Bezos AI Lab is valued at nearly 38 billion, raising funds to move into the physical artificial intelligence market

AI lab under Jeff Bezos, “Project Prometheus,” is nearing the completion of a new round of fundraising totaling $10 billion, with institutional investors such as JPMorgan and BlackRock participating. After this fundraising round is completed, the company’s valuation is expected to reach about $38 billion. Project Prometheus has already completed a $6.2 billion seed round, recruiting more than 100 employees from top AI labs such as OpenAI.

Embodied AI and LLMs: distinctly different technical paths

The core positioning of Project Prometheus is to build a new kind of AI system that can understand physical laws and interact with the real world—especially with a focus on manufacturing and industrial processes—fundamentally different from companies such as OpenAI and Anthropic that concentrate on large language models (LLMs).

Use cases for these systems include operating factory machinery, optimizing supply chains, and automating aerospace and semiconductor production processes. Its AI can not only generate text or images, but also directly intervene in how the physical world operates.

Data moat: the biggest competitive barrier for embodied AI

The biggest challenge facing embodied AI is the barrier to obtaining data. LLMs can be trained using massive amounts of text and images scraped from the internet, while embodied AI needs interaction data from the real world—sensor readings, manufacturing processes, haptic feedback, failure data in chaotic environments, and the like. Such data is typically proprietary and very costly to collect. Tesla is a typical example of this data advantage: roughly 5-6 million electric vehicles equipped with fully automated driving hardware accumulate more than 50 billion real-world driving miles each year, allowing it to maintain a sustained lead in autonomous driving capabilities.

Business layout: a holding-company strategy and a grand vision of $100 billion

To address the problem of obtaining embodied data, Project Prometheus has adopted a unique holding-company strategy. Bezos and Bajaj are raising hundreds of billions of dollars for a holding company positioned as a “tool for industrial transformation.” The funds will mainly be used to acquire companies in engineering, construction, and design, and to obtain real-world data through these investments to train its AI systems. According to a report by The New York Times, Bezos is also holding early discussions with investors in the Middle East and Southeast Asia about raising as much as $100 billion.

Frequently asked questions

What is embodied artificial intelligence, and how is it fundamentally different from LLMs like ChatGPT?

LLMs primarily process digital data such as text and images, producing outputs mainly as text or images. The goal of embodied AI is to understand physical laws and interact with real environments—operating factory machinery, perceiving three-dimensional space, and making real-time decisions in complex industrial settings. Its training data includes physical-world data such as sensor readings and mechanical motion trajectories. The technical path is fundamentally different from that of LLMs.

Why did Bezos choose to bet on embodied AI right now?

Generative AI has become relatively saturated at the software layer, while AI penetration in the physical world remains very low. The addressable markets in areas such as industrial manufacturing, aerospace, and semiconductors are enormous. Combined with Bezos’s deep experience in supply chains and industrial infrastructure accumulated at Amazon, it gives him a significant innate advantage in the next main battleground of the AI race.

What are the main competitive challenges facing Project Prometheus?

The biggest challenge is the barrier to obtaining embodied data. Unlike LLMs that can get vast training data from the internet, embodied AI requires data that is expensive and proprietary. Tesla has already established a significant first-mover advantage in autonomous-driving data. New startups such as Periodic Labs are also entering the same track. However, Bezos’s scale of capital and his experience with Amazon’s industrial infrastructure are core competitive advantages that are difficult to replicate quickly.