AMD to acquire Fei-Fei Li's World Labs for $8.2 billion in spatial intelligence push

AMD acquires World Labs for $8.2 billion

AMD has agreed to acquire World Labs, a startup building deep learning models that understand physical reality, in a deal valued at $8.2 billion, according to TechCrunch. World Labs' own announcement confirms a definitive agreement, with the transaction expected to close by the end of 2026, subject to regulatory approvals and other customary conditions.

World Labs founder Dr. Fei-Fei Li will join AMD as executive vice president and chief scientist, working directly with CEO Dr. Lisa Su. Co-founders Justin Johnson and Ben Mildenhall will continue leading the World Labs team as it becomes part of what the companies describe as a frontier research organization within AMD.

What World Labs builds

Li, a Stanford computer science professor considered a pioneer of computer vision for creating the ImageNet database and the AI competitions it inspired, founded World Labs in 2024. The company's core argument, as TechCrunch notes, is that true general intelligence requires models grounded in physics and able to reason about data beyond text.

"World model" remains a loosely defined term, spanning everything from language models that understand visual inputs to systems capable of generating and sustaining high-fidelity simulations of reality. World Labs' first product, Marble, is pitched both as a tool for creating entertainment experiences and as a way to build simulated environments for training robots.

How the deal came together

The acquisition grew out of a technical partnership formed last year, which began with model training and inference optimization on AMD GPUs. According to World Labs' announcement, working closely together convinced both sides that combining software, hardware, foundation models and applications was a natural fit. Li had also appeared as a guest at AMD's CES presentation earlier this year, TechCrunch reports.

World Labs framed the sale as a matter of scale, saying in a statement that AI development requires close collaboration across model research, systems and compute, and that accelerating its progress meant scaling its efforts, widening its reach and getting closer to the hardware. AMD, in turn, says that understanding frontier workloads like those created at World Labs will shape its chipmaking roadmap. The stated shared ambition is an end-to-end open AI ecosystem spanning hardware, software, platforms and widely accessible open models.

Why it matters

The move is AMD's most aggressive step yet toward competing with Nvidia's ecosystem strategy. Nvidia already publishes a suite of open-weight world models under the Cosmos brand, while AMD has so far only offered text- and video-based models to the public. Buying a leading world model developer hands AMD both the models and the researchers needed to close that gap.

World models are also increasingly seen as critical infrastructure for robotics. With useful real-world training data for general-purpose robots in short supply, synthetic data generated by world models is expected to be central to the ambitions of companies building autonomous vehicles, industrial robots and humanoids, including Tesla and Figure. A chipmaker that owns both the silicon and the model stack for that work would be unusually well positioned.

The deal also reflects a broader consolidation of frontier AI research around companies that control compute. Rather than partnering loosely, AMD has bought the lab outright and installed one of the field's most prominent researchers at the top of its research organization. It is a bet that the next phase of AI — systems that perceive and act in the physical world — will be shaped as much by hardware as by the models themselves.