Fei-Fei Li's World Labs Raises $1B for Spatial AI Models

Fei-Fei Li, the Stanford computer scientist who helped catalyze modern AI through her work on ImageNet, has emerged from stealth mode with World Labs, a San Francisco startup valued at $1 billion that's building what it calls "spatial intelligence."

The company's first product, Marble, launched in November and represents a fundamental shift in how AI systems understand the world. Rather than processing text or generating flat images, Marble creates interactive 3D environments from photos or written descriptions—what the industry calls "world models."

From text to spatial understanding

World models differ from large language models in a crucial way: they're designed to comprehend physical space and geometry, not just language patterns. Feed Marble a few iPhone photos of a room, and within minutes it generates a navigable 3D replica complete with spatial relationships, lighting, and structure.

The technology isn't perfect yet. When a reporter tested Marble by photographing World Labs' own lobby—featuring a rainbow-tiled wall, grand piano, and oversized iridescent spheres—the resulting simulation showed scrambled surfaces "like a cubist painting left out in the rain," according to Fast Company, which first reported details of the company's progress.

But Li, serving as CEO, acknowledged Marble represents just the beginning. "Marble is the first part of [our] journey—it's not a fully matured model," she said during a tour of the company's offices.

Commercial traction in creative industries

Despite its early-stage limitations, World Labs has already secured paying customers across multiple sectors. The company reports active use cases in virtual film production, architectural design, and robotics training—domains where generating realistic 3D spaces offers immediate practical value.

The shift from language models to world models reflects broader momentum in AI investment. Venture capital is increasingly flowing toward systems that can interact with physical reality, particularly as robotics and spatial computing applications mature.

Why it matters

World Labs' $1 billion valuation signals investor confidence that the next wave of AI breakthroughs will come from systems that understand three-dimensional space, not just text. For industries from manufacturing to entertainment, AI that can model physical environments could unlock applications impossible with today's language-focused systems. Li's track record—her ImageNet dataset became foundational training data for modern computer vision—gives World Labs credibility that spatial intelligence represents a genuine frontier, not just rebranded hype.

Li, often called the "godmother of AI" (a nickname she reportedly dislikes), brings significant scientific credibility to the spatial intelligence thesis. Her ImageNet project, which organized millions of labeled images, became the foundation for training modern computer vision systems.

Details of World Labs' technology and business progress were first reported by Fast Company.