Why Nvidia and AMD Backed Fei-Fei Li’s World Labs in a $1 Billion Spatial AI Megaround

AI pioneer Dr. Fei-Fei Li and her newly minted startup, World Labs, have secured a staggering $1 billion in fresh funding to tackle the next grand challenge of artificial intelligence: teaching machine learning models to understand, simulate, and interact with the physical, three-dimensional world. The massive transaction elevates the stealthy startup into unicorn status overnight and draws an extraordinary coalition of backing from semiconductor rivals Nvidia and AMD, alongside venture giants and software institutions.

For the past two years, the tech industry has been obsessed with large language models (LLMs) that manipulate text, code, and flat 2D pixels. But the limits of scaling these text-based architectures are becoming obvious, and the race is shifting toward what Dr. Fei-Fei Li calls "spatial intelligence." By building foundational models that grasp physical laws, spatial dimensions, and object relationships, World Labs is trying to create the cognitive engine required for advanced robotics, spatial computing, and physical world simulations.

The Physics of Spatial Intelligence

At its core, World Labs spatial AI funding is a massive bet that the next platform shift in AI will not occur in a chat box. Today's generative AI models are essentially correlation engines operating on tokenized sequences; they do not understand gravity, depth, structural integrity, or how an object behaves when it is pushed off a table. If you ask an LLM or a diffusion model to generate a room, it draws a flat picture that looks convincing but lacks structural logic. If you try to navigate that room in virtual space, the illusion instantly falls apart.

World Labs is designing "large world models" (LWMs) capable of generating fully interactive, geometrically consistent 3D environments. This technology represents the missing middle layer between passive generative art and active physical robotics. By teaching models the physical priors of our world, World Labs aims to enable software that can not only observe the environment but predict how its structure changes over time and react to those changes in real-time.

The Silicon Alliance: Why Nvidia and AMD Joined Forces

Perhaps the most significant signal of the World Labs round is its cap table. Securing capital simultaneously from Nvidia and AMD is an exceedingly rare feat in the current hyper-competitive silicon landscape. It highlights a fundamental truth of the next computing era: spatial AI is a voracious consumer of compute, and whoever wins this paradigm wins the next decade of chip sales.

For Nvidia, investing in World Labs aligns perfectly with its Omniverse initiative and its push into industrial robotics. Nvidia CEO Jensen Huang has repeatedly stated that the next wave of AI will be physical. By backing Dr. Fei-Fei Li's team, Nvidia ensures its GPUs remain the default development environment for spatial foundations. For AMD, the investment is a strategic land grab, proving that its ROCm software ecosystem and Instinct accelerators are capable of supporting the most cutting-edge, trillion-parameter physical world models alongside its chief rival.

The round also drew capital from design software powerhouse Autodesk, Emerson Collective, Fidelity, and global tech conglomerate Sea. The presence of Autodesk is highly telling; a spatial model that understands 3D geometry is the ultimate tool for future architects, industrial designers, and game developers, transforming static CAD files into dynamic, AI-simulated worlds.

The Developer Layer: Code Metal’s Rapid Rise

While World Labs captured the lion's share of headlines this week, it wasn't the only startup proving that investors are shifting capital toward actionable, physics-adjacent AI. Developer-focused startup Code Metal secured $125 million in a Series B round led by Salesforce Ventures—occurring a mere three months after its Series A round.

Code Metal is targeting the notoriously complex world of edge computing and hardware-level code translation. While tools like GitHub Copilot excel at high-level languages like Python and TypeScript, writing efficient code for edge devices, microcontrollers, and IoT hardware requires deep architectural awareness. Code Metal’s AI translates complex software structures into highly optimized machine code tailored for specific silicon configurations, drastically reducing the time required to deploy smart models to physical hardware. The rapid succession of its funding rounds underscores a growing urgency to optimize software for the real, constrained physical devices that exist outside of central cloud data centers.

The Barbell Strategy of Late-Stage Venture Capital

The broader funding landscape this week reflects a clear barbell strategy among late-stage venture capitalists. On one side of the barbell, we see massive, speculative bets on deep-tech paradigms that could take years to fully mature, such as World Labs' spatial intelligence. On the other side, investors are pouring hundreds of millions into highly practical, cash-generative, and sticky enterprise software infrastructure.

For instance, savings and fintech platform Vestwell raised $385 million in Series E funding at a $2 billion valuation, led by Blue Owl Capital and Sixth Street Growth. Meanwhile, workflow state management platform Temporal Technologies brought in $300 million, pushing its valuation to $5 billion. These rounds demonstrate that while foundational AI research commands eye-watering premiums, the plumbing of the modern digital economy—reliable financial ledgers and resilient backend orchestration—still commands immense, premium-tier valuations.

The Takeaway

The era of treating AI as a disembodied digital oracle is ending. By backing World Labs with $1 billion, the tech industry's elite have declared that the future of machine intelligence belongs to systems that can navigate, understand, and simulate the physical 3D world we inhabit. Startups that can successfully bridge the gap between digital computation and physical reality are no longer just building software; they are building the operating system for the physical future.