XDOF, a startup building data pipelines and annotation systems for robot training, is in late-stage discussions to raise a Series B round valuing the company at approximately $1.2 billion, led by venture capital firm 8VC.
The funding talks come less than three months after XDOF publicly launched. The company was co-founded in 2024 by UC Berkeley researchers Philipp Wu, CEO, and Fred Shentu, CTO.
XDOF's Series A, which closed in June for $70 million, included Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. Company management had not planned to raise again so quickly, but annualized revenue approaching $50 million prompted venture capitalists to initiate discussions for the new financing.
The core economics are straightforward: frontier AI labs and robotics companies need large-scale training datasets but lack efficient internal systems to generate them. XDOF functions as an outsourced data-supply chain for the sector.
Unlike large language models, which trained on billions of internet-sourced documents, physical robots face a hard constraint — there is no publicly available corpus of real-world manipulation data at scale. The gap between what robotics companies need and what exists is the business opportunity.
Wu's PhD research at UC Berkeley identified the problem. He and Shentu built GELLO, a low-cost teleoperation system that lets human operators remotely control robotic arms to generate training data. Their subsequent paper became influential in robotics circles.
The company's go-to-market combines two data-collection methods. Remote teleoperators steer robots in real time. Separately, human collectors wearing body sensors record everyday tasks — folding clothes, flattening boxes — to capture egocentric movement patterns.
XDOF is also partnering with UC Berkeley's AI Research Lab to release a dataset called ABC, which it describes as the largest assembly of high-quality robot training data compiled to date.
The startup currently serves 20 customers, primarily frontier AI laboratories. It plans to expand by hiring and training data-collection teams globally.
Investors draw parallels to Scale AI and Mercor, which supplied the data-labeling infrastructure that fueled the large-language-model wave. The robotics sector faces an analogous bottleneck, with capital flowing to whoever can industrialize data supply.
Other startups pursuing similar strategies include Mecka AI and human-data platforms like expa. The specific size of XDOF's Series B round and whether the $1.2 billion valuation reflects the new capital remain undisclosed. XDOF and 8VC did not respond to requests for comment.
