SAN FRANCISCO: Jeff Bezos is a CEO again. Prometheus, the industrial-AI startup he co-founded in late 2024, announced on June 11, 2026 that it has closed a $12 billion Series B round at a valuation of roughly $41 billion — one of the largest private financings of the year and Bezos’s first executive role since he stepped down from Amazon in July 2021. The round was confirmed in a CNBC interview with co-CEOs Bezos and Vik Bajaj and detailed by Axios, GeekWire, and Semafor.
Investors in the Series B include JPMorgan, BlackRock, Goldman Sachs, DST Global, and Arch Venture Partners. Bezos, who was the single largest backer of the company’s $6.2 billion Series A, told CNBC’s David Faber he participated in the new round as well. The valuation jumps from the roughly $38 billion mark reported in April, when Bloomberg first surfaced the company’s earlier financing.
What is Project Prometheus and what does it build?
Prometheus — the company has now dropped “Project” from its name — is building what Bezos describes as an “artificial general engineer.” The pitch is to do for physical-product development what large language models did for text: ingest data from the physical world and compress the full arc of design, performance modeling, prototyping, and manufacturing into an end-to-end AI workflow.
Bajaj, a Stanford School of Medicine professor and co-founder of Alphabet’s life-sciences arm Verily, framed the company’s core technical claim in concrete terms. “What has changed in the last few years is the ability to formulate even something as complicated as that, from design to manufacturing, as an end-to-end AI problem,” he said, using a jet engine — a project that can take teams of engineers a decade or more to design and ship — as the example.
The target list is deliberately broad: chips, smartphones, jet engines, skyscrapers, bridges. The company has been explicit that this is not a robotics or factory-automation play, although Prometheus tools could be used to design and build robots and factories. The focus is on the pre-production loop — making the design-build cycle “10 times faster or more,” in Bajaj’s words — rather than swapping out humans on the factory floor.
Why $41 billion for a company founded 18 months ago
Prometheus was incorporated at the end of 2024 and now has about 150 employees split across San Francisco, London, and Zurich. The headcount is small relative to the price tag, but the talent is concentrated. The company has recruited from OpenAI, Google DeepMind, and Nvidia, and is led by a CEO with the most successful operating track record in modern technology paired with a co-founder who has built large research organizations in both the public and private sector.
The funding scale also reflects the cost of the bet. Training models against physical-world data — sensor logs, materials properties, manufacturing telemetry, and simulation outputs — is compute-heavy and data-heavy in ways that text-trained LLMs are not. The size of the round suggests Prometheus is funding a multi-year buildout of proprietary data pipelines, simulation infrastructure, and partnerships with industrial operators, not just GPU rentals. It mirrors the pattern set by other megarounds where capital is being used to lock up scarce compute capacity, as covered in our piece on Anthropic’s Colossus 1 deal with SpaceX.
Bezos and Bajaj also pushed back on the “stealth startup” framing the press has applied to the company. “We’re not being secretive,” Bezos told CNBC, noting that the company has been hiring publicly and publishing research. The narrative the founders want is “industrial AI lab” rather than “Bezos’s secret moonshot.”
What does Prometheus mean for founders building outside text-only AI?
The Prometheus thesis lands at a useful moment for founders. The 2024–2025 cycle was dominated by chatbots, copilots, and AI wrappers on top of foundation models. Most of the venture capital chasing AI has gone toward software-only products. Prometheus is pointing at a different frontier: AI that designs and builds atoms, not bits.
That has three near-term implications for builders. First, vertical AI for hard-tech disciplines — semiconductors, materials science, aerospace, civil engineering, biotech instrumentation — is being validated as a category that can attract megaround capital, not just specialist seed checks. Second, the moat is shifting toward proprietary physical-world data. Founders sitting on sensor logs, manufacturing telemetry, simulation outputs, or test-bench data have an asset class that pure software AI companies cannot easily replicate. Third, the competitive set for industrial AI is now a 150-person company with $18 billion in committed capital and the most valuable founder brand in technology. Anyone building in this space should plan for the partnership-or-be-acquired conversation earlier than they otherwise would.
The signal extends beyond hard tech. When the smartest pools of late-stage capital — JPMorgan, BlackRock, Goldman Sachs — back a physical-AI thesis at $41 billion, they are voting that the next leg of AI value creation runs through manufacturing, engineering, and the built environment rather than another chat interface. Founders should expect more competition for industrial data, more LP pressure on funds to find their own “physical AI” thesis, and faster valuation compression for AI categories that look like commodity software.
What’s next for Prometheus
The company has not announced a product launch date or named flagship customers. Bezos and Bajaj signaled in their interviews that the next phase is partnerships with industrial operators willing to share data in exchange for early access to the design tools. Watch for joint pilots with semiconductor, aerospace, or heavy-industrial firms in the second half of 2026, and for Prometheus to publish or open-source pieces of its physical-world data infrastructure as a recruiting and validation play.
The other thing to watch is Bezos’s calendar. Taking a CEO role at a company this large is a binding signal: Prometheus is not a hobby fund or a side bet. The most influential founder in modern technology has put his time, his money, and his title behind the claim that the next decade of AI will be defined by the machines that design and build the physical world.
