Valor Equity Partners, Point72 Ventures and Seven Seven Six are among the new investors in a proposed round that would fund robotics, compute and hiring.
By RuntimeWire Staff · Published
Primary source: TechCrunch
Why it matters #
The proposed round would value de Witte's gameplay-data thesis at more than twice June's mark before broad robot deployment has been demonstrated. The investors involved appear willing to price the proprietary data and research team aggressively while commercial-scale transfer remains unproven.
Pim de Witte (@PimDeWitte) is negotiating another financing for General Intuition, this time at a reported $6 billion pre-money valuation, TechCrunch reported on August 24.
The talks bring in Valor Equity Partners, Point72 Ventures and Seven Seven Six as new investors. Existing backers Khosla Ventures and General Catalyst are also participating, according to the report. General Intuition is still finalizing the financing, and TechCrunch cited a source close to the deal who described it as oversubscribed. The proposed investment amount was not reported.
The discussions surfaced exactly 60 days after General Intuition announced a $320 million Series A at a $2.3 billion valuation. RuntimeWire reported in June that the round brought General Intuition's disclosed funding to $454 million, following a $134 million seed in October 2025. A financing at the newly reported terms would more than double General Intuition's headline valuation in two months.
That fundraising pace amounts to an unusually fast repricing of de Witte's central bet: the most useful training data for physical AI may already be sitting inside video games.
A founder's data bet, repriced in 60 days
De Witte built the asset behind General Intuition over years rather than buying it after the robotics pitch became fashionable. As a teenager in the Netherlands, de Witte built a private RuneScape server that reportedly generated more than $1 million in annual revenue. He later worked in humanitarian programs, including with Doctors Without Borders, before founding Medal, a platform for recording and sharing gameplay clips.
General Intuition spun out of Medal in October 2025. Its initial advantage comes from pairing Medal's gameplay footage with "action labels," the records of which buttons players pressed and when. Video shows what happened on screen. The associated inputs provide examples of decisions, timing and consequences, giving General Intuition training material for models designed to perceive an environment, anticipate changes and choose an action.
At the October spinout, TechCrunch reported that Medal had about 10 million monthly active users. General Intuition now describes its underlying archive as hundreds of millions of hours of gameplay.
De Witte assembled a research-heavy founding group around that data. Eloi Alonso studied reinforcement learning and world models during his Ph.D. at the University of Geneva. Adam Jelley is a Ph.D. candidate at the University of Edinburgh researching efficient reinforcement learning and world models, after initial doctoral training at University College London. Vincent Micheli worked on generative models for decision-making and previously held research positions at Wayve and Microsoft Research.
The combination matters. Medal supplies interactive data at consumer-platform scale, while Alonso, Jelley and Micheli bring research that predates the current rush to fund physical AI. Their earlier work includes IRIS, Delta-IRIS and DIAMOND, a sequence of papers exploring how agents can learn and act inside generated environments.
The new money has a specific job
TechCrunch reported that General Intuition intends to spend the proposed financing on improving its general model for robotic embodiments, expanding compute infrastructure and hiring. General Intuition already works with CoreWeave for computing capacity.
The August financing would extend the plan attached to June's Series A. General Intuition said then that most of the $320 million would fund compute and pre-training, with part reserved for expanding access to its commercial API. General Intuition's website now says it has onboarded initial partners in gaming, simulation and robotics and is selecting a limited group ahead of a broader model release. General Intuition has not named those partners publicly.
De Witte is pursuing a hardware-agnostic model business. General Intuition plans to provide the intelligence layer to customers building games, simulations and machines, rather than manufacture its own fleet of robots. Each customer could also provide new interaction data from a different device or environment, feeding future model training.
The early robotics demonstration shows the promise and the distance still to cover. In June, the same underlying model controlled a game-playing agent and a quadruped robot inside General Intuition's New York office. A General Intuition employee told TechCrunch that the robot required eight minutes of real-world robotics data for fine-tuning. During the demonstration, it still clipped chair legs and bumped into bins.
Eight minutes is a compelling demo result. It does not establish that gameplay-trained models can reliably control different machines through long, safety-critical tasks. The reported $6 billion valuation prices General Intuition well ahead of that proof.
Physical AI is consuming venture capital
General Intuition is raising into a physical AI market where training data, computing capacity and access to real machines have become expensive strategic assets. Skild AI announced a $1.4 billion Series C in January at a valuation above $14 billion for a foundation model intended to control multiple types of robots. World Labs raised $1 billion in February for spatial intelligence models and acquired robotics developer SceniX in July.
General Intuition's approach gives investors a different entry point. De Witte already controls a large stream of human interaction data through Medal, reducing General Intuition's dependence on collecting every training example from physical robots. The approach could lower the cost of teaching models general behaviors if virtual skills transfer cleanly into real environments.
Valor, Point72 and Seven Seven Six are being asked to underwrite that transfer before it has been demonstrated at commercial scale. The proposed financing suggests that the investors involved are willing to price a proprietary data source and research-heavy team aggressively, even before commercial-scale transfer has been demonstrated. General Intuition raised its seed less than a year ago, completed its Series A in June and is already discussing a $6 billion pre-money valuation.
The capital can buy more compute, researchers and robot experiments. General Intuition's harder task is turning Medal's record of how gamers act into a dependable model for how machines should act when the walls, obstacles and consequences are real.