General Intuition, a New York-based AI startup backed by Jeff Bezos and Eric Schmidt, has raised $320 million at a $2.3 billion valuation — betting big that video game data can train the next generation of AI agents to navigate the real world.
The Big Picture
While most of the AI world has focused on large language models (LLMs) that parse text and images, General Intuition is chasing something different: spatial-temporal reasoning — the ability to understand how objects move through space and time. The company believes LLMs hit a ceiling when it comes to understanding physics, cause-and-effect, and embodiment, and that video game telemetry data is the missing key.
Spun out of Medal TV — a clip-sharing platform for gamers — the startup tapped into hundreds of millions of hours of gameplay footage paired with exact button-press data. This “action-labeled” data gives AI models a far richer signal than raw video alone, the company argues.
The Demo That Stunned TechCrunch
On the company’s R&D floor, an AI agent trained on Fortnite gameplay was running non-stop for 100 hours. The same AI model also powered a quadruped robot that walked autonomously around the office, navigating chairs, tables, and unexpected obstacles. According to the team, it took just eight minutes of real-world robotics data to fine-tune the model for the physical robot.
“The same brain powering the agent playing the game is powering the robot,” CEO Pim de Witte told TechCrunch.
Who’s Investing?
The $320M round was led by Khosla Ventures and included General Catalyst, Jeff Bezos, Eric Schmidt, former F1 champion Nico Rosberg, and researchers from Google DeepMind and MIT. To date, General Intuition has raised $454 million in total.
“In world models, I think the quantum leap is the emergence of intuition in the AI — a human intuition-like capability,” said Vinod Khosla. “The human action data and reaction data you have in games is the key part to the emergence of intuition.”
What This Means for the Industry
General Intuition isn’t building a self-driving car company or a robotics firm. Instead, it wants to be the backbone infrastructure — a model provider like OpenAI or Anthropic — powering everything from warehouse robots to gaming NPCs to drone navigation. Its API is expected to become broadly available by the end of summer 2026.
The startup also launched Nerve, a platform that lets gamers earn money through data labeling and robot teleoperation — creating a data flywheel while giving displaced workers a stake in the AI economy.
Key Takeaways
- Funding: $320M Series Unknown at $2.3B valuation
- Total Raised: $454M
- Key Backers: Khosla Ventures, Jeff Bezos, Eric Schmidt, Google DeepMind, MIT
- Technology: World models + agentic AI trained on video game action data
- Vision: A generalized AI backbone for robots, games, and simulations
FAQ
How is General Intuition different from OpenAI or Anthropic?
Those companies build LLMs focused on text. General Intuition builds world models that understand physics, movement, and causation — skills LLMs struggle with. Their AI doesn’t just generate text; it takes actions in simulated and physical environments.
Why video game data?
Video games provide massive amounts of action-labeled data — exactly what buttons were pressed and when — that teaches an AI about cause-and-effect. This is much richer than passive data like text or video, where you have to guess what the human intended.
Will military robots use this technology?
CEO Pim de Witte has drawn a clear ethical line: General Intuition will not build lethal autonomous weapons. The models can be used for search-and-rescue and humanitarian applications, but the company refuses military contracts targeting human harm.
When will the API be available?
General Intuition plans to make its API broadly available to developers by the end of summer 2026. Early access is already available for select partners in gaming, simulation, and robotics.
Source: TechCrunch — “General Intuition’s $2.3B bet that video games can train AI agents for the real world” (June 25, 2026)