Jeff Bezos is building a new artificial intelligence venture under the code name Project Prometheus. The effort is still quiet, but recent hiring moves show clear intent. The company has brought in Kyle Kosic, a co-founder of xAI, to lead key infrastructure work.
Kosic adds strong technical depth to the project. He helped build the Colossus supercomputer at xAI and also worked at OpenAI. His focus at Prometheus will center on large-scale systems that support advanced AI models. That includes compute clusters, data pipelines, and tools that allow models to run in real-world settings.
The startup is led by Jeff Bezos and former Google executive Vikram Bajaj. It is growing across San Francisco, London, and Zurich. These locations suggest a plan to tap into top research talent and global engineering hubs.
Kosic’s move also reflects a wider shift inside xAI. Reports say all 11 original co-founders have now left the company. This points to bigger changes within the team built by Elon Musk. At the same time, xAI continues to hire, which shows it is still in expansion mode despite leadership turnover.
The Prometheus Blueprint: Building AI for the Physical World
Project Prometheus stands apart from many AI startups. Most firms focus on chatbots or coding tools. Prometheus is taking a different path. It aims to build systems that can understand and act within physical environments. That means working with real-world processes like manufacturing, logistics, and engineering systems.
This approach relies on domain-specific data. Instead of training models on general text, the company wants data from industries such as aviation, architecture, and design. These data sets can help models learn how physical systems behave. The goal is to solve complex industrial problems, not just answer questions.
The company is also hiring engineers who can handle large-scale infrastructure. These roles are critical. AI systems that interact with the physical world need stable, high-performance systems behind them. They must process data in real time and operate with high reliability.
Bezos and Bajaj appear to have a long-term plan. Some observers compare the structure of Prometheus to Berkshire Hathaway. That means it may act as both an AI developer and an investment vehicle. The company could invest in businesses that provide useful data, then use that data to improve its models.
This model creates a feedback loop. Partner companies supply real-world data. Prometheus uses that data to train its systems. The improved systems then help those same industries operate more efficiently. Over time, this could shorten the pace of technology adoption. Changes that once took a decade may happen much faster.
To support this vision, the startup is seeking large amounts of capital. It is exploring partnerships with sovereign wealth funds in Singapore and Gulf countries. These investors can provide the scale needed for global infrastructure and long-term research.
Bringing AI from the Screen to the Physical World
The focus on physical systems may give Prometheus an edge. Many AI tools today work well in digital tasks but struggle in real-world settings. Factories, transport networks, and energy systems have strict constraints. They require precision, safety, and reliability. Building AI that can handle these conditions is a harder challenge, but it also offers greater impact.
Kosic’s experience fits this need. His work on high-performance systems at xAI and OpenAI gives him the background to lead this effort. He understands how to build the backbone that advanced AI requires.
Project Prometheus is still in its early stage. It has not released products or detailed plans. But its hiring choices and structure show a clear direction. It aims to connect AI with the physical world and reshape how industries use data.
If it succeeds, the impact could reach far beyond software. It could change how machines are designed, how supply chains run, and how complex systems are managed. That is a different kind of AI story one that moves from screens into the real world.
