AMD has agreed to buy World Labs, the artificial intelligence startup co-founded by Stanford professor Fei-Fei Li, for US$8.2 billion, according to a report by The Malaysian Reserve.

The deal ranks among the largest acquisitions AMD has pursued and signals how far the company is willing to spend to strengthen its position in AI beyond silicon. Fei-Fei Li, often described as one of the founding figures of modern deep learning through her earlier work on the ImageNet dataset, co-founded World Labs to build what the startup calls spatial intelligence — AI systems that understand and reason about three-dimensional environments rather than only text or images.

The $8.2 billion price tag puts World Labs in the same tier as the most expensive private AI acquisitions of the current cycle. For AMD, which has spent the past two years trying to close the gap with Nvidia in AI accelerators, the purchase targets a different layer of the stack: the software and modeling capabilities that determine whether customers can actually deploy AI systems on AMD hardware.

AMD's core AI business rests on its Instinct GPU family, which competes against Nvidia's dominant data-center accelerators. Hardware alone has not been enough. Nvidia's CUDA software ecosystem has locked in much of the AI development community for nearly two decades, and AMD has responded with its open ROCm software stack, acquisitions of AI software firms, and aggressive pricing on large MI-series deployments. Buying World Labs extends that strategy from tooling into frontier AI research talent.

Fei-Fei Li's involvement gives the deal weight beyond its balance-sheet figure. Her reputation in the research community — she directed Stanford's Stanford AI Lab and later co-led the Stanford Human-Centered AI Institute — makes World Labs a rare acquisition target that combines academic credibility with a commercial roadmap. The startup's focus on spatial intelligence also aligns with emerging demand for AI in robotics, autonomous systems, and embodied applications, areas where three-dimensional world modeling is a prerequisite rather than a convenience.

The reported deal value of US$8.2 billion would make World Labs one of the most expensive AI startup exits to date for a company founded only in 2024. The acquisition price reflects the premium that established chipmakers are paying for scarce AI talent and differentiated model capabilities, a market dynamic that has pushed valuations of young AI firms sharply higher across the sector.

For AMD, the commercial logic runs through its data-center franchise. The company has forecast billions of dollars in AI accelerator revenue and has secured design wins with major cloud and model providers, but it continues to trail Nvidia in both market share and software maturity. Owning a frontier AI lab headed by a researcher of Li's stature gives AMD an internal capability to shape workloads, benchmarks, and model architectures toward its own hardware — the same flywheel Nvidia built with its close ties to leading AI labs.

The transaction also fits a broader pattern of consolidation across the semiconductor industry, as chipmakers move up the stack through acquisition rather than organic development. AMD previously bought Xilinx for $35 billion in 2022 and Pensando for $1.9 billion the same year, deals that added adaptive computing and networking to its portfolio. World Labs would be its largest software-and-models acquisition to date, and its first aimed squarely at frontier AI research.

Completion of the deal will depend on standard regulatory review. Should it close, AMD gains not just a startup but a research agenda — spatial intelligence — that could define the next generation of AI applications in robotics and physical-world computing, and a competitive lever against Nvidia in the contest over where the next wave of AI models gets built.