Fei-Fei Li
Co-founder and CEO, World Labs
Fei-Fei Li's ImageNet made a heretical claim stick: that data, not algorithms, was computer vision's real bottleneck — and AlexNet's 2012 breakthrough on her benchmark proved it. She now leads World Labs as co-founder and CEO, where the Marble model generates persistent, editable 3D worlds, alongside her Stanford professorship and co-directing role at Stanford HAI.
Why they matter
Li's second act applies her original insight to a new frontier: World Labs argues that spatial understanding — how scenes, objects, and physics fit together — is a capability language models cannot encode, and shipped Marble as the first commercial world model while rivals kept theirs in research previews. Investors backed the thesis with $1 billion, Autodesk anchoring $200 million of it.
Leadership intelligence
TECHi profile brief
Fei-Fei Li is co-founder and CEO of World Labs and a Stanford professor who created ImageNet, the dataset that catalyzed modern computer vision. World Labs released Marble, its first commercial world model for generating persistent 3D environments from text, images, and video, in November 2025 on freemium subscription tiers. The company went on to raise $1 billion in a round anchored by Autodesk's $200 million commitment, with NVIDIA, AMD, Andreessen Horowitz, and Fidelity participating, following reports of a roughly $5 billion valuation.
TECHi thesis
Li keeps winning the same argument twice. When the field believed algorithms were the bottleneck, ImageNet proved data was — AlexNet's 2012 breakthrough ran on her dataset. Now the field believes language is the path to general intelligence, and World Labs bets that spatial understanding is what text can never deliver. Marble is the test: a commercial 3D world generator shipped on freemium tiers while Google and OpenAI kept their world models in research demos. Being first matters less than whether 3D generation finds paying professional users in games, film, and simulation before the giants productize theirs. If Marble stalls as a curiosity, spatial intelligence gets absorbed as a feature of bigger models — and the second act ends as a footnote to the first.
Distribution leverage
86/100World Labs is small, but Li's reach spans three channels most founders never touch: a shipped product in Marble, Stanford HAI's policy pipeline, and a cap table — Autodesk, NVIDIA, AMD — that doubles as a distribution partner list.
Execution record
82/100ImageNet is arguably the highest-leverage artifact in modern AI relative to what it cost to build, and Marble shipped commercially before Big Tech's world models left the lab. Sustained product revenue is the unproven piece.
Audience demand
78/100Search interest concentrates on spatial intelligence, Marble's launch, and World Labs' funding — a narrower but higher-intent audience than celebrity AI executives draw.
Career map
Inflection points
Origin
ImageNet era
Created ImageNet as a professor, betting career capital that scale of labeled data would beat cleverer algorithms — the bet AlexNet vindicated in 2012.
Scale
Human-centered AI
Built Stanford HAI into a leading academic center for AI's societal effects while remaining an active vision researcher.
Watch
Spatial intelligence
Whether Marble converts freemium curiosity into paying professional adoption in games, film, and simulation before Google and OpenAI commercialize their own world models.
Search answers
People ask about Fei-Fei Li
Who is Fei-Fei Li?
Fei-Fei Li is co-founder and CEO of World Labs, a Stanford computer science professor, and the creator of ImageNet, the dataset that launched the deep-learning era of computer vision. She also co-founded Stanford's Institute for Human-Centered AI, the university's hub for AI policy and societal-impact research.
Why is Fei-Fei Li important in AI?
She proved that data, not algorithms, was the binding constraint on computer vision — AlexNet's breakthrough ran on her ImageNet benchmark. Now she is making the parallel case for spatial intelligence: that 3D understanding of scenes and physics is a capability language models cannot reach, and building World Labs to deliver it commercially.
What is Fei-Fei Li known for?
ImageNet, the human-centered AI agenda at Stanford HAI, and World Labs' Marble — the first commercially released world model, which generates persistent, editable 3D environments from text, images, and video on freemium subscription tiers.
Can you invest in World Labs stock?
No — World Labs is private. It raised $1 billion in a round anchored by Autodesk's $200 million commitment, with NVIDIA, AMD, Andreessen Horowitz, and Fidelity among the backers, after reports of a roughly $5 billion valuation. Public-market exposure is limited to listed investors such as NVDA, AMD, and ADSK.