XDOF, a startup that collects real-world data to train versatile robots, is in advanced talks to raise a Series B funding round at a valuation of approximately $1.2 billion, according to people familiar with the deal. 8VC is expected to lead the round, but its terms have not yet been finalized and could change before closing.
This comes less than three months after the company emerged from stealth and after a $70 million Series A round announced in June, with participation from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. The total amount targeted in the new round is unclear, as is whether the reported valuation includes the new capital.
Rapid Growth After the First Funding Round
XDOF was founded in 2024 by Philipp Wu, who serves as chief executive officer, and Fred Shentu, chief technology officer, both UC Berkeley researchers. According to the source, the company had not planned to raise new funding so quickly, but its growth prompted investors to reach out. Its annualized revenue run rate is approaching $50 million, and it is already working with 20 customers, including leading AI labs.
XDOF and 8VC did not respond to requests for comment, so there is no direct confirmation from either party regarding the round’s value or timeline.
Robot Data as a Bottleneck
The company aims to build data-processing pipelines, collection tools, and labeling systems for AI labs and robotics companies that may find it difficult to develop this infrastructure internally. The process includes remotely operating robots by people, along with using data collectors wearing sensors to record everyday tasks such as folding clothes and flattening boxes.
The idea originated with Project GELLO, a low-cost system that enables a person to remotely control a robotic arm to generate training data. The project was based on Wu and Shentu’s research into robot learning from large datasets, research that led to an influential paper in the field of robotics.
What Is Changing in Practice?
Unlike large language models, which initially benefited from massive quantities of internet content, physical robots do not have a comparable supply of ready-made real-world data for training. XDOF is therefore focusing on building a data supply chain that includes collection, remote operation, and annotation, rather than merely developing a single robotic model.
The company plans to hire and train teams of data collectors around the world, including remote robot operators and operators wearing body sensors to record motion from a first-person perspective. It is also collaborating with the UC Berkeley AI lab to launch a dataset called ABC, which it describes as the largest collection of high-quality robot training data gathered to date.
The significance of the round remains contingent on its actually closing and on XDOF’s ability to turn growing demand into usable data at scale. The funding amount and valuation terms, both essential to understanding the size of the deal, have also not yet been disclosed.