Jeff Bezos is lifting the curtain: His AI startup Prometheus officially emerged from stealth mode on Thursday – with numbers that stand out even amid the current wave of AI investment. The company has raised a total of $12 billion, at a valuation of $41 billion. That instantly makes Prometheus one of the most highly valued AI startups in the world – before it has presented a single product.
Alongside Bezos himself, who is contributing a substantial share of the capital, a roster of heavyweight investors is on board: JPMorgan, BlackRock and Goldman Sachs, as well as venture capital firms DST Global and Arch Venture Partners. The startup had already secured an initial $6.2 billion in funding in November 2025.
Bezos’ First CEO Role Since Amazon
For Bezos, Prometheus is more than just another investment vehicle. For the first time since stepping down as Amazon CEO in July 2021, the tech billionaire is taking on an operational CEO role again – together with co-CEO Vik Bajaj, a professor at Stanford University School of Medicine and co-founder of Alphabet’s research lab Verily. Bezos recently said the project excites him so much that he is investing significant time and energy in it. AI now dominates his entire entrepreneurial life: whether at Amazon, where he remains Executive Chairman, at his space company Blue Origin, or at Prometheus – the topic is front and center everywhere, Bezos told CNBC.
With this move, the Amazon founder joins a remarkable trend: the founding generation of Big Tech is personally stepping back into the ring. Google co-founder Sergey Brin has returned to Mountain View, where he works hands-on on the Gemini models; Larry Page is reportedly building Dynatomics, an AI startup for the manufacturing industry; and Meta CEO Mark Zuckerberg has made AI development a top personal priority with his Superintelligence Labs – including personal poaching campaigns for top researchers. The AI race is no longer just a competition between corporations, but between their founders.
“Artificial General Engineer”: AI for the Physical Economy
But what is Prometheus actually about? The company wants to do for engineering and manufacturing what large language models have done for text. Bezos speaks of an “artificial general engineer” – an AI system designed to make it dramatically easier for engineers to design physical objects, from skyscrapers to smartphones to jet engines. Instead of distilling text from the internet, Prometheus trains its models on data from the physical world: established laws of physics, its own testing data, and information from collaborations with manufacturing companies it declined to name.
Bezos’ math goes like this: If a project that today keeps 100 engineers busy for ten years can in the future be handled by ten engineers in a single year, far more things will simply get built. He dismisses concerns about job losses – on the contrary, he predicts a labor shortage, because AI will make building new things cheaper and easier. For workers, AI is like a “bulldozer instead of a shovel,” he says.
Prometheus explicitly has nothing to do with robotics, Bezos emphasizes. The company employs around 150 people across offices in San Francisco, London and Zurich, operates its own large GPU cluster, and is actively recruiting talent from OpenAI, Google DeepMind and Nvidia. The co-CEOs are still largely silent about the second part of their vision: According to reports, the plan is to build a Berkshire Hathaway-style portfolio of companies that uses Prometheus’ AI models to acquire existing industrial giants or build new competitors from scratch – a direct push into the physical economy, analogous to the way LLMs are currently upending knowledge work.
Industrial AI: Mistral Is Betting on the Field Too – With Technology from Linz
Prometheus is not alone in pursuing this direction. In Europe, too, industrial AI has gained significant momentum recently: French AI champion Mistral AI just acquired the Linz-based deep-tech startup Emmi AI – for well over €200 million, according to industry insiders. Founded only in late 2024 by Johannes Brandstetter, Dennis Just and Miks Mikelsons, the company develops so-called Large Engineering Models: physics-based AI models that simulate processes such as airflow, thermodynamics and material deformation in real time, with the aim of drastically shortening development cycles in industries like automotive, energy, semiconductors, and aerospace.
With the acquisition, Linz becomes an official Mistral location alongside Paris, London and San Francisco; Brandstetter will lead the AI for Science division as Vice President. The parallel to Prometheus is obvious: On both sides of the Atlantic, AI companies are betting that the next big leap won’t come from chatbots, but from models that understand the physical world – and help build it faster, cheaper and more precisely.
In the US, Periodic Labs, founded by researchers from leading AI firms, is pursuing a similar approach, aiming to use AI to accelerate discoveries in physics and chemistry. The race for “AI for the physical economy,” as Prometheus itself describes its mission, has officially begun.
