XDOF: The Robotics Data Startup Racing to a $1.2B Valuation

Three months out of stealth, the teleoperation-data startup is already in late-stage Series B talks — because physical robots have no internet to train on.

Published: 2026-09-05 Category: Quick Take Sources: TechCrunch

The Data Bottleneck Nobody Solved

Large language models had a gift the robotics industry never got: the entire internet. GPT-class systems trained on the accumulated text of humanity, which is why they scaled so fast. Physical robots have no equivalent. There is no "internet of folding laundry" or "corpus of flattening boxes" sitting around waiting to be scraped. That gap is the entire thesis of XDOF, a startup co-founded by UC Berkeley researchers Philipp Wu (CEO) and Fred Shentu (CTO) in 2024.

The company collects real-world teleoperation data for training general-purpose robots, and investors are treating it like the missing piece of the physical-AI puzzle. TechCrunch reports XDOF is in late-stage talks for a Series B at roughly a $1.2 billion valuation, led by 8VC — less than three months after emerging from stealth and hot on the heels of a $70 million Series A in June that included Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital.

Why the Speed Matters

The velocity is the story. XDOF wasn't planning to raise again so soon, but annualized revenue approaching $50 million changed the calculus — VCs came to them. That kind of growth in a data-supply business signals that frontier labs and robotics companies are desperate for exactly what XDOF sells: the pipelines, collection tools, and annotation systems they can't easily build themselves.

Investors describe XDOF as "the Scale AI or Mercor for physical robotics" — a reference to the data-labeling giants that fueled the LLM boom. The comparison is apt. Scale AI became a giant by industrializing the grunt work of AI training. XDOF is betting the same playbook works when the data is physical motion rather than text.

The GELLO Origin Story

The company's roots are academic. As a PhD student, Wu studied how robots learn from large datasets and hit the same wall everyone does: not enough data. He teamed up with Shentu on GELLO, a low-cost teleoperation system letting a human operator control a robotic arm remotely to generate training data. That work produced an influential robotics paper and became the foundation for XDOF.

The approach combines remote robot teleoperation with human collectors wearing sensors to record everyday tasks — folding clothes, flattening boxes. It's unglamorous, labor-intensive work, which is precisely why it's valuable. XDOF is also 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, dubbed ABC.

The Takeaway

XDOF's trajectory is a reminder that the AI boom's real bottleneck is increasingly boring infrastructure, not clever algorithms. Every frontier lab wants general-purpose robots; almost none want to build the data-supply chain to get there. If XDOF closes this round at $1.2B, it will be one of the fastest ascents in robotics — and a signal that the "picks and shovels" of physical AI are now as valuable as the models themselves.

Source: TechCrunch, "XDOF, just three months out of stealth, is in talks for a Series B at a $1.2B valuation" (September 4, 2026).