General Intuition's $320M bet that video games can train real-world AI

AUSTIN - General Intuition, an AI lab building 'large action' models trained on gameplay footage, has raised a $320 million Series A at a $2.3 billion valuation, the company said. Khosla Ventures led the round, joined by General Catalyst, Hedosophia, Bezos Expeditions, Innovation Endeavors and others, bringing total disclosed funding to roughly $454 million after the lab's launch last October.

The company's pitch is unusual: it uses hundreds of millions of hours of uploaded gameplay - complete with the exact button presses players made and when - to teach AI systems to understand, predict and act in dynamic environments. The end goal is software that can reason about space and time well enough to guide robots in the physical world. General Intuition said it plans to spend most of the new capital on compute and to open its API to more developers by late summer.

Why it matters

General Intuition is headquartered in New York, not Austin - but the spatial-reasoning problem it is funding is the exact frontier Austin's robotics cluster, from humanoid-maker Apptronik to autonomy startup Saronic, is racing to solve. As capital floods into physical AI, the talent and supplier networks forming in Central Texas stand to benefit from whichever lab cracks general-purpose machine intuition first.

The company's training uses hundreds of millions of hours of uploaded gameplay as its initial dataset to train in spatial-temporal reasoning, with the key ingredient being the action labels embedded in those clips.— TechCrunch, June 25, 2026