San Francisco: Jeff Bezos is close to finalizing a $10 billion funding round for Project Prometheus, his physical AI laboratory, at a $38 billion valuation. JPMorgan and BlackRock anchor the deal. The round had not formally closed at publication time, according to the Financial Times, which first reported the raise. If completed, total capital raised by the lab would cross $16 billion in just five months after launch.
Bezos co-founded Project Prometheus in November 2025 alongside Vikram Bajaj, who serves as co-CEO. Bajaj holds a PhD in physical chemistry from MIT and previously led early work at Google X on Wing and Waymo.
He also co-founded Alphabet’s Verily and AI drug discovery company Xaira Therapeutics. The lab has since grown to over 120 employees drawn from OpenAI, xAI, Meta, and DeepMind. It operates from San Francisco, with additional offices in London and Zurich.
The lab sits in a different category from the large language model companies that have dominated AI investment since 2022. Prometheus builds physical AI systems trained on real-world experimental data, robotics interactions, and engineering workflows rather than text and images alone.
Target applications span manufacturing, aerospace, robotics, drug discovery, and logistics. Physical AI requires proprietary data that cannot be scraped from the internet. That scarcity creates both a barrier to entry and a structural advantage for whoever accumulates it earliest.
The $38 billion valuation arrives in a week that underscores how dramatically AI deal scale has shifted. Amazon separately committed up to $25 billion in new investment to Anthropic, which pledged $100 billion in AWS spend over ten years in return. Prometheus, at $10 billion in a single round, is raising more than most AI companies have raised across their entire histories.
Behind the funding round sits an even larger ambition. Bezos is separately pitching a $100 billion holding company to sovereign investors in the Middle East and Southeast Asia. The plan would acquire industrial businesses affected by AI disruption, then feed their operational data directly back into Prometheus’s models. That data flywheel, if it closes, would give the lab a training advantage that no amount of compute spending alone can replicate.
Previous
