Robotics Data Startup XDOF Nears $1.2B Round as Embodied AI Hits Data Wall

For the AI community, the XDOF story is less about the round and more about the data. LLMs scaled by training on the entirety of the internet, but a general-purpose robot has no equivalent corpus — every grasp, insertion, and manipulation sequence must be collected from the physical world. XDOF's answer, rooted in the GELLO low-cost teleoperation system its UC Berkeley founders created, is to industrialize that collection: building the pipelines, tools, and annotation systems that frontier labs can't easily stand up themselves. A ~$1.2 billion Series B on ~$50 million annualized revenue is the market's verdict that embodied AI's data bottleneck is the next great moat to own.

XDOF, a robotics data-infrastructure startup founded by two UC Berkeley researchers, is in late-stage talks to raise a Series B at a valuation of about $1.2 billion led by 8VC, according to several people with knowledge of the deal cited by TechCrunch on September 4, 2026. The development is remarkable on timing alone: the company emerged from stealth less than three months ago, and its $70 million Series A — backed by Thrive Capital, Andreessen Horowitz, Lux Capital, and Spark Capital — was only reported in June. XDOF had not planned to raise again so soon, the sources said, but its rapid revenue growth forced the issue: annualized revenue is approaching $50 million, and investors began approaching the company about a new round rather than the other way around.

“

The development is remarkable on timing alone: the company emerged from stealth less than three months ago, and its $70 million Series A — backed by Thrive Capital, Andreessen Horowitz, Lux Capital, and Spark Capital — was only reported in June.

The company occupies the picks-and-shovels layer of the embodied-AI build-out. XDOF builds data pipelines, collection tools, and annotation systems for training general-purpose robots — infrastructure that frontier AI labs and robotics companies cannot easily build themselves. Investors have taken to describing XDOF as "the Scale AI or Mercor for physical robotics," referencing the data-labeling giants that helped fuel the large-language-model boom. The analogy is instructive. LLMs trained on the entirety of the internet, but physical robots have no equivalent corpus of real-world interaction data. XDOF's founders encountered that gap directly: as a PhD student, CEO Philipp Wu studied how robots learn from large datasets and found the field starved for "large-scale data to work with." With CTO Fred Shentu, Wu built GELLO, a low-cost teleoperation system that lets a human operator control a robotic arm remotely to generate training data. That work produced an influential robotics paper and became the technical foundation of XDOF, which the pair founded in 2024.

The valuation math underscores how hot the robotics data category has become. At a $1.2 billion valuation against roughly $50 million of annualized revenue, XDOF would command a multiple of about 24 times revenue — a software-like premium for what is, in significant part, an operations-heavy data collection business. That premium reflects scarcity: there are few scaled suppliers of real-world teleoperation data, and demand from humanoid-robot and general-purpose-robotics developers is rising faster than the data supply chain can mature. The rapid re-rating from a $70 million Series A to a potential $1.2 billion Series B within a single quarter also signals a founder-friendly environment in which top-tier investors are willing to preempt rounds to secure allocation in category-defining companies.