Swiss AI startup raises €114M to train robots using gaming data

Swiss AI startup General Intuition has raised €114 million in one of 2025’s largest seed funding rounds, led by Khosla Ventures and General Catalyst with participation from Raine. The New York and Geneva-based company, spun off from Dutch video platform Medal, plans to bridge gaming data with real-world applications in robotics and drones, leveraging over 2 billion annual gameplay clips to train AI agents for both virtual and physical environments.

What you should know: General Intuition emerged from Medal’s extensive gaming content library, which processes nearly 1 billion video uploads annually from 10 million monthly active users.

• The startup uses organically uploaded gameplay footage featuring dramatic successes and failures as training data for AI systems with spatial and temporal awareness.
• Unlike traditional AI companies relying on curated datasets, General Intuition leverages naturally occurring edge cases from real player interactions across tens of thousands of gaming environments.

The big picture: This funding round significantly outpaces typical European AI and robotics investments, highlighting a stark contrast in both scale and ambition within the sector.

• Most European startups in embodied AI, agentic systems, or robotics have raised substantially smaller amounts—Energy Robotics secured €11.5 million, Unchained Robotics raised €8.5 million, and Cyberwave closed a €7 million round.
• While most 2025 AI and robotics rounds reported by EU-Startups ranged between €3 million and €12 million, General Intuition’s raise underscores the company’s cross-continental structure and ambitious scope.

How it works: The company’s AI agents learn purely through visual input, mimicking human perception and spatial navigation patterns.

• “When you play video games, you essentially transfer your perception, usually through a first-person view of the camera, to different environments,” explained CEO and Co-founder Pim de Witte.
• Agents only see what human players would see and move through space by following controller inputs, enabling natural transfer to physical systems like robotic arms, drones, and autonomous vehicles.
• “You get this selection bias towards precisely the kind of data you actually want to use for training work,” de Witte added.

Why this matters: The approach addresses critical data constraints in next-generation AI development while targeting real-world applications beyond gaming.

• “This next frontier in AI requires large scale interaction data, but is severely data constrained. Meanwhile, nearly 1 billion videos are posted to Medal each year,” said de Witte.
• The company’s emphasis on embodied AI has practical applications including search-and-rescue drones that can interpret and navigate unfamiliar terrains without GPS reliance.
• General Intuition believes large language models (LLMs) alone are insufficient for achieving artificial general intelligence due to their lack of understanding of physical and spatial dynamics.

What’s next: The startup will use funding to scale its research team and advance AI agent development for both virtual and real-world environments.

• Commercial plans include launching AI-powered non-player characters (NPCs) and simulation tools by the first half of 2026, offering interactivity and adaptability beyond deterministic, rule-based bots.
• Research targets include agentic systems capable of learning from unstructured video, world models that simulate dynamic environments for training, and video understanding applications beyond gaming.
• Structured as a public-benefit corporation, the company aims to enhance rather than replace creative roles in the gaming industry.