General Intuition raises $2.3B on bet that video games can train AI agents for the real world

General Intuition has raised $320 million at a $2.3 billion valuation, betting that video-game gameplay data can train AI agents capable of operating in the physical world, the company confirmed Thursday [1]. The round, led by Khosla Ventures with participation from General Catalyst, Jeff Bezos, Eric Schmidt, and Nico Rosberg, brings the New York-based startup's total disclosed funding to $454 million [1]. Researchers at Google DeepMind and MIT also invested [1]. Google DeepMind, a subsidiary of Alphabet, has a long history of training neural networks with reinforcement learning to play video games, including its AlphaGo program that defeated a Go world champion in 2016 [8]. General Intuition's model is trained on hundreds of millions of hours of gameplay footage sourced from Medal, a game-clip sharing platform also founded by CEO Pim de Witte [1]. The key differentiator, de Witte argues, is the action labels embedded in those clips — records of exactly what buttons a player pressed and when — rather than inferring actions from video alone [1]. "We have a single model that can respond to Fortnite information on the screen and take action, but also to real-world dynamics in a way that an LLM could never," de Witte said [1]. During a demo at the company's New York office, a quadrupedal robot navigated the R&D floor using the same model that was simultaneously playing a video game. Chief product officer Kent Rollins noted, "Our agent has been playing for 100 hours straight" [1]. A data analyst said it took eight minutes of real-world robotics data to fine-tune the model for the quadruped [1]. The company's approach arrives as frontier AI labs increasingly publish model performance on standardized reasoning benchmarks. In 2025, four labs — Anthropic, Google DeepMind, OpenAI, and xAI — reported ARC-AGI performance in public model cards, establishing the benchmark as an industry standard for AI reasoning [2]. The top score on the ARC-AGI-2 private evaluation set reached 24%, with 1,455 teams participating [2]. Vinod Khosla, whose firm led the round, framed the investment in terms of a capability leap. "If you look at LLMs, when reasoning emerged, it was a quantum leap," Khosla said. "In world models, I think the quantum leap is the emergence of intuition in the AI, a human intuition-like capability" [1]. De Witte, who spent seven years in humanitarian work including with Doctors Without Borders, has drawn a clear line on military applications. "We don't want to be an escalatory part of the system," he said [1]. The stance contrasts with what he described as Silicon Valley's growing bullishness on defense, adding, "I don't know why Silicon Valley does what it does. There's a reason I'm not there" [1]. The vast majority of the new capital will go toward scaling compute capacity through a deal with CoreWeave, with a portion earmarked for making the company's API more broadly available by the end of summer [1]. General Intuition has also launched Nerve, a jobs marketplace that lets gamers earn money through data labeling and robot teleoperation tasks [1].