Robot training-data startup XDOF nears $1.2B valuation just three months after leaving stealth
XDOF, a startup that collects real-world data to train general-purpose robots, is in late-stage talks for a new funding round at roughly a $1.2 billion valuation led by 8VC, less than three months after its Series A.

XDOF, a startup focused on collecting real-world teleoperation data for training general-purpose robots, is reportedly in late-stage talks to raise a Series B round at a valuation of about $1.2 billion. The round would be led by venture firm 8VC, according to people familiar with the matter.
The company was co-founded in 2024 by UC Berkeley researchers Philipp Wu, who serves as CEO, and Fred Shentu, the CTO. Just three months ago, XDOF closed a $70 million Series A round with participation from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital.
XDOF reportedly hadn't planned to raise again so soon, but rapid growth — with annualized revenue nearing $50 million — drew renewed investor interest. The total amount being raised and whether the valuation includes new capital remain unclear, and terms of the deal are not yet finalized.
From research project to startup
The company's roots trace back to Wu's PhD research on how robots learn from large datasets, which was hampered by a lack of large-scale training data. Together with Shentu, he built GELLO, a low-cost teleoperation system that lets a human remotely control a robotic arm to generate training data. That work led to an influential research paper and eventually became the foundation for XDOF.
Investors reportedly describe XDOF as a version of Scale AI or Mercor for physical robotics. Unlike large language models, which were trained on vast internet data, robots lack an equivalent real-world dataset, making data collection a major bottleneck for the industry.
XDOF is partnering with UC Berkeley's AI Research lab on a project called ABC, aiming to build what it calls the largest high-quality robot training dataset assembled to date. Data is gathered through remote robot teleoperation as well as human collectors wearing sensors while performing everyday tasks such as folding clothes. The company plans to build teams of data collectors globally and already counts 20 customers, including several frontier AI labs.
/nginx/o/2026/09/03/17895100t1h0ff3.png)

