Project Prometheus (Bezos)
Last reviewed
Jun 2, 2026
Sources
9 citations
Review status
Source-backed
Revision
v1 · 1,463 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Last reviewed
Jun 2, 2026
Sources
9 citations
Review status
Source-backed
Revision
v1 · 1,463 words
Add missing citations, update stale details, or suggest a clearer explanation.
Project Prometheus is an artificial intelligence startup co-founded by Jeff Bezos, who serves as its co-chief executive alongside the scientist Vik Bajaj. The company was first reported by The New York Times on November 17, 2025, which described a venture that had launched with roughly $6.2 billion in funding and a focus on "physical AI" for engineering and manufacturing across sectors such as computers, aerospace, and automobiles [1][2]. The venture marks Bezos's most significant operational role at a company since he stepped down as chief executive of Amazon in 2021 [2][3].
Project Prometheus operates with an unusually low public profile for a company of its size, and much of what is known comes from press reporting that cites anonymous sources rather than from the company itself. Care should therefore be taken to attribute specific figures and claims to the reporting that produced them.
Disambiguation: This article concerns the Bezos-backed AI startup. It is unrelated to Prometheus, the multi-gigawatt data center cluster that Meta is building in New Albany, Ohio, which is a separate project with a coincidentally identical name [4].
Project Prometheus is positioned around the idea of applying artificial intelligence to physical engineering and manufacturing rather than to text or imagery alone. Reporting describes the company as pursuing "physical AI," a category in which models learn from real-world experimentation and sensor data instead of relying solely on digital corpora [1][2]. This places it in the broader area sometimes called physical AI or embodied AI, and observers have compared its likely research direction to that of other physics- and science-oriented labs such as Periodic Labs [1].
At launch the company was described as one of the best-capitalized early-stage AI ventures in the world, owing to the roughly $6.2 billion it had reportedly raised [1][3]. It is based in San Francisco and has been reported to maintain additional offices in London and Zurich [5].
Project Prometheus is led by two co-founders who share the chief executive role.
| Person | Role | Background |
|---|---|---|
| Jeff Bezos | Co-founder, co-CEO | Founder and former CEO of Amazon; founder of the space company Blue Origin. First operational company role since leaving the Amazon CEO position in 2021 [2][3]. |
| Vik Bajaj | Co-founder, co-CEO | Physicist and chemist; former director at Google's X moonshot lab; co-founder of the Alphabet health-sciences company Verily; most recently co-founder and CEO of the AI incubator Foresite Labs, which he left to start Prometheus [1][2][6]. |
Bajaj's earlier work at Google's X has been linked in some reporting to early efforts that became the self-driving car program Waymo, although the precise extent of that involvement is described differently across accounts and is best treated with caution [2]. His most consistently reported credentials are his physics and chemistry training, his time at X, his co-founding of Verily, and his leadership of Foresite Labs [1][6].
For Bezos, the venture represents a return to a hands-on executive seat. The New York Times and subsequent coverage framed his participation as notable precisely because he had largely stepped back from day-to-day company management after handing the Amazon chief executive role to Andy Jassy in 2021 [2][3].
The figure most often associated with Project Prometheus is approximately $6.2 billion raised at or around its launch. This number originated with The New York Times report of November 17, 2025, which attributed it to people familiar with the matter; Fortune characterized the funding as including "some from Bezos himself," without specifying the precise share contributed by Bezos or the identities of other backers [1][2][3].
Later reporting pointed to a substantially larger follow-on raise. In April 2026, the Financial Times reported, and Bloomberg relayed, that Project Prometheus was nearing a roughly $10 billion funding round that would value the company at about $38 billion, with institutions including JPMorgan and BlackRock named among the investors anchoring the syndicate [7][8].
| Round | Reported amount | Reported timing | Notes |
|---|---|---|---|
| Initial funding | ~$6.2 billion | November 2025 | Reported by The New York Times; said to include some funding from Bezos [1][2] |
| Follow-on round | ~$10 billion | April 2026 (in progress) | ~$38 billion post-money valuation; JPMorgan and BlackRock reported among investors [7][8] |
Because these amounts come from press accounts citing unnamed sources rather than from official company disclosures, they should be read as reported figures rather than confirmed financials.
Project Prometheus is reported to be building AI products for engineering and manufacturing, with named target sectors including computers, aerospace, and automobiles [1][2][3]. The unifying theme across reporting is "physical AI": systems intended to understand and act in the physical world, trained on multimodal data gathered from real-world interactions rather than on text alone [5].
The company has signaled this orientation in the limited public footprint it maintains, describing itself in brief terms as building "AI for the physical economy" [5]. As part of assembling its capabilities, Project Prometheus acquired General Agents, a startup founded by former Google DeepMind researcher Sherjil Ozair that developed video-language-action (VLA) models, an approach that interprets visual inputs and acts on natural-language instructions and that is directly relevant to robotics and manufacturing [5]. Some coverage has connected the ambition to the notion of an "artificial general engineer," an AI system meant to act as a general-purpose engineering partner to human designers, though this framing comes from secondary summaries rather than a detailed company statement [9].
The work draws on talent recruited from established AI organizations. By late 2025, reporting described a team approaching 100 people, with researchers drawn from firms including OpenAI, Google DeepMind, and Meta's AI efforts; later accounts placed the headcount above 120 [1][6][5].
Project Prometheus has been characterized as secretive, and as of the reporting available it had no conventional public website and was operating largely in stealth [5]. The clearest, most consistently corroborated facts are the dual leadership of Bezos and Bajaj, the rough $6.2 billion launch funding, the physical-AI engineering-and-manufacturing focus, and the November 2025 timing of the first report [1][2][3].
Several details remain comparatively thin or single-sourced and warrant caution. These include the precise breakdown of the initial funding and the full investor list, the exact scope of products under development, the company's specific technical roadmap, and the finer points of Bajaj's earlier project history. Where this article gives such details, it attributes them to the reporting that produced them rather than presenting them as settled fact.
The venture drew attention chiefly because of who is behind it and how it is structured. Bezos taking a co-chief-executive title represents an unusual re-entry into operational leadership by one of the most prominent figures in technology, and it did so in service of physical-world AI rather than the consumer software or cloud businesses with which he is most associated [2][3]. The scale of the reported funding, both the initial round and the much larger follow-on, also placed Project Prometheus among the most heavily capitalized young AI companies, intensifying discussion about the amount of capital flowing into frontier AI ventures [3][7].
Project Prometheus emerged during a period of rapid capital formation around AI for the physical world, including robotics, world models, and laboratory automation, in which several well-funded labs sought to extend machine learning beyond language and images into manufacturing, science, and engineering. Its emphasis on learning from real-world experimentation situates it alongside other efforts exploring world models and robotics, even as its specific approach remained undisclosed. The episode also fed a wider conversation, which Bezos himself addressed in remarks describing an "industrial bubble" in AI investment, about whether the surge of funding into such companies was sustainable [9].