A World Models Startup Just Tripled Its Valuation in Two Months

Two months ago, General Intuition was worth $2.3 billion. Now it's reportedly closing in on $6 billion, built entirely on video game footage.

August 25, 2026
A World Models Startup Just Tripled Its Valuation in Two Months AI Investment

Summary: General Intuition, a startup training world models on hundreds of millions of hours of gameplay footage from CEO Pim de Witte's platform Medal, is reportedly finalizing a funding round that would value the company at $6 billion, up from $2.3 billion just two months ago. The pace mirrors a broader rush into world models this summer, with Amazon backing Odyssey and Alibaba launching its own Qwen robotics suite within days of each other in June.

worldmodelsgeneralintuitionairoboticsaifundingkhoslaventuresaistartups

Two months ago, General Intuition was worth $2.3 billion. Now it's reportedly closing in on $6 billion.

TechCrunch reported Monday that the startup is finalizing a new funding round at that valuation, citing unnamed sources. The company hasn't confirmed the report.

If accurate, the jump is striking on its own terms. Most startups take years, not two months, to nearly triple in value.

From Gaming Clips to Robot Training Data

General Intuition trains world models, AI systems built to understand physical space rather than generate text. Its training data comes from an unusual source.

That distinction is the whole premise behind the world models category. A language model learns from text. A world model needs to learn how objects and bodies actually move.

CEO Pim de Witte also founded Medal, a video game clip-sharing platform. General Intuition's model draws on hundreds of millions of hours of gameplay footage from Medal's 17 million monthly users.

The company describes its own mission bluntly. It called itself "the frontier lab for acting in space and time" in a June post announcing its Series A.

That round raised $320 million at a $2.3 billion valuation. Backers included Khosla Ventures, former Google chairman Eric Schmidt and Amazon founder Jeff Bezos.

Having Schmidt and Bezos personally involved, rather than just their firms, is itself a signal. Individual investors of that stature don't typically write checks into unproven technology categories casually.

Before that, General Intuition raised a $133.7 million seed round last October. The company plans to use new capital for compute infrastructure and talent, with a specific focus on robotics.

The pattern across all three rounds is consistent. Bigger checks, faster, from increasingly high-profile backers, with the underlying technology still unproven at commercial scale.

A Global Race, Not Just One Company

General Intuition's pace isn't happening in isolation. Three major world models announcements landed within about ten days of each other in June alone.

Amazon backed and partnered with Odyssey, a startup building AI that simulates 3D physical environments. That round valued Odyssey at $1.45 billion.

Alibaba unveiled its own robotics AI suite the same week. It includes Qwen-RobotWorld, a system that predicts how physical scenarios will unfold.

MAIN covered a similar race playing out in China earlier this month. Tsinghua-founded startups there are racing to build the same underlying technology.

This is the American and Chinese sides of the identical bet, made almost simultaneously.

What This Means for Miami

World models represent a real shift in where AI investment is heading. Less text generation, more systems that understand and predict physical space.

That matters beyond robotics labs. Logistics, construction, real estate and manufacturing all depend on understanding physical environments, not just processing language.

None of these companies has shipped a robot doing real commercial work at scale yet. The valuations are running well ahead of the actual product.

For Miami investors watching AI valuations move this fast, General Intuition's trajectory is a useful marker. A company can nearly triple its valuation in two months right now. That's true even in a technology category still working out its own definition of success.