Less than three months after leaving stealth, XDOF — a startup that collects real-world teleoperation data to train general-purpose robots — is in late-stage discussions to raise a Series B at an approximate $1.2 billion valuation, according to people familiar with the deal. The round is reported to be led by 8VC.
Founding and previous financing
XDOF was co-founded in 2024 by University of California, Berkeley researchers Philipp Wu (CEO) and Fred Shentu (CTO). TechCrunch reported that the company closed a $70 million Series A in June, with participation from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital.
Sources say XDOF did not originally plan to raise again so soon after the Series A, but rapid growth — with annualized revenue approaching $50 million — drew renewed interest from venture capital firms.
Terms and current status
The discussions are in late stages but the terms are not final and could change. It is unclear whether the roughly $1.2 billion valuation includes the new funding, and the total amount being raised was not disclosed. Neither XDOF nor 8VC responded to requests for comment.
What XDOF does and why it matters
XDOF builds data pipelines, collection tools, and annotation systems that frontier AI labs and robotics companies may struggle to construct in-house. The company positions itself as an outsourced data supply chain for the robotics industry.
As a PhD student, Philipp Wu investigated how robots learn from large datasets and encountered a major barrier: a lack of large-scale real-world data. Wu and Shentu collaborated on a project called GELLO, a low-cost teleoperation system that lets a human operator control a robotic arm remotely to generate training data. Their work produced an influential robotics paper and laid the groundwork for XDOF.
Investors compare XDOF to data-labeling firms such as Scale AI or Mercor for physical robotics: unlike large language models, which initially trained on massive internet corpora, physical robots lack an equivalent, comprehensive real-world dataset, making data collection a critical bottleneck for building general-purpose machines.
ABC dataset and data collection approach
XDOF is partnering with UC Berkeley’s AI Research lab to publish what it calls ABC, a collection the company believes is the largest high-quality robot training dataset ever assembled. To capture these data, XDOF combines remote robot teleoperation with human collectors who wear sensors to record everyday tasks like folding clothes and flattening boxes.
The startup plans to hire and train data collection teams globally, including teleoperators who steer robots remotely and egocentric operators who wear body sensors to capture movement data.
Customers and competitors
XDOF previously told TechCrunch it is already working with 20 customers, including several frontier AI labs. Other companies attempting to gather real-world robot training data include Mecka AI, and broader human-data platforms expanding beyond language models, such as Scale AI and Micro1.
Conclusion
XDOF’s fast revenue growth has prompted renewed investor interest and late-stage Series B talks at an estimated $1.2 billion valuation. The terms are not finalized, and the company continues to focus on scaling its data collection and annotation capabilities, including releasing the ABC dataset in partnership with UC Berkeley AI Research.
