General Intuition, an artificial intelligence startup focused on building AI agents capable of understanding and navigating the physical world, has raised $320 million in a Series A funding round, taking its post-money valuation to $2.3 billion. The funding underscores growing investor confidence in a new category of AI that moves beyond language models and into real-world reasoning through video game data.
The round was led by Khosla Ventures and included participation from General Catalyst, Hedosophia, Bezos Expeditions, Innovation Endeavors and former Formula One world champion Nico Rosberg. The company, founded by Medal CEO Pim de Witte, plans to use the capital to accelerate development of foundation models designed to teach AI systems how to perceive, reason and act in dynamic environments.
Unlike many AI startups focused on text generation or conversational assistants, General Intuition is building models for spatial-temporal reasoning. Its central thesis is that video games provide one of the richest sources of training data for AI because they capture how humans navigate complex environments, solve problems, react to uncertainty and make decisions in real time.
The startup draws on Medal's extensive gaming platform, where users upload roughly two billion gameplay videos each year from tens of thousands of games. According to the company, these videos expose AI models to diverse scenarios involving movement, planning, exploration and interaction, creating datasets that are difficult to replicate through conventional robotics training.
The long-term ambition extends well beyond gaming. General Intuition believes the same models trained on virtual environments can eventually support autonomous robots, drones, industrial automation systems and self-driving vehicles operating in the physical world. The approach reflects growing industry interest in so-called embodied AI, where machines must understand both language and the physical environments around them.
Interest in world models has accelerated across the AI industry. Companies including Google DeepMind have introduced systems capable of generating interactive environments for training AI agents, while robotics companies increasingly rely on simulated worlds to prepare autonomous systems for real-world deployment. These developments are encouraging investors to back startups focused on physical intelligence rather than language generation alone.
The funding also illustrates how data ownership is becoming a competitive advantage in AI. Rather than relying solely on publicly available internet content, General Intuition is building its models around proprietary gaming datasets collected through Medal's platform. Investors increasingly view exclusive data sources as a differentiator as foundation model developers compete to improve AI capabilities.
The latest investment follows one of the largest AI seed rounds announced last year, when the company raised approximately $134 million shortly after spinning out from Medal. The new valuation represents a significant increase in less than a year, highlighting continued appetite for frontier AI companies despite growing competition across the sector.
For enterprise technology leaders, the announcement signals that the next phase of AI investment is increasingly focused on systems capable of interacting with the physical world. As AI expands into robotics, logistics, manufacturing and autonomous mobility, demand is expected to rise for models that can understand movement, space and decision-making beyond text.
The funding also reflects a broader shift in how the AI industry is approaching intelligence. Rather than viewing games purely as entertainment, companies are beginning to treat virtual environments as large-scale learning laboratories for future AI agents. If General Intuition's approach proves successful, video game data could become an important building block for autonomous systems that eventually operate across factories, warehouses, cities and other real-world environments.