AMD Acquires World Labs for $8.2 Billion to Rival Nvidia in the Frontier of World Models

This focus is increasingly viewed as the "North Star" for the next phase of AI development, particularly in the fields of robotics, autonomous systems, and advanced simulation. By acquiring this capability, AMD is positioning itself to compete directly with Nvidia’s Cosmos world models and its Omniverse platform. The strategic necessity of this move cannot be overstated; as Nvidia expands its presence in the robotics sector, AMD must provide a comparable software and model ecosystem to ensure its Instinct accelerators remain a viable alternative for developers building the next generation of physical AI.

The technology portfolio coming to AMD is already showing signs of maturity. World Labs conducted a public launch of its "Marble" world model in November 2025, which was quickly followed by the release of the World API in January 2026, offering cloud-based access to its spatial intelligence tools. More recently, the company unveiled "Atlas," described as an "omni world model." Technically categorized as a multimodal autoregressive diffusion transformer, Atlas is designed to serve as a foundational engine for 3D generation and physical simulation. Early access for Atlas was granted to a select group of enterprise customers in early September, and the technology is expected to be integrated into future iterations of the Marble platform under the AMD umbrella.

However, the acquisition is as much about human capital as it is about software. Leading the transition is Fei-Fei Li, who served as the CEO of World Labs and will now take on the role of Chief Scientist at AMD. Li is widely regarded as one of the most influential figures in the history of artificial intelligence. With a PhD in electrical and electronics engineering from Caltech and a background in physics from Princeton, she rose to international prominence as the director of the Stanford AI Lab. Her tenure as Chief Scientist of AI/ML at Google Cloud in 2017 and 2018 ended amid internal company friction regarding military contracts, after which she returned to Stanford to co-direct the Institute for Human-Centered AI. Her move to AMD represents a major coup for CEO Lisa Su, providing the company with a figurehead who commands immense respect across academia and the private sector.

The rest of the founding team brings equally impressive credentials. Justin Johnson, a former AI research scientist at Meta and an adjunct professor at the University of Michigan, has a long history of collaboration with Li dating back to their time at Stanford. Ben Mildenhall, a former research scientist at Google, is a specialist in computer vision and generative 3D models. Interestingly, the AMD announcement did not explicitly mention co-founder Christoph Lassner, a veteran of Amazon, Meta, and Epic Games, leading to speculation regarding his continued role within the newly integrated structure.

According to internal reports, Johnson and Mildenhall will continue to oversee the day-to-day operations of the World Labs unit within AMD. Li, meanwhile, will report directly to Lisa Su. This reporting structure suggests that Li’s influence will extend far beyond the narrow confines of world models for robotics. She is expected to play a pivotal role in shaping AMD’s broader AI roadmap, potentially steering the company toward the development of frontier-level generative models that could challenge the current market leaders.