A robot data startup is closing in on unicorn status remarkably fast. XDOF is in late-stage talks to raise a Series B at a valuation of roughly $1.2 billion. Notably, this comes less than three months after the company first emerged from stealth. Venture firm 8VC is leading the round, according to people familiar with the deal.
The startup collects real-world teleoperation data used to train general-purpose robots. UC Berkeley researchers Philipp Wu and Fred Shentu co-founded XDOF in 2024. Wu serves as CEO while Shentu acts as CTO. TechCrunch previously reported on the company’s $70 million Series A back in June.
That earlier round drew notable backers including Thrive Capital and Andreessen Horowitz. Importantly, XDOF was not planning to raise again so soon. However, the company’s rapid growth changed those plans quickly. With annualized revenue approaching $50 million, investors approached XDOF about a new round themselves.
The startup addresses a critical bottleneck in modern robotics. Unlike large language models, physical robots lack an equivalent internet-scale dataset. As a result, data collection remains a major obstacle to building capable machines. XDOF builds the data pipelines and annotation systems that frontier labs cannot easily create alone.
Investors now describe XDOF as the Scale AI for physical robotics. This references the data-labeling giants that helped fuel the broader AI boom. Essentially, XDOF acts as an outsourced data-supply chain for the entire robotics industry.
The company’s origins trace back to Wu’s PhD research. He struggled with a lack of large-scale data for his work. So he teamed up with Shentu on a low-cost teleoperation system called GELLO. That project let human operators control robotic arms remotely to generate training data.
As they put it themselves:
Our mission is to unlock a world with abundant, useful physical intelligence.
We partner closely with the world’s leading robotics labs and most promising robotics companies to deliver production-scale datasets, advanced robotic systems, and the tooling needed to support the next generation of physical AI.
We believe strongly in full-stack expertise – from hardware to operations to policy training. Our co-founders authored GELLO, now commonly deployed for teleoperation. Our multi-disciplinary DNA allows us to deeply understand our customers, and we’re able to bring new offerings that push their thinking.
To capture this data, XDOF combines remote teleoperation with human collectors wearing sensors. These collectors record everyday tasks like folding clothes and flattening boxes. The startup plans to hire and train collection teams worldwide.
XDOF is already working with 20 customers, including several frontier AI labs. However, the terms of the new deal are not final and could still change. Other startups pursuing similar robot training data include Mecka AI, alongside expanding players like Scale AI and Micro1.
