- XDOF is in late-stage talks for a Series B at roughly $1.2 billion, reportedly led by 8VC
- It closed a $70M Series A just three months ago, backed by Thrive Capital, Andreessen Horowitz, Lux, Spark Capital and WndrCo
- The San Mateo startup builds the data pipelines and teleoperation tools that teach general-purpose robots to work in the physical world
XDOF is in late-stage talks to raise a Series B at a valuation of about $1.2 billion, with 8VC reportedly set to lead the round, according to people familiar with the deal cited by TechCrunch.
What’s clear is the speed. XDOF only came out of stealth on June 17, 2026, after closing a $70 million Series A. It wasn’t planning to raise again this soon, but annualized revenue approaching $50 million pushed investors to come knocking, TechCrunch reported.
The Series A brought in Thrive Capital, Spark Capital, Andreessen Horowitz, Lux Capital and WndrCo, taking XDOF’s total funding to roughly $78 million once its earlier seed round is counted, according to PitchBook.
What XDOF actually does
XDOF is effectively a data supply chain for robotics. Frontier AI labs know how to build language models; almost none of them have built the infrastructure to collect physical-world training data at scale, and XDOF is betting they’d rather pay someone else to do it.
Its platform combines teleoperation tools, human data collectors and annotation systems: some people remotely pilot robotic arms to demonstrate tasks, others wear body sensors to record how humans move through everyday chores like folding laundry.
The idea traces back to a data problem co-founder and CEO Philipp Wu hit as a PhD student at UC Berkeley. He and co-founder Fred Shentu built GELLO, a low-cost teleoperation rig, to solve it; the resulting paper became widely used across the field. Wu, Shentu and a third co-founder, Nemo Jin, launched XDOF in October 2024.
The dataset behind the pitch
XDOF’s clearest calling card so far is ABC-130K, a dataset it released with UC Berkeley’s AI Research Lab: over 130,000 episodes across 195 bimanual manipulation tasks, according to XDOF’s own write-up.
The company plans to keep expanding its network of human operators globally, split between teleoperators who steer robots remotely and “egocentric” collectors who wear sensors. It has around 60 employees and about 20 customers, including several frontier AI labs it won’t name.
The race towards physical AI
Investors are reportedly describing XDOF as the Scale AI or Mercor of physical robotics. Scale AI raised $1 billion at a $13.8 billion valuation in May 2024, and Mercor went from $100 million at a $2 billion valuation in February 2025 to a reported $10 billion by October, but both built their businesses sourcing human expertise for digital-world AI.
XDOF, along with rivals like Mecka AI and Encord, which raised $60 million to chase the same physical-AI data problem, is trying to solve something harder: there’s no internet-scale archive of the physical world to scrape. Someone has to go generate it by hand.
That’s the vision underneath the valuation talk. LLMs got lucky — the training data already existed. Robotics companies don’t have that luxury, and the humanoid robotics field is already absorbing billions on that assumption.
Whether XDOF’s data-collection model scales as fast as the checks investors are reportedly willing to write is the open question nobody, including XDOF and 8VC, who aren’t talking yet, has answered.
