From Fortnite to robots: General Intuition Secures $2.3B Funding to Advance AI Trained via Video Games

General Intuition announced on Thursday that it has raised $320 million in a funding round, bringing its valuation to $2.3 billion. This latest investment increases the company's total disclosed funding to $454 million, following a $134 million round at its launch last October.

The New York-based startup focuses on developing AI agents capable of generalizing from video gameplay to real-world robotics. The company’s co-founder and CEO, Pim de Witte, explained that their AI model, initially trained on hundreds of millions of hours of gameplay data collected through Medal—a platform for sharing video game clips—can navigate both virtual and physical environments.

Medal’s data includes not only gameplay footage but also detailed action labels indicating the exact buttons pressed by players, a feature that distinguishes General Intuition’s approach from competitors who attempt to infer actions from video alone. De Witte emphasized that this embedded action data is crucial for the model’s spatial-temporal reasoning and its ability to understand causality.

During a demonstration, the AI agent was shown playing Fortnite for over 100 hours continuously. The same AI system also powered a large quadrupedal robot navigating an office space, using a single camera to explore its surroundings. According to data analyst Josh Duplantis, it required only eight minutes of real-world robotics data collected on the street to fine-tune the model for the robot’s operation.

General Intuition’s world model generates simulated environments frame-by-frame rather than relying on traditional game engines. This simulation enables the AI to learn physical constraints, such as walls being impassable and ladders being climbable, based on the gameplay data. The company views this simulated environment as a training ground—referred to internally as “the gym”—rather than a product itself.

The startup’s recent funding round was led by Khosla Ventures, with participation from General Catalyst, Jeff Bezos, Eric Schmidt, Nico Rosberg, and researchers from Google DeepMind and MIT. Most of the new capital will be allocated to expanding compute capacity, with plans to pre-train the next model iteration and broaden access to its API by the end of summer through a partnership with CoreWeave.

Vinod Khosla, head of the leading venture firm, highlighted the significance of the human action data derived from games, describing it as essential to the emergence of AI intuition—an ability akin to human intuition. He contrasted this with the development of large language models, where reasoning capabilities marked a quantum leap.

General Intuition originated from Medal, which was declined acquisition offers from major labs. The founders—Pim de Witte, Eloi Alonso, Adam Jelley, and Vincent Micheli—along with their investors, are focused on building a long-term company rather than seeking an exit. Khosla underscored the value of the proprietary data as a reason to consider General Intuition a generational investment rather than a typical acquisition target.

De Witte, who previously worked in humanitarian roles including with Doctors Without Borders, has set ethical boundaries for the company’s technology, explicitly ruling out its use in lethal autonomous systems. He expressed concern about Silicon Valley’s increasing interest in defense applications but welcomed uses such as search and rescue missions.

The company’s European-influenced culture is reflected in its team composition and values. Chief of Staff Brianna Martin joined partly due to her public resignation from Palantir over its collaboration with U.S. Immigration and Customs Enforcement. De Witte remarked on his distance from Silicon Valley’s prevailing trends.

Addressing the impact of AI on employment, General Intuition recently launched Nerve, a jobs marketplace enabling gamers to monetize their existing setups. Participants can start with data labeling and progress to roles like robot teleoperation. De Witte noted that Medal’s user base is among those most vulnerable to AI-driven job displacement, and he aims to provide them with opportunities in the evolving landscape.

General Intuition positions itself as an ecosystem enabler similar to Anthropic or OpenAI, offering foundational models that others can build upon. The startup currently serves a handful of customers across gaming, simulation, and robotics sectors. De Witte stated that the company does not intend to build applications like self-driving cars but aims to simplify such developments for others.

Once its API is more widely available, General Intuition plans to test various use cases, including robots operating in digital factory twins, human-like bots in gaming studios, and quadrupeds navigating hazardous environments. Besides quadrupeds, the company has experimented with drones and driving games, with its AI model adaptable to any device controllable via game controllers or keyboard and mouse.

De Witte highlighted the goal of creating a data flywheel by selecting customers who can provide diverse real-world data and maintain agile internal teams for collaborative development. Khosla emphasized that the startup’s proprietary data has been key to its progress and will remain critical, especially since the challenge of scaling simulation-to-real-world transfer remains unresolved.