Workspaces
Log in Sign up

Jeff Bezos’s Prometheus Raises $12 Billion to Build the "Artificial General Engineer"

Updated

Sources

Jeff Bezos’s Prometheus Raises $12 Billion to Build the "Artificial General Engineer"

As the battle for large language models remains highly crowded, Amazon founder Jeff Bezos is making a massive, high-stakes play for the next frontier of physical artificial intelligence. His startup, Prometheus (officially founded in November 2025), closed a monumental $12 billion Series B funding round in June 2026, catapulting its valuation to $41 billion before ever shipping a commercial product.

Co-led by Bezos and Vikram Bajaj (former director of Google X/Verily), Prometheus is explicitly focused on building an "artificial general engineer" (AGE). Rather than training chatbots on internet text, Prometheus aims to build AI-driven world models that encode physical laws and materials science to accelerate industrial design, manufacturing supply chains, and physical workflows.

To power this vision, Prometheus has aggressively poached elite infrastructure and AI talent. Key hires include:

  • Kyle Kosic, an xAI co-founder and former OpenAI infrastructure lead who oversaw the deployment of xAI's massive Colossus supercluster.
  • Sherjil Ozair and William Guss (founders of General Agents, acquired by Prometheus).
  • Drew Jaegle (former Google DeepMind researcher).
  • Cristian Bodnar (former Microsoft researcher and co-lead of the Aurora Earth System Forecasting Model).

"This is an age-old dream, the idea that you might build a set of tools that can actually do engineering." — Jeff Bezos, CNBC / Observer

"The ‘artificial general engineer’ framing is ambitious, but the underlying bet that physical world data creates a moat that internet-trained models cannot replicate is the most strategically sound position123 I have seen staked out in this funding cycle..." — Shashi Bellamkonda, Info-Tech Research Group / Observer

Prometheus joins a tier of ultra-high-valuation, early-stage AI ventures founded by prominent industry figures—such as Mira Murati's Thinking Machines Lab ($12B) and Yann LeCun's AMI Labs ($3.5B)—that are bypassing the current LLM landscape to focus entirely on agentic reasoning and physical world modeling.

  1. An instance of Industrial world data is the ultimate moat against internet-trained AI models. — Prometheus is bypassing the crowded internet-trained chatbot race to invest billions in world models encoding physical and material engineering data. ↩︎

  2. An instance of Industrial world data is the ultimate moat against internet-trained AI models. — The text directly states that physical world data, materials science, and physical physics laws represent a major protective moat that internet-trained text models cannot match. ↩︎

  3. An instance of Dual-use AI systems require physical testing environments to bridge simulation gaps. — It establishes that physical engineering data creates an irreplaceable, real-world competitive boundary that web-scraped data cannot replicate. ↩︎

Part of

This finding is an example of a pattern recurring across your work:

Revision history

  • Write a new finding note on Jeff Bezos's physical AI startup Prometheus, its $12B funding round, and the poaching of xAI co-founder Kyle Kosic from OpenAI.
    · by the agent
  • Write a new finding note on Jeff Bezos's physical AI startup Prometheus, its $12B funding round, and the poaching of xAI co-founder Kyle Kosic from OpenAI.
    · by the agent
  • Write a new finding note on Jeff Bezos's physical AI startup Prometheus, its $12B funding round, and the poaching of xAI co-founder Kyle Kosic from OpenAI.
    · by the agent