XDOF in Late-Stage Talks for $1.2 Billion Series B Led by 8VC, TechCrunch Says

TechCrunch was unable to learn the total amount being raised or whether the valuation includes the new funding. Terms are not final and could still change. XDOF and 8VC did not respond to requests for comment.

XDOF was founded in 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, who serve as CEO and CTO, respectively. In June, TechCrunch reported on the startup's $70 million Series A, with participation from Thrive Capital, Andreessen Horowitz, Lux and Spark Capital.

The people said XDOF had not planned to raise again so soon after that round, but its rapid growth, with annualized revenue approaching $50 million, prompted venture investors to approach the company. XDOF aims to build data pipelines, collection tools and annotation systems that frontier AI labs and robotics companies cannot easily build themselves, effectively acting as an outsourced data-supply chain for the robotics industry.

The startup came out of research by Wu and Shentu on GELLO, a low-cost teleoperation system that lets a human operator remotely control a robotic arm to generate training data. Wu told TechCrunch in June that, as a PhD student studying how robots learn from large datasets, a major obstacle was the lack of large-scale data to work with. The research led to an influential paper in robotics.

Investors describe XDOF as the Scale AI or Mercor for physical robotics, a parallel to the data-labeling companies that helped fuel the AI boom. Unlike large language models, which were initially trained on large parts of the internet, physical robots do not have an equivalent real-world dataset to draw from, making data collection a critical bottleneck in building general-purpose machines.

XDOF is partnering with UC Berkeley's AI Research lab to release what it believes is the largest collection of high-quality robot training data ever assembled, called ABC. To capture data, XDOF combines remote robot teleoperation with human collectors who wear sensors to record everyday tasks such as folding clothes and flattening boxes. It plans to hire and train data-collection teams worldwide, including teleoperators who steer robots remotely and egocentric operators who wear body sensors to capture movement data.

The startup previously told TechCrunch that it is already working with 20 customers, including several frontier AI labs. Other startups attempting to collect real-world data for robot training include Mecka AI, as well as human-data platforms expanding beyond large language models, such as Scale AI and Micro1.