Fei-Fei Li

Who they are

Fei-Fei Li is Cofounder and CEO of World Labs — the AI company she founded to build frontier spatial-intelligence models, which launched its first product, Marble, in November 2025.

Person

Fei-Fei Li is one of the architects of the modern deep-learning era — best known for creating ImageNet, the large-scale visual dataset that catalyzed the computer vision revolution of the 2010s. Her academic career ran through Caltech and Stanford, where she spent years leading the Stanford AI Lab and Human-Centered AI Institute. She stepped out of academia to serve as Chief Scientist of AI/ML at Google Cloud from 2017 to 2018, one of the few researchers of her stature to take a senior industry role — then returned to Stanford before founding World Labs. The company she founded, World Labs, is focused on spatial intelligence: building AI models that understand and generate 3D worlds for storytelling, robotics, creativity, and scientific discovery. The through-line is a sustained bet that visual and spatial understanding — not just language — is where foundational AI progress happens. She writes and speaks publicly on AI ethics, human-centered AI, and the responsibilities of researchers building powerful systems.

Company

World Labs closed a $1 billion funding round in February 2026, bringing total funding to $1.23 billion and a valuation of $5 billion — the largest capital raise for a company explicitly focused on spatial intelligence. Also in February 2026, World Labs announced a partnership with Autodesk to integrate its AI models with Autodesk's 3D design tools, targeting media and entertainment use cases first. The company had launched its first product, Marble, in November 2025 — a tool that generates 3D virtual worlds from image or text prompts. Backers in the February round include Nvidia, AMD, and Autodesk, signaling deep hardware and toolchain alignment from the start.

Market

World Labs sits at the intersection of generative AI and 3D spatial computing — a field where the main named competitors include Matterport, Cognite, and Hexagon. Among its 254 active competitors tracked by Tracxn, World Labs holds the highest total funding and ranks 4th in the peer set, giving it a meaningful capital advantage. US-China geopolitical competition in AI is creating technology blocs that could affect international collaboration and market access, a dynamic that shapes the environment for any frontier AI company with global ambitions.

How they likely show up

Long tenure in academic leadership at Stanford (Stanford AI Lab, HAI) → likely operates with deep patience for foundational, multi-year research bets rather than short product cycles.
Founder role at World Labs after decades as a researcher → high agency, willing to leave institutional security to own outcomes directly.
Created ImageNet as a large-scale infrastructure project before its value was obvious → comfortable building things whose payoff is non-linear and deferred.
Public writing and speaking on AI ethics and human-centered AI → brings a normative lens to technical decisions; won't respond well to pitches that treat safety as an afterthought.
Stepped into a senior Google Cloud industry role and then returned to academia → able to context-switch between research and commercial environments, but academia and research credibility are core to her identity.

Conversation tips

→ Reference ImageNet or the spatial-intelligence thesis specifically — she'll distinguish between people who know the research arc and people who just know the name.
→ Engage seriously on the science, not just the product roadmap — she founded World Labs around a research conviction, and the commercial layer follows from it.
→ Ask about the Autodesk partnership and the media-and-entertainment entry point — it's a deliberate wedge strategy worth unpacking.
→ Don't compress the conversation to 'what does World Labs do' — she thinks in terms of what spatial intelligence enables across robotics, creativity, and science, and will appreciate a question that spans those domains.

Toolbox

Openers

The February 2026 Autodesk partnership is a live integration story — asking how World Labs is scoping the media-and-entertainment entry point opens a concrete product conversation.
The $1.23 billion in total funding and the $5 billion valuation set a high bar for infrastructure partners — any vendor who can credibly serve at that scale has an immediate anchor.
Marble launched only in November 2025 — the product is weeks into its commercial life, which means the tooling and workflow questions around 3D world generation are genuinely open.

Discovery questions

How does the Autodesk partnership shape which use cases World Labs prioritizes first — is media and entertainment a proving ground before robotics, or a parallel track?
Marble generates 3D worlds from image or text prompts — where does the human creative workflow fit in, and how are you thinking about the feedback loop between the model and the user?
With Nvidia, AMD, and Autodesk all in the cap table, how do you manage the tension between being a neutral platform and being aligned with specific hardware and toolchain bets?

Avoid

Don't pitch language-model or text-AI capabilities as the core value prop — her entire research identity and company thesis is built on visual and spatial intelligence being a distinct and underinvested frontier.

Make it yours