Skip to content
News Funding & M&A

General Intuition in talks to raise $300M at $2B valuation to train AI world models on gaming video

· by Pondero Newsdesk

The short version

The New York startup, spun out of game-clip platform Medal, trains embodied AI agents on 2 billion gaming video clips per year and counts Jeff Bezos and Eric Schmidt among its backers.

General Intuition in talks to raise $300M at $2B valuation to train AI world models on gaming video

A New York startup that trains AI on gaming video footage is in talks to raise around $300 million at a valuation just above $2 billion, sources familiar with the matter told TechCrunch on June 18. The company: General Intuition.

What

Eight months ago, General Intuition spun out of Medal, a gaming clip-sharing platform, with a $134 million seed round. The new raise would bring the valuation to just over $2 billion, which is roughly triple the earlier figure, per TechCrunch.

The core asset is Medal's dataset. That platform ingests approximately 2 billion gaming video clips per year from more than 10 million monthly active users. General Intuition takes those clips and uses them to train world models: AI systems built to understand how objects and agents move through physical space and time. Pim de Witte, who co-founded Medal, leads General Intuition alongside researchers Eloi Alonso, Adam Jelley, and Vincent Micheli, each of whom specializes in world modeling and simulation.

Jeff Bezos and Eric Schmidt are among the backers, joined by existing investors Khosla Ventures and General Catalyst. The round has not yet closed and no lead investor has been named publicly. New capital will go toward scaling compute, with a product launch targeted before the end of summer or early fall, a source told TechCrunch.

Why it matters

Training data is the contested resource in the world model race, and gaming footage has properties that are hard to replicate elsewhere. It is first-person, interactive, and physics-consistent: exactly the kind of environment that teaches a model how agents act inside a space rather than simply what a scene looks like. General Intuition's pitch rests on the claim that no comparable dataset exists at scale.

The value of that data was apparently visible to OpenAI before the spin-out even happened. According to reporting cited by TechCrunch, OpenAI attempted to acquire Medal. Other large AI labs have since approached General Intuition as well, per TechCrunch, though no acquisition went through.

Runway, Decart, and World Labs (founded by Fei-Fei Li) have all released world models in the past twelve months. Google's Genie 3 began incorporating Street View imagery for real-world simulation. What separates General Intuition's approach is where the output goes: the company builds world models to train AI agents, not to license as a foundation model. The agents are the product.

That distinction matters for buyers. Enterprise teams building embodied AI, robotic systems, or autonomous agents that need spatial reasoning get a packaged agent rather than raw infrastructure. Whether the product delivers on that framing is something the planned summer launch will begin to answer.

What to watch next

Two things: the formal close, which will confirm the round size and name the lead investor; and the summer product release, which would be General Intuition's first commercial output since spinning out of Medal. A public benchmark for spatial reasoning would also be significant. The company has not yet released one.

Sources