General Intuition Raises $320M at $2.3B Valuation for AI Game Training
Original: General Intuition’s $2.3B bet that video games can train AI agents for the real world
Why This Matters
Demonstrates viable path for scaling AI robotics using existing gameplay data, addressing critical bottleneck in embodied AI development.
General Intuition secured $320 million in Series B funding at a $2.3 billion valuation to develop AI agents trained on video game footage. The company uses gameplay data with embedded action labels to train models for real-world robotics applications, with a quadrupedal robot demonstration showing the technology's effectiveness.
General Intuition, co-founded by 31-year-old Pim de Witte, announced a $320 million Series B funding round on June 25, 2026, bringing total disclosed funding to $454 million following a $134 million seed round in October 2025. The New York-based startup focuses on training AI agents using video game data to control real-world robots and autonomous systems. The company demonstrated an AI agent playing Fortnite for 100 hours continuously, with the same underlying model powering a quadrupedal robot navigating the company's office. The robot required only eight minutes of real-world robotics data collected on the street to fine-tune the model, showcasing generalization capabilities across virtual and physical environments. General Intuition's competitive advantage lies in using action labels embedded in gameplay footage from its parent company Medal—a platform for sharing video game clips—rather than inferring actions from video alone. De Witte stated the approach represents spatial-temporal reasoning that traditional large language models cannot achieve. The model processes both virtual game information and real-world dynamics to control embodied agents, positioning the technology as a foundational pre-training method for future AI systems.