XDOF, a startup collecting real-world teleoperation data for general-purpose robots, is in late-stage talks to raise a Series B at a valuation of about $1.2 billion. This comes less than three months after the company emerged from stealth mode, according to several sources familiar with the deal.
The speed of the funding
Co-founded in 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, the firm raised $70 million in a Series A last June. That round included participation from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. The company had no intention of raising again so soon. However, revenue is approaching $50 million annually, a rate that prompted venture capitalists to approach them for a new round.
Details on the total capital being raised remain unclear. It is not known whether the $1.2 billion figure includes the new funding. The terms are not final and could change.
Neither XDOF nor 8VC responded to requests for comment.
Building the supply chain
The startup builds data pipelines, collection tools, and annotation systems that frontier AI labs and robotics firms cannot easily construct themselves. It effectively acts as an outsourced data-supply chain for the robotics industry.
Wu was a PhD student studying how robots learn from large datasets. He told TechCrunch in June that a major impediment to his research was the lack of large-scale data to work with.
He teamed up with Shentu on a project called GELLO. This low-cost teleoperation system allows a human operator to control a robotic arm remotely to generate training data. Their work produced an influential paper in robotics.
That research formed the foundation for XDOF. Investors now describe the firm as the Scale AI or Mercor for physical robotics. These data-labeling giants helped fuel the AI boom. Unlike large language models, which initially trained on the entirety of the internet, physical robots lack an equivalent real-world dataset. This makes data collection a critical bottleneck for 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. This dataset is dubbed ABC.
To capture this data, the company combines remote robot teleoperation with human collectors. These workers wear sensors to record everyday tasks like folding clothes and flattening boxes.
The startup plans to hire and train teams of data collectors worldwide. This includes teleoperators who steer robots remotely and egocentric operators who wear body sensors to capture movement data.
XDOF 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. There are also human-data platforms expanding beyond LLMs, such as Scale AI and Micro1.
What it means
This funding validates a specific bottleneck in the industry. While models for text and code have been trained on vast digital archives, physical robots require physical interaction to learn. XDOF is solving this by industrialising the collection of that physical experience, turning a manual process into a scalable service for companies building the next generation of machines.
