Jeff Bezos and Vik Bajaj Start New Company to Speed Up Product Design With AI
Jeff Bezos and Vik Bajaj Start New Company to Speed Up Product Design With AI
Jeff Bezos and Vik Bajaj have co-founded a company called Prometheus that builds AI tools to help engineers design and manufacture physical products more quickly, according to CNBC. They announced the venture on June 11, 2026, and both are serving as co-CEOs.
Prometheus focuses on what the founders call "the physical economy" — the world of objects and hardware, rather than software and digital files. The company's AI tools are aimed at engineering work that leads to real products: how mechanical parts fit together, how things get manufactured, and especially how computer chips are designed.
The emphasis on chip design is significant. Creating semiconductors ranks among the most difficult, expensive, and time-consuming types of engineering work that exists. Moving a new chip design to manufacturing, called a "tape-out," takes months. The physical templates needed to fabricate chips can cost tens of millions of dollars. If AI can shorten that cycle, the savings ripple through the entire technology industry.
Bajaj previously co-founded a biotech company called Verily and worked as an investor. Bezos is known for building massive operations at Amazon — the cloud computing division, the shipping logistics network, and his space company Blue Origin all required managing extraordinarily complex, long-term physical infrastructure. Prometheus brings both founders' experience to bear on one question: can AI make the process of designing and building hardware faster.
In their CNBC interview, Bezos and Bajaj described Prometheus as infrastructure for the physical economy at large. Chips are one important example, not the only market they intend to address.
Why Now
AI has already transformed software work. Tools like GitHub Copilot help software engineers write code faster by predicting what comes next, much like autocomplete on a smartphone. Those tools work well because the feedback is instant — a program runs in seconds, and the engineer immediately sees if it works or breaks.
Designing physical things is slower. You cannot test a chip design in milliseconds the way you test code. Also, much of the data engineers use to design hardware is proprietary — each company keeps its own designs secret, making it hard to train AI models on a broad range of examples. Because of these constraints, AI adoption in hardware and manufacturing has lagged well behind software.
Several established companies and startups have tried to inject AI into hardware engineering over the past few years. Cadence and Synopsys, which sell design software that engineers already rely on, have added AI features to their existing tools. Smaller companies have explored AI for material science. But no company has emerged as a clear leader in this space. That opening is what Prometheus sees.
The Unknown Territory
What remains unclear is exactly how Prometheus will approach the problem. The company might build AI helpers that sit on top of the design tools engineers already use — similar to how Copilot works with existing code editors. Alternatively, they might rebuild the underlying design process from the ground up.
This distinction matters. The existing companies that sell design software have been entrenched in this market for decades. They have strong relationships with customers and a lot of intellectual property locked into their platforms. Displacing them would require either a dramatically better workflow or a way to work alongside them without triggering a competitive response.
The co-CEO structure is also worth noting. Two leaders sharing the top role can work well if they split responsibility cleanly — one handling external strategy and the other running the product and science. If both are trying to drive the same decisions, friction usually surfaces within about eighteen months.
Prometheus is very early. The company has not demonstrated a working product publicly, announced any customers, or disclosed funding details beyond what the founders themselves are putting in. What exists so far is a mission statement and two founders with enough resources and reputation to attract talented engineers from the chip design, semiconductor, and AI research worlds. In a field where progress depends on the quality of the people building the models, that recruiting advantage is real.
The idea is straightforward: let AI do for hardware design what it started doing for software. The execution is the hard part. Encoding the physical constraints that real manufacturing must obey, the tolerances that parts must meet, and the realities of supply chains into AI models that can learn and generalize — that is technically tractable but genuinely difficult.
