Jeff Bezos's $41B bet on an AI engineer for the real world

Generating text and writing code was just the warm-up. Silicon Valley's deepest pockets are now rushing headfirst into the physical world.

Prometheus, a physical AI startup co-founded by Jeff Bezos and former Verily co-founder Vik Bajaj, just announced a massive $12 billion funding round at a staggering $41 billion valuation. The round was backed by Bezos himself alongside Wall Street titans JPMorgan Chase, Goldman Sachs, and BlackRock.

According to CNBC, this is actually the stealthy venture's second round. Prometheus quietly launched late last year with an eye-popping $6.2 billion initial raise. The company currently employs just 150 people across offices in San Francisco, London, and Zurich, making its per-employee valuation almost surreal.

Meet the artificial general engineer

Prometheus isn't building another chatbot. It is setting out to create an "artificial general engineer" — sophisticated software capable of automating the design and manufacturing of complex physical systems, ranging from commercial jet engines to novel drug compounds.

While the ambitious goal is to replace vast swathes of traditional engineering labor with AI, Bezos is surprisingly downplaying fears of an unemployment crisis. In an interview with CNBC, he argued that massive AI productivity gains will actually create "labor scarcity," meaning demand for human workers will outstrip supply.

His reasoning? A productivity boom will dramatically lift living standards, allowing two-earner households to comfortably drop down to one earner, while salaried employees can stop clocking overtime. It is a bold perspective coming from the executive chairman of Amazon, a company with over 1.5 million employees that has laid off tens of thousands over the past year while pushing its own automation.

For now, Prometheus is keeping what it has already built strictly under wraps. Bezos did note that a substantial portion of the new $12 billion windfall will be channeled directly into massive compute resources.

Why it matters

At $41 billion, Prometheus instantly becomes one of the most richly valued AI startups in history and marks the biggest bet yet on physical AI.

VCs and institutional investors are piling in because pure software is starting to look commoditized. A wrapper app or a software coding agent has thin moats against open-source models, but the messy physical world is different. Designing jet engines, navigating thermodynamics, and formulating chemical compounds create real-world data and testing moats that pure code cannot breach.

For engineers and manufacturing teams, there is nothing to test today since Prometheus remains in stealth. But down the road, this could completely upend legacy industrial software. Giants like Siemens and Dassault Systemes, which have dominated computer-aided engineering for decades, may soon face AI that designs parts from scratch rather than just assisting human draftsmen.

That said, we think the valuation raises huge questions. Handing a 150-person team a $41 billion price tag before revealing a single public benchmark leaves zero room for error. Simulating physical reality is notoriously capital-intensive, and if moving from synthetic simulations to working physical prototypes hits snags, burning billions on compute could take years to yield a viable return.