AI Is Designing Jet Engines — Bezos's $12B Prometheus and the Dawn of Physical AI Wars

Chatbots write text. Coding AIs write code. But who designs jet engines?

Jeff Bezos put $12B on that question. techcrunch

What is physical AI, anyway? — How's it different from chatbots?

While ChatGPT generates text and Cursor writes code, the most complex things humanity has ever built — jet engines, semiconductor chips, drug molecules — have remained untouched by AI.

There's a reason. "Digital AI" is ultimately pattern matching. It learns from billions of texts to predict the next word. But designing a jet engine is different. Aerodynamics, thermodynamics, materials science, and manufacturing processes — dozens of physical laws must interact simultaneously in ways text patterns can't capture. You need to understand the causal relationships of the physical world.

Bezos's startup Prometheus calls this an "Artificial General Engineer (AGE)." One letter different from AGI (Artificial General Intelligence), but a completely different goal. Not an AI that thinks like a human — an AI that designs like an engineer. Physical Intelligence (π)'s founders call it the beginning of "a Cambrian explosion of robotics applications." techcrunch +1

$41B / 150 employees — what this math is saying

Prometheus currently has 150 employees, with offices in San Francisco, London, and Zurich. And a valuation of $41 billion. techcrunch

That's $270 million per employee. Here's the thing — in physical AI, compute costs far more than people.

To automate jet engine design with AI, the AI needs to simulate tens of millions of design combinations. Each simulation has to compute real physical laws — which requires enormous computing power. That's why a substantial portion of the funding reportedly goes to compute. They're running a team of thousands of GPUs, not 150 people. techcrunch

Goldman Sachs, JPMorgan Chase, and BlackRock being on the investor list is notable. These firms don't typically back AI startups. Their bet signals that Prometheus's target isn't just another AI tool — it's a business that could reshape the design infrastructure of aerospace, pharma, and manufacturing industries worth hundreds of trillions. techcrunch

Co-founder Vik Bajaj was previously the co-founder of Verily, Google's life sciences division. That background explains why drug compound design made it onto the first target list. Biological molecules are also a domain governed by physical laws. techcrunch

So does Bezos think "mass unemployment" is coming?

If AI automates engineering design, doesn't that mean engineering jobs disappear? Bezos's answer is surprising.

"Significant productivity in the economy is going to raise the standard of living."

— Jeff Bezos, at the Prometheus announcementtechcrunch

His logic: AI boosts productivity → real incomes rise → more spending creates more demand for goods and services → that demand requires more people. He predicts a world where dual-income households become single-income, and overtime work disappears — not "mass unemployment," but an era of "labor scarcity."

This doesn't sound empty for a reason. Amazon currently employs over 1.5 million people worldwide, while aggressively pushing warehouse automation full-speed. The most aggressive automation company is simultaneously one of the world's largest employers — Bezos has been watching this paradox play out for over 20 years, firsthand. techcrunch

There's healthy skepticism too. Prometheus hasn't released any specific product or public demo yet. For the "Artificial General Engineer" concept to materialize, it also needs to clear regulatory approval from aviation (FAA) and pharma (FDA). The $41B valuation is among the most aggressive ever for an AI startup. techcrunch

How to prepare for the physical AI era

Whether Prometheus succeeds or fails, "physical AI" as a category is already real. Here's what you can do now.

1. Map your industry's physical bottlenecks
   If you're in manufacturing, logistics, R&D, or engineering, list the most repetitive and time-consuming design/verification processes. Those bottlenecks are what physical AI will eat first.
2. Get familiar with simulation and digital twin tools now
   Physical AI runs on simulation. Starting with NVIDIA Omniverse, Ansys, or Siemens NX today lowers the entry barrier when AI tools become practical.
3. Protect your domain expertise
   Physical AI still needs engineers who understand the physical world to validate and steer it. Domain knowledge becomes rarer and more valuable in the AI era.
4. Start structuring your design data now
   For AI to learn physical systems, it needs structured data. Systematically document design parameters, test results, and failure cases — this becomes your company's AI training data.
5. Experiment with π0 open-source models today
   Prometheus isn't public yet, but Physical Intelligence's π0 model is available as open source. NVIDIA is already expanding manufacturing partnerships. The channels to experience physical AI today are open. pi.website +1

Want to go deeper?

Jeff Bezos's Prometheus raises $12B for physical world AI Full announcement of Prometheus's Series B, founders, investors, and technical goals. techcrunch.com

Physical Intelligence (π) — Foundation Models Blog π0.7 release and open-source model details, Vision-Language-Action technology explained. pi.website

NVIDIA Physical AI for Manufacturing AI digital twin deployments with Foxconn, BMW, and Kawasaki — real-world manufacturing applications. nvidia.com

CB Insights — Physical AI and Manufacturing Robotics Market map of 115+ physical AI and manufacturing robotics startups and technology trends. cbinsights.com