical and energy frontiers.

---

What Prometheus actually is

Prometheus is an AI company that Jeff Bezos co-founded in late 2025 with Vik Bajaj, a chemist and physicist who previously co-founded Alphabet's life-sciences arm Verily. The two serve as co-CEOs — notable because it is reportedly Bezos's first hands-on operating role since he stepped down as Amazon's chief executive in 2021.

The company stayed largely under wraps until June 11, 2026, when it confirmed a second funding round of $12 billion at a valuation of roughly $41 billion. That follows an initial raise of about $6.2 billion late last year, according to CNBC. The new investors reportedly include JPMorgan Chase, Goldman Sachs and BlackRock, alongside Bezos himself.

The product pitch is unusually concrete for an AI moonshot. Prometheus is building what it describes as an "artificial general engineer" — software meant to automate the design and manufacturing of complex physical systems. Reporting from TechCrunch and Axios frames the target industries broadly: aerospace, semiconductors, energy and pharmaceuticals, anything where a physical object has to be designed, simulated, tested and built.

The "dream-build loop"

The core idea Bezos has described publicly is compressing the engineering cycle. Speaking after the raise, he said the company is building tools to make the "dream-build loop" — the path from concept to a working physical product — roughly ten times faster, per CNBC and GeekWire reporting. He also pushed back on the "stealth startup" label, telling reporters the company is "not being secretive."

In practice, that means AI systems that can reason about the physical world: materials, tolerances, thermodynamics, manufacturing constraints. This is a different problem from the text-and-code generation that dominates today's large language models. It belongs to a category many in the industry now call "physical AI" — models trained to understand and act on real-world systems rather than just language.

To accelerate that, Prometheus has already made at least one acquisition. It bought General Agents, a startup co-founded by former Google DeepMind researcher Sherjil Ozair that builds a video-language-action (VLA) model — software that interprets visual input and acts on natural-language commands, according to multiple reports. The company employs roughly 150 people across San Francisco, London and Zurich.

Project Prometheus: the key facts

The other bet: Flourish

A week earlier, on June 4, 2026, Bezos surfaced in a very different AI deal. Flourish, a New York startup, closed a $500 million round at a roughly $2.5 billion valuation. Several outlets, including TechTimes and The Next Web, reported that Bezos anchored the round — but with a personal check of around $100 million, not the full $500 million. He reportedly committed about $50 million initially, then nearly doubled it after other backers piled in. Other investors include Lux Capital, GV (Alphabet's venture arm) and Catalio Capital.

This is worth flagging because the figure is easy to misstate: the $500 million is the total round, while Bezos's personal stake is a fraction of that. The distinction matters for understanding how concentrated his bet really is.

Flourish is chasing a problem that has become AI's quiet crisis: power. The company, co-founded by Thomas Reardon — who built Internet Explorer at Microsoft and later founded the brain-computer-interface firm CTRL-labs, which Meta acquired in 2019 — and former Amazon executive Rob Williams, is pursuing "brain-inspired" AI. Its flagship effort, Cortex AI, draws on connectomics, the painstaking mapping of real neurons and their connections, to hunt for what the founders call the brain's "core algorithm."

The payoff they are aiming for is efficiency. The human brain runs on roughly 20 watts; Flourish says it is targeting AI architectures that operate in the 20-to-50-watt range, against the far heavier draw of today's GPU clusters, where a single server-grade chip can consume many times that. Crucially, Flourish says it wants to get there through algorithms and architecture rather than custom silicon — a different path from chipmakers like Cerebras or Groq. The company has no commercial product yet, so the valuation rests heavily on its founding team and the urgency of the energy problem.

How the two bets fit together

On the surface, an "artificial general engineer" and a 20-watt AI brain look unrelated. But they target the two walls the current AI boom keeps running into. Prometheus attacks the application wall — getting AI to do hard, valuable work in the physical economy rather than generating text. Flourish attacks the energy wall — making AI cheap enough to run at the scale those applications would demand.

That pairing tells you something about how Bezos is thinking. Both companies sit outside the crowded chatbot race and aim instead at infrastructure-level problems with enormous potential payoffs. It is a pattern we have seen across the wave of next billion-dollar AI companies to watch: the biggest checks are increasingly flowing to teams tackling the physical and energy constraints of AI, not just the software layer.

It also reflects a broader shift from AI that talks to AI that does. The same logic driving interest in AI agents and workflow automation — systems that take actions, not just answer questions — extends naturally to Prometheus's vision of agents that design and build real machines. If you are tracking where this capital is heading next, our rundown of emerging AI startups to watch covers more of the teams chasing these frontier problems.

What to watch — and what's still unproven

A few caveats are worth keeping front of mind. First, Prometheus has shown almost no public product. The "10x faster" claim is a stated ambition, not a demonstrated result, and an "artificial general engineer" remains a marketing term rather than a defined technical milestone. Second, the valuations are extraordinary for pre-revenue companies — $41 billion for Prometheus and $2.5 billion for Flourish — which raises the usual questions about AI-era pricing. Third, much of the detail comes from a single news cycle in June 2026; some figures, including Flourish's valuation, vary slightly across outlets.

What is well corroborated: the existence of both companies, Bezos's leadership role at Prometheus, the $12 billion raise, and his personal backing of Flourish. The grander claims about redesigning how jet engines and medicines get built sit firmly in the promise column.

The bottom line

Jeff Bezos has placed two large, complementary bets on AI's hardest frontiers. Prometheus, with $12 billion and a $41 billion valuation, wants to turn AI into a general-purpose engineer for the physical world; Flourish, with Bezos's roughly $100 million inside a $500 million round, wants to make AI itself far more energy-efficient by copying the brain. Neither has a shipping product, and the headline ambitions remain unproven. But the direction is unmistakable: the richest person on the planet is wagering that the next phase of AI is not about better chatbots, but about machines that design and build — and that can run on a fraction of today's power.