Jeff Bezos, founder of Amazon and one of the world’s richest individuals, is returning to hands-on leadership with a new AI venture.
The company, Project Prometheus, begins life with $6.2 billion in funding — an extraordinary sum for an early-stage startup — according to three people familiar with the plans who shared the details with The New York Times.
Bezos will help run the new firm as co-chief executive. He has not taken on a formal executive role since stepping down as Amazon’s chief executive in July 2021. Although he remains deeply involved with Blue Origin, his growing space venture, his official title there has remained “founder.” Project Prometheus marks the first time since leaving Amazon that he will once again manage an organization’s operations at the top.
A new player in an intensifying AI race
Prometheus, a Titan from Greek mythology, is known for his intelligence and foresight. But he is also known for stealing fire for humanity and his subsequent eternal punishment.
In the real world, Project Prometheus emerges at a moment of explosive competition in AI, as both established tech giants and well-funded startups compete to shape the future of automated reasoning, robotics, scientific discovery, and industrial design. Companies such as Google, Meta, Microsoft, OpenAI, and Anthropic are investing heavily into foundational AI systems, while smaller entrants are trying to carve out specialized roles.
The new venture’s enormous initial funding gives it an immediate presence in this ecosystem. Comparable AI startups, even those backed by prominent research talent, typically launch with far less. Earlier this year, Thinking Machines Lab — founded by former OpenAI employees — drew attention for raising $2 billion, which until now stood as one of the largest rounds in the space.
Project Prometheus has already hired nearly 100 employees, poaching talent from OpenAI, Google DeepMind, and Meta. This level of recruitment suggests that Bezos intends the company to be a major force from inception, not a side project or experimental lab.
AI for the physical world
The company’s focus aligns closely with Bezos’ long-standing interest in technology that operates in the physical world — especially in aerospace. Project Prometheus aims to build AI systems for engineering and manufacturing across several industries, including computers, automobiles, and spacecraft.
This emphasis is notable because it represents a shift away from the text-based AI models that dominate the current public conversation. Large language models, which power tools such as OpenAI’s ChatGPT, learn by analyzing vast volumes of text. While they can generate code, solve certain math problems, and simulate reasoning, they remain fundamentally limited by their training on digital content rather than real-world experimentation.
New companies like Periodic Labs — and now Project Prometheus — are taking a different path. They aim to create AI models that learn from physical experimentation, where robots run scientific tests at scale and feed the results back into machine-learning systems. The idea is that AI may not only accelerate discoveries in chemistry, physics, and materials science, but may eventually design and test new technologies in an automated cycle.
This approach could reduce costs in drug discovery, shorten timelines for developing new aerospace materials, or significantly increase the speed of designing complex components for spacecraft and advanced robotics. If successful, it could mark a major shift in how scientific research and industrial engineering are conducted.
Vik Bajaj joins Bezos at the helm
Bezos will share leadership with Vik Bajaj, a physicist and chemist with a long résumé in Silicon Valley’s research ecosystem. Bajaj previously worked at Google X — often called “The Moonshot Factory” — where he collaborated with Google co-founder Sergey Brin on ambitious, far-future projects such as Wing’s drone delivery system and the self-driving car initiative that later became Waymo.
Bajaj also co-founded Verily, Alphabet’s life sciences lab, and later became chief executive of Foresite Labs, an incubator for AI-driven data science startups. His background positions him as a bridge between AI research, scientific experimentation, and commercialization — precisely the areas that Project Prometheus aims to unite.
Implications for the AI landscape
The arrival of a Bezos-backed enterprise with billions of dollars in capital carries significant implications for the AI sector.
Large players such as Google DeepMind and Meta have been investing heavily in scientific AI research, including DeepMind’s AlphaFold. Bezos’ entry, with substantial resources and hiring power, forces incumbents to accelerate their work or risk losing top scientific talent.
Companies such as OpenAI and Meta have claimed their language models are approaching breakthroughs in mathematics and theoretical physics. Yet specialized systems designed to learn from physical experiments could prove more effective in domains requiring real-world validation. Project Prometheus could push the field toward hybrid AI models that combine digital reasoning with experimental data.
If Project Prometheus succeeds at automating engineering design and manufacturing processes, industries that typically involve long, expensive prototyping phases — automotive, computer hardware, and spaceflight — could see transformative reductions in cost and development time.
The company dovetails with his interests at Blue Origin. Advanced AI could help design new rocket engines, optimize spacecraft manufacturing, or test materials for extraterrestrial environments. This move positions Bezos even more prominently in the evolving privatized space race.
At Baidu World 2025 event, the company unveiled ERNIE 5.0, a new multimodal model that can simultaneously understand and generate text, images, audio, and video.