XDOF, a startup that collects real-world data to train general-purpose robots, is in talks to raise a Series B round that would value the company at roughly $1.2 billion, according to a report from TechCrunch, which said the round is being led by investment firm 8VC. The talks have not been confirmed as finalized, and the size of the new investment was not disclosed.
The company was co-founded in 2024 by Philipp Wu and Fred Shentu, two UC Berkeley researchers who built GELLO, a low-cost teleoperation system for controlling robot arms remotely, which became the technical basis for XDOF's approach. Wu had previously researched how robots can learn from large datasets as a doctoral student, an area where the field has long lagged behind large language models, which can train on vast troves of internet text. Physical robots have no equivalent dataset of real-world movement and manipulation to draw on, and that gap is what XDOF is trying to fill.
XDOF's method combines two kinds of data collection, according to TechCrunch's reporting: teleoperators who remotely pilot robotic arms through everyday tasks like folding laundry or flattening boxes, and "egocentric" operators who wear body sensors to record their own movements performing similar tasks, which can then be mapped onto a robot's range of motion. The company builds the data pipelines, annotation tools and quality-control systems that let frontier AI labs and robotics companies buy training data rather than build their own collection operations from scratch, effectively acting as an outsourced data-supply chain for the physical-AI industry.
A fast climb from a $70 million Series A
The company raised a $70 million Series A only in June, with participation from Thrive Capital, Andreessen Horowitz, Lux Capital and Spark Capital. A jump from that round to a reported $1.2 billion valuation within roughly three months would mark one of the steeper valuation climbs in the current robotics-funding boom. XDOF has said it already works with 20 customers, including several frontier AI labs, and that its annualized revenue is approaching $50 million — figures that, if accurate, would help explain investor appetite for the company despite its short public track record.
The talks are part of a broader surge in venture funding for robotics and "physical AI" companies this year, as AI labs and robot makers alike look for ways to give machines the kind of real-world training data that has powered progress in text and image generation. XDOF's pitch is that supplying that data is itself a large and defensible business, rather than a service any single robotics company would want to build in-house. Neither XDOF nor 8VC has publicly commented on the funding talks, and terms could still change before any deal is signed.
