Scaling AI for Robotics
AI Startup General Intuition Targets $6 Billion Valuation to Advance Physical Robotics
Key Points
Medal-derived data is powering General Intuition's models: the dataset includes hundreds of millions of hours of gameplay and action labels from roughly 17 million monthly Medal users generating billions of clips annually, with a trained model showing continuous gameplay for over 100 hours and needing as little as 8 minutes of real-world data to adapt to physical navigation tasks.
The company is investing in compute infrastructure with a CoreWeave partnership to scale training for large action foundation models.
The funding round is reported as oversubscribed and targeting a $6 billion pre-money valuation, following a $320 million Series A at a $2.3 billion valuation; new investors include Valor Equity Partners, Point72 Ventures, and Seven Seven Six.
Valor Equity Partners, known for backing SpaceX, would be making its first AI-lab investment since SpaceX, underscoring the deal's significance for anchoring an AI hardware-to-robotics effort.
General Intuition, an AI startup training robots to understand the physical world, is closing in on a $6 billion pre-money funding round PYMNTS. The New York-based company's new backers include Valor Equity Partners, Point72 Ventures, and Seven Seven Six, alongside existing investors Khosla Ventures and General Catalyst TechCrunch. The capital will fund expanded foundation models for robotics and a partnership with compute provider CoreWeave to scale training infrastructure.
This valuation leap—from $2.3 billion just months ago—reflects booming investor interest in "physical AI" systems that learn from real-world data Crypto Briefing. General Intuition trains AI agents using hundreds of millions of hours of gameplay footage from Medal, a gaming clip platform the company spun out from, to teach machines how to move, perceive, and act in actual environments.
Medal Data Powers AI Training at Scale
General Intuition's secret weapon is Medal's dataset: roughly 17 million monthly users generate billions of gaming clips annually TechBuzz. The startup trained one model to play continuously for over 100 hours straight. More impressively, it needs just 8 minutes of real-world video to adapt that same model to physical navigation tasks, showing how gameplay data transfers to robotics.
From Gaming Clips to Real Robots in Motion
General Intuition was founded by Pim de Witte, who previously built Medal. The company extracts action labels from hundreds of millions of hours of gameplay—what players do, how they react, how they navigate obstacles. This "action foundation model" teaches AI agents to generalize across tasks Yahoo Finance. The approach works because human gamers already demonstrate how to solve problems in complex, dynamic environments.
Valor's First Major AI Lab Bet Since SpaceX
Valor Equity Partners, known for backing SpaceX, marks this as its first significant investment in an AI lab since the rocket company TechCrunch. The signal carries weight: Valor typically backs moonshot hardware-to-software ventures. Its participation underscores investor confidence that General Intuition can bridge AI models with physical robotics—the next frontier after language models.
Compute Infrastructure as the Bottleneck
Training massive "action foundation models" demands enormous compute power. General Intuition's CoreWeave partnership addresses this: CoreWeave provides specialized GPU clusters optimized for large-scale AI training TechBuzz. This infrastructure play mirrors how companies like OpenAI scaled language models—secure reliable compute first, then push model size and capability forward. The $6 billion round funds both talent and silicon.
Sentiment
Mostly PositivePublishers
10
Articles
5
Reach
15
