General Intuition has raised $220 million at a valuation of $6.2 billion. As recently as June 2026, when the company closed a $320 million Series A, that valuation stood at $2.3 billion. In less than three months, the paper value of the Netherlands-founded AI company has nearly tripled.
The round was led by Valor Equity Partners and Atreides, with additional participation from 776, Point72, Khosla Ventures and General Catalyst. Khosla and General Catalyst had already participated in the earlier Series A. In total, General Intuition has raised more than $650 million since October 2025.
The company is simultaneously announcing that its models will become available to external partners. Until now, it has worked with a small group of companies in gaming, simulation and robotics through a closed API. With the new round, it intends to broaden access and expand its team, both in New York and in Europe.
Why the valuation rose so quickly
The acceleration is driven by a combination of data growth and forthcoming product availability. General Intuition trains its models on action-labelled gameplay videos from Medal, the game-clip platform through which the company was created in 2025 by CEO Pim de Witte. In August 2026, Medal passed the two-billion mark for cumulative uploaded clips, with approximately 195 million new clips being uploaded per month. The company is on track for three billion uploads per year.
For comparison: the frontier robotics models that General Intuition references train on less than one percent of the volume of action data the company has at its disposal. That difference in scale is central to the investment thesis: more diverse, action-linked video content should lead to models that handle unexpected situations more effectively, both in simulation and in the physical world.
In 2025, the company already generated approximately $40 million in revenue from early-access agreements with partners, which is substantial by early deeptech standards. The company turned down an acquisition offer from OpenAI for Medal's data, reported at $500 million, in 2024. General Intuition is legally registered in the Netherlands and operates as a public-benefit corporation.
What the models do
General Intuition builds two types of models. Action models determine which action an agent performs; world models predict the consequences of those actions. The underlying idea is that people control video games and robots through the same inputs and outputs, visual perception and controller actions. By training on these at scale, in a manner comparable to the approach Tesla used for its Full Self-Driving system but across a broader range of actions and environments, General Intuition aims to build models with broad applicability, including outside the gaming domain.
A concrete example is MIRA, a playable multiplayer world model trained on ten thousand hours of Rocket League data, developed together with the French AI lab Kyutai and in collaboration with Epic Games. The model runs in real time at twenty frames per second based on keyboard input from multiple players. For robotics and autonomous systems, the reasoning is that a model that encounters rare or dangerous situations in simulation far more frequently than would ever be observable in the real world will be better equipped to handle such situations when they do occur.
For training infrastructure scalability, General Intuition works with CoreWeave. The co-founders alongside De Witte are Eloi Alonso, Adam Jelley and Vincent Micheli. The company has offices in New York, Geneva, Paris and London; its data and intellectual property are managed from Naarden.
Position in the European AI landscape
General Intuition's rise in valuation fits within a broader trend of European AI companies attracting larger rounds more quickly in 2026. Mistral AI closed a Series D of €3 billion at a valuation above €21 billion in early September, the largest equity financing round for a European tech company on record. General Intuition operates at a different layer, specifically the combination of proprietary training data and foundation models for physical AI, but is drawing investors from the same global pool of venture capital funds.
For founders and investors in the Netherlands, it is notable that the company maintains its legal base and data management in the Netherlands, even as its operations shift increasingly towards New York. This demonstrates that a Dutch corporate structure can remain viable even through rapid international scaling. Whether the announced expansion of the European team will translate into substantial growth of the Dutch office is not discernible from the information currently available.