XDOF nears Series B talks at about $1.2 billion valuation

XDOF enters late-stage Series B discussions

XDOF, a startup building real-world data infrastructure for general-purpose robot training, is in late-stage discussions to raise a Series B at a valuation of about $1.2 billion. Several people familiar with the proposed deal said 8VC is expected to lead the round. The terms are not final and may change.

The company emerged from stealth less than three months ago. It was founded in 2024 by UC Berkeley researchers Philipp Wu, its chief executive, and Fred Shentu, its chief technology officer. XDOF and 8VC did not respond to requests for comment.

TechCrunch reported in June that XDOF had raised a $70 million Series A with participation from Thrive Capital, Andreessen Horowitz, Lux and Spark Capital. The startup had not intended to seek another round so quickly, but investor interest followed rapid growth: its annualized revenue is approaching $50 million. The total amount sought in the new financing, and whether the reported valuation includes fresh capital, were not available.

A data-supply model for physical robotics

XDOF is developing data pipelines, collection tools and annotation systems for frontier AI laboratories and robotics companies. Its role is intended to address a constraint specific to physical AI: robots do not have an equivalent of the internet-scale corpus that helped fuel early large language model training.

The company’s origins lie in GELLO, a low-cost teleoperation system created by Wu and Shentu that lets a human operator remotely control a robotic arm to produce training data. That research became the basis for XDOF. Its first major financing, detailed in XDOF’s initial robotics data infrastructure funding, supported the company’s move into data infrastructure for training robots.

XDOF is also working with UC Berkeley’s AI Research lab on ABC, which it describes as the largest collection of high-quality robot training data assembled to date. The dataset is gathered through remote robot teleoperation and human collectors wearing sensors while performing everyday activities, including folding clothes and flattening boxes.

Operational scale becomes the test

The startup plans to hire and train data-collection teams globally. These include teleoperators who guide robots remotely and egocentric operators who wear body sensors to record movement. XDOF has said it already serves 20 customers, including several frontier AI labs.

Competition includes Mecka AI, as well as platforms such as Scale AI and Micro1 that are expanding beyond language-model data. For businesses developing robotics products, the proposed financing underlines the practical importance of assessing how training data is collected, annotated and scaled alongside the robot model itself.