XDOF is moving unusually fast for a robotics infrastructure startup. Less than three months after emerging from stealth with a $70 million Series A, the Berkeley-founded company is reportedly in late-stage talks for a new Series B that could value it at roughly $1.2 billion.
The reported talks matter because XDOF is not primarily selling a humanoid robot. Its business is building the training-data and deployment infrastructure around general-purpose robotics: teleoperation systems, data collection, robot operations and policy-training support for companies trying to train physical-AI models.
Robot AI Needs Real-World Data, Not Just Simulations
The current robotics boom has made one constraint increasingly obvious: foundation models need enormous amounts of useful physical interaction data. Synthetic environments help, but real robots still need demonstrations of grasping, tool use, manipulation, recovery and task sequencing across messy environments.
BitcoinVersus.tech has been tracking that same bottleneck from several directions. Trossen and Stereolabs are building a physical-AI robot learning stack, AMD is pushing physical AI from cloud systems into robots, and Boston Dynamics has opened an Atlas factory training center.
XDOF Sells the Infrastructure Around the Model
XDOF was founded in 2024 by Berkeley researchers Philipp Wu and Fred Shentu. Rather than betting on one robot body, the company positions itself as an infrastructure partner to robotics teams that need hardware operations, teleoperation, data capture and policy-training support at scale.
The company has also been explicit about the type of training data it believes matters. In this first-party update, XDOF points to paired demonstration data—linking human or robot demonstrations with the observations a policy sees at decision time.
That approach overlaps with the wider shift toward increasingly data-hungry robot policies. BitcoinVersus.tech recently covered Qualcomm’s move deeper into robotics software through PickNik, Aeva and LG Innotek moving 4D LiDAR toward robot production, and NVIDIA Isaac ROS 5.0 bringing AI agents into robot development.
$70 Million Came Before the New Talks
The reported Series B discussions come almost immediately after XDOF’s June financing. CEO Philipp Wu described that round in this announcement, saying the company had raised $70 million to build robotic infrastructure for foundation-model development.
That same post also announced ABC-130K, an open-source teleoperation dataset created with collaborators from UC Berkeley, Carnegie Mellon, MIT and Amazon FAR. The dataset angle is important because XDOF is trying to make robot-data infrastructure valuable both as a commercial service and as part of a broader research ecosystem.
Revenue Growth Is Driving Investor Attention
According to the people cited in the September report, XDOF’s annualized revenue is approaching $50 million, which is one reason investors are reportedly pushing for another round so soon. The discussions are said to be led by 8VC, but the financing has not been announced as closed.
That distinction matters. A reported valuation in fundraising talks is not the same thing as a completed financing. The final round size, lead investors, ownership terms or valuation could still change before any transaction closes.
The Business Case Is the Data Flywheel
If XDOF can keep expanding robot deployments, each deployment can potentially generate more demonstrations, more edge cases and more training data. That can improve policies, which can make deployments more useful, which can create still more data.
The risk is that robot foundation models remain difficult to generalize across hardware and environments. A teleoperation stack that works well for one robot type or manipulation domain may not automatically create transferable intelligence across every platform. XDOF’s value therefore depends not just on collecting data, but on collecting the right data in formats that customers can actually use.
Still, the reported valuation talks show how quickly the market is assigning value to the infrastructure behind robot intelligence—not just the robot itself. As physical AI moves from demos toward deployment, data operations may become as strategically important as actuators, sensors and compute.
BitcoinVersus.Tech Editor’s Note
Support independent technology reporting: BTC donations may be sent to 3C9o19EH5HSiwEPyCTmEKzxhNCbo2X6TTb.
Follow BitcoinVersus.tech on X/Twitter for robotics, AI hardware, semiconductor, Bitcoin mining and data-center reporting.
Disclaimer: BitcoinVersus.tech is not a financial advisor. This media platform reports on financial subjects purely for informational purposes.
Leave a comment