General Intuition Eyes $6 Billion Valuation in Rapid-Fire Funding Round for Physical AI

General Intuition is finalizing a new financing round that values the physical intelligence startup at $6 billion pre-money, according to people familiar with the transaction. The dramatic step-up arrives merely weeks after the company closed a $320 million tranche at a $2.3 billion valuation.

The oversubscribed round draws fresh capital from Valor Equity Partners, Point72 Ventures, and Seven Seven Six. Existing backers Khosla Ventures and General Catalyst are also re-upping their commitments. For Valor, known for its long-standing stake in SpaceX, the deal represents a rare high-conviction push into frontier foundation model laboratories.

Spun out last October by CEO Pim de Witte from gameplay-sharing network Medal, the New York-based startup takes an unconventional path toward building large action models. Rather than relying purely on static images or text tokens, the lab trained its initial systems on hundreds of millions of hours of video game sessions paired with granular input telemetry, mapping visual frames directly to the specific controller actions executed by players.

Backers argue this telemetry creates a dynamic spatial baseline that standard vision-language models lack. Vinod Khosla previously noted that these action labels serve as a foundational mechanism for the "emergence of intuition," granting software the capacity to navigate and generalize across unfamiliar environments without bespoke task programming.

Proceeds from the round will primarily fund compute capacity and talent acquisition as the startup pivots its spatial architecture toward hardware embodiments and robotic control. General Intuition already maintains an infrastructure partnership with specialized cloud operator CoreWeave to power its training runs.

The Embodied AI Premium

The company's rapid ascent reflects an intense investor appetite for physical AI architectures capable of bridging digital simulation and physical reality. While foundation models have transformed text and image generation, mastering three-dimensional interaction remains an elusive frontier; whether high-speed gaming telemetry translates smoothly to real-world mechanical actuators remains the billion-dollar question.