AI Startup General Intuition Hits $6.2 Billion Valuation After New Funding
Nijmegen, Friday, 2 October 2026.
Dutch entrepreneur Pim de Witte’s startup, General Intuition, secured $220 million, tripling its valuation to $6.2 billion by training advanced robotics AI using billions of video game clips.
A Massive Capital Injection in Silicon Valley’s AI Race
The New York-based artificial intelligence lab General Intuition, co-founded by Dutch entrepreneur Pim de Witte and Moritz Baier-Lentz, has successfully raised $220 million in its latest funding round [6]. This significant cash injection has propelled the company’s valuation to $6.2 billion, nearly tripling its previous valuation of $2.3 billion secured during its Series A round just months prior [6]. The funding round attracted a prominent syndicate of global investors, including Valor Equity Partners, Atreides Management, Seven Seven Six, Point72, Khosla Ventures, and General Catalyst [3][6]. This rapid valuation growth represents a percentage increase of 169.565% over its previous valuation [6].
Resisting Early Acquisition to Maintain Autonomy
With this latest round, General Intuition’s total equity funding has reached $673.7 million [6]. The startup’s current multi-billion-dollar status vindicates a bold decision made by De Witte in late 2025, when he rejected a €500 million acquisition offer [1]. Reflecting on his choice to decline the half-billion-euro buyout, De Witte noted that accepting such an acquisition would mean giving up his operational freedom [1]. By maintaining independence, the Nijmegen-born founder has kept the liberty to steer the company’s unique research direction [1].
Leveraging Gaming Data for Physical Autonomy
General Intuition distinguishes itself from other AI labs by training its robotics and world models on billions of action-labeled video clips and controller data [1][3]. This massive data pipeline is fed directly by Medal.tv, a game-clip sharing platform co-founded by De Witte that is currently on track to reach 3 billion uploaded videos per year [1][3][5]. Because of this integration, General Intuition has access to a dataset that is orders of magnitude larger than those of competing robotics foundation models, which are often trained on less than 1% of the action data General Intuition utilizes [3].
Training in Pixel Space for Real-World Generalization
The core methodology relies on training AI models in “pixel space” to enable deep spatiotemporal reasoning [2][4]. De Witte compares this approach to how Tesla trained its Full Self-Driving system, but scaled across a far wider variety of environments and possible actions [3]. Humans interact with both video games and physical robots using the same fundamental inputs and outputs—namely visual feedback and physical controller actions [3]. By training directly on this shared paradigm, General Intuition’s models develop a generalized capability to execute autonomous tasks in diverse, real-world scenarios [1][3].
From Virtual Simulations to Surgical Robotics
The scale of General Intuition’s simulated training allows its systems to encounter and learn from edge cases at an unprecedented rate. On any given day, the company’s simulations record more accidents than occur in reality across the entire United States [1][3]. This intensive exposure allows General Agents and World Models to successfully navigate and act in real time within complex physical environments they have never previously encountered [1][3]. While the startup currently sells its AI models to factories and drone manufacturers, De Witte maintains that profit is not the sole driver [1].
Expanding Global Footprint and Research Teams
Looking forward, De Witte aims to bridge the gap between AI, physics, and medical science [1]. The ultimate long-term goal for General Intuition is to integrate its adaptive AI models into medical robotics, potentially enabling machines to perform complex surgeries by determining patient-specific actions in real time [1]. To support these ambitious objectives, the company is actively scaling its engineering and research teams across its offices in New York and Europe [3][6]. Meanwhile, the company has opened a waitlist for commercial partners seeking early access to its autonomous models [1][3], preparing for a rapidly changing technological landscape over the next five years [1].