The archive · AI & Models · Strategic decision · 2024–2026
XDOF's robot-data bet: $70M Series A, then Series B talks at $1.2B in three months
XDOF supplies the teleoperation data physical AI lacks — 20 customers, $70M raised, and annualized revenue approaching $50M within months of stealth exit.
XDOF
What the business is
XDOF (pronounced ecks-doff, a play on degrees of freedom) is a robot-training-data company: it collects real-world teleoperation and egocentric sensor data, builds the pipelines, and annotates it so frontier AI labs and robotics companies can train general-purpose physical models without running their own data factories.
How it started
Philipp Wu hit the missing-input problem as a UC Berkeley PhD student: he wanted robots to learn skills from large datasets, but the datasets barely existed. With Fred Shentu he built GELLO, a low-cost teleoperation system that lets a human drive a robotic arm to generate training data; the paper became influential as labs adopted the device for collection. In October 2024 Wu, Shentu and Nemo Jin launched XDOF to industrialize that loop.
What happened
XDOF spent 2025 building the operation in stealth, then emerged on 2026-06-17 with a $70M Series A and the ABC dataset release with UC Berkeley: 130,000 trajectories of robot manipulation data, 300 hours of simulation and 100 hours of evaluations, with the data already used to train robots at tasks like folding shirts and flattening boxes. The company told TechCrunch it was working with 20 customers including several frontier AI labs and planned to hire and train teleoperators and egocentric data collectors worldwide across three tiers: task-specific teleoperation, general GELLO-style collection, and everyday human egocentric data.
How it ended up
The growth outran the plan: people familiar with the deal told TechCrunch on 2026-09-04 that annualized revenue was approaching $50M and that investors approached XDOF for a Series B at about a $1.2B valuation led by 8VC, less than three months after the Series A — though terms were not final and XDOF declined to comment. TechCrunch quoted investors describing XDOF as the Scale AI or Mercor of physical robotics.
Background
XDOF was founded in October 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, joined by Nemo Jin, to solve the input problem of physical AI: robots cannot train on an internet-scale dataset because that dataset does not exist. The company collects real-world data through remote teleoperation and egocentric body sensors, then builds the cleaning, tooling and annotation around it.
The thesis traces to GELLO, the low-cost teleoperation arm Wu and Shentu built during Wu's PhD; its paper became influential as robotics labs adopted it for data collection. XDOF emerged from stealth on 2026-06-17 with a $70M Series A from Thrive Capital, Spark Capital, a16z, Lux and WndrCo, about 60 employees, 20 customers including several unnamed frontier AI labs, and the ABC dataset released with UC Berkeley's AI Research lab — 130,000 manipulation trajectories plus simulation and evaluation data.
Three months later TechCrunch reported late-stage Series B talks at about a $1.2B valuation led by 8VC, with people familiar saying annualized revenue was approaching $50M and that investors had approached the company. The open question is whether outsourced physical-data collection stays a defensible franchise as labs build internal capture, as they did with language data, or whether XDOF's operator network and dataset flywheel keep it the Scale AI of robotics.
What has to be true
- The missing input is real and dated: robot learning researchers had no large-scale physical dataset, which is why GELLO's cheap teleoperation design spread through the field first.
- The business model mirrors proven AI infrastructure: like Scale AI for text data, XDOF sells frontier labs an outsourced data supply chain instead of asking them to run warehouses of robots.
- Growth evidence is concrete: 20 customers at stealth exit in June 2026 and annualized revenue approaching $50M by September 2026, per people familiar with the company's finances.
- The capital escalation is checkable: a $70M Series A on 2026-06-17 was followed within three months by Series B talks at about $1.2B, terms still not final.
What can be applied
Own the boring bottleneck: while labs raced to train models, XDOF sold the missing input — physical data — and turned a PhD problem into a supply chain labs outsource.
Aftermath
As of September 4, 2026 XDOF is in late-stage talks to raise a Series B at about a $1.2B valuation led by 8VC, less than three months after its June stealth exit; people familiar with the deal told TechCrunch annualized revenue was approaching $50M and that investors approached the company, which had not planned to raise again so soon. Terms are not final. The company still declines to name its roughly 20 customers among frontier AI labs, plans worldwide teams of teleoperators and egocentric sensor wearers, and faces Mecka AI plus expanding human-data platforms like Scale AI and Micro1.
Sources
- Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it.
- XDOF, just three months out of stealth, is in talks for a Series B at a $1.2B valuation
- How Robot Startup XDOF Became a Unicorn Contender in Under a Year
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card