Robotics data startup XDOF raises Series B at $1.2 billion valuation as physical AI demand surges
Three months after exiting stealth, Philipp Wu’s UC Berkeley spinout secures late-stage backing from 8VC to scale its teleoperation data supply chain.
Original cover art, generated for this story. THE VISSION does not republish third-party press imagery.
- Physical robotics data startup XDOF is in late-stage talks for a Series B funding round led by 8VC, valuing the company at $1.2 billion.
- The startup is approaching $50 million in annualized revenue from 20 customers, serving as a physical data supplier to leading frontier AI labs.
- Co-founded in 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, XDOF developed the low-cost GELLO arm teleoperation system.
Physical robotics data startup XDOF is in late-stage talks to raise a Series B funding round led by venture capital firm 8VC at a $1.2 billion valuation, according to reporting from TechCrunch. The round comes just three months after the startup officially exited stealth, reflecting the intense investor appetite for companies that can provide high-fidelity, real-world data to train general-purpose embodied AI and physical robotics systems.
Founded in 2024 by UC Berkeley researchers Philipp Wu (CEO) and Fred Shentu (CTO), XDOF operates as an outsourced data-supply chain for the physical robotics industry, drawing comparisons to what Scale AI or Mercor provide for language models. The startup has experienced explosive commercial growth, approaching $50 million in annualized revenue from its 20 customers, which include several of the world's leading frontier AI laboratories.
XDOF's rapid rise is built on its low-cost robotic teleoperation hardware and collaborative data collection. The startup is the creator of GELLO, an open-source, low-cost teleoperation system that allows humans to easily control robotic arms to capture training data. In partnership with UC Berkeley's AI Research lab, XDOF is also releasing the ABC dataset, a massive library of high-quality physical manipulation data designed to help robots learn general-purpose tasks like sorting, assembly, and packaging.
Embodied AI systems cannot learn physics from text or static images; they require large-scale, high-quality physical interaction data. XDOF's $1.2 billion valuation just three months out of stealth demonstrates that the robotics bottleneck has shifted from hardware design to data ingestion. By commercializing low-cost teleoperation like GELLO and building a massive physical dataset, XDOF is establishing itself as a critical utility provider for the next wave of physical AI developers.