General Intuition, an AI startup spun out of gaming clip platform Medal, has raised $320 million at a $2.3 billion valuation on a contrarian thesis: that player-annotated video game footage is the most underrated training data in robotics. The round was led by Khosla Ventures, with General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg and researchers from Google DeepMind and MIT participating.
The bet on gameplay data
Medal has roughly 10 million monthly active users who upload around 2 billion clips a year. Crucially, those clips can carry the controller inputs behind the action — not just what happened on screen, but what a person did to make it happen. That pairing of visible outcome and human intent is exactly the spatial-temporal data robotics labs struggle to collect at scale. Founder Pim de Witte reportedly turned down a $500 million OpenAI acquisition of Medal to build the company instead.
Compute and the road to an API
The round brings disclosed funding to $454 million after a $134 million launch raise in October 2025. Most of the new capital goes toward compute: the next model will train on CoreWeave infrastructure, with an API planned by the end of summer 2026. The skeptic's case is fair — game engines are cleaner than factory floors — but General Intuition argues scale and real human decisions, not synthetic trajectories, are what set its dataset apart.
A crowded race for robot brains
The deal fits a surge of capital into embodied AI, where venture funding has climbed from $4 billion in 2019 to well over $20 billion in 2026. It joins heavyweight raises like Skild AI's $1.4 billion round, Neura Robotics' Series C and data-centric plays such as Mecka AI. The hard part remains unproven: a big dataset can get a company funded, but it cannot by itself make a robot useful.
Reporting based on coverage from TechCrunch, Khosla Ventures and Startup Fortune.