At some point, most of us have admired the 50-year-old “godmother of AI” because of her pioneering, foundational work in computer vision and deep learning, as well as her advocacy for human-centered technology.
Fei-Fei Li spent more than a decade as a professor at Stanford before co-founding ImageNet. That’s the dataset often credited with igniting the modern AI revolution.
Her career has been defined by a simple belief that the next frontier of AI is not language but space — how machines perceive, model, and interact with the physical world.
“Real recognize real,” as the saying goes. On Sept. 28, AMD CEO Lisa Su made a bet that Li is right.
Advanced Micro Devices, Inc. (AMD) confirmed it has agreed to acquire World Labs, the spatial AI startup Li co-founded, in an all-stock transaction valued at approximately $8.2 billion.
Li will join AMD as executive vice president and chief scientist, reporting directly to Su. The deal is expected to close by the end of 2026, subject to regulatory approval.
Even before we go deeper, describing AMD’s performance as “incredible” would be an understatement. It is up 183.84% year to date, according to Yahoo Finance.
With a $1 trillion-plus market cap built on explosive data-center growth and agentic AI momentum, Su is now betting her company’s hard-earned gains on where AI goes next.
ALSO READ: Advanced Micro Devices Inc. Latest News
What World Labs actually does, and why the AMD acquisition is a big deal
The core question you’d ask as an investor in an $8.2 billion acquisition is: What exactly am I buying?
World Labs, headquartered in San Francisco, develops spatial intelligence models. Those are AI systems that generate, reconstruct, and simulate interactive 3D environments from text, image, and video inputs.
World Labs also develops tech for robotic learning and simulation. In simple terms, it is AI designed to understand the physical world, not just language.
More AMD:
Jim Cramer sends a blunt message to AMD stock investors
Bank of America sends wake-up call to AMD stock investors
Nvidia just sent a strong signal to AMD and Intel investors
The distinction is what we need to understand. Remember, Nvidia’s dominant position in AI infrastructure was built around language and image models that are computationally structured in ways that map naturally onto GPU architecture.
Now, the next wave — reasoning AI, robotics, autonomous systems, physical simulation — places very different demands on compute.
World Labs gives AMD a gem of a research team working directly at that frontier, helping shape future hardware roadmaps around workloads that have yet to reach commercial scale but could soon become mainstream.
“Building the compute platforms for the next generation of AI requires a deep understanding of how models are evolving,” Su said in AMD’s announcement. “Fei-Fei and the World Labs team bring exceptional research leadership and model expertise.”
The Nvidia challenge and why this is about software as much as chips
We all know that Nvidia’s most durable competitive advantage is CUDA, not hardware. CUDA is the software ecosystem that has become the default programming interface for AI researchers worldwide.
Millions of researchers and engineers have built careers, workflows, and products on top of CUDA. Switching away from it is painful, expensive, and slow.
In fact, AMD’s ROCm software ecosystem has been trying to close that gap for years with limited success.
So the World Labs acquisition represents a different strategy. Rather than trying to replicate CUDA’s ecosystem, AMD is vertically integrating a frontier AI research team that can design the next generation of workloads from scratch on AMD hardware. Excellent, right?
Related: Bank of America resets AMD price target after major milestone
When robotics simulation, physical AI, and spatial intelligence become major commercial compute categories — and every major technology investor is betting they will — they will not necessarily have the same CUDA dependency current LLM infrastructure does. AMD is positioning itself as the primary and natural home for that next category.
We saw that AMD’s Q2 2026 data center revenue was $6.7 billion, up 107% year over year, accounting for 58% of the total company revenue of $11.5 billion, according to AMD’s Q2 earnings release.
The Anthropic partnership includes plans to deploy up to 2 gigawatts of MI450 Series GPUs. Microsoft Azure, Oracle, Meta, and OpenAI all have meaningful AMD Helios deployments underway.
AI systems that matter most a decade from now may not be the ones that can just read text and generate language.Shutterstock
The bigger picture of what this means for the AI compute market
Nvidia now sits at a $5 trillion market cap. That number remains almost incomprehensible. AMD at $1 trillion is a fifth of that, but growing faster from its current base.
Of course, the World Labs acquisition does not change that gap overnight.
What it changes is the research agenda for shaping the next decade of compute infrastructure. By hiring one of AI’s most respected scientists and the team she built around spatial intelligence, AMD is buying revenue, patents, direction, and much, much more.
Li understands that the AI systems that matter most a decade from now may not be the ones that can just read text and generate language. They could be the systems operating physical robots, or simulating environments for autonomous-vehicle training and generating the synthetic data that physical AI will need at massive scale.
I think we may be getting there sooner than most expect. Li is thinking several moves ahead, and AMD is betting $8.2 billion that she can see which hardware those systems will need.
The answer does not have to be Nvidia.
Related: AMD just hit a milestone that reshuffles the AI chip race