Less than three months after emerging from stealth mode, XDOF is advancing in late-stage discussions to raise a Series B round at a valuation of approximately $1.2 billion. The financing is expected to be led by 8VC, according to individuals familiar with the matter.
XDOF, co-founded in 2024 by UC Berkeley researchers Philipp Wu and Fred Shentu, focuses on collecting real-world teleoperation data to train general-purpose robots. The company previously raised a $70 million Series A in June, which included investments from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. At the time, XDOF had no immediate plans to return to the market so soon, but rapid growth—driving annualized revenue toward $50 million—prompted venture capitalists to approach the startup with new funding opportunities.
Details regarding the total capital being raised and whether the current valuation reflects the new investment remain uncertain. Deal terms are not yet finalized and may shift before completion. Representatives for XDOF and 8VC declined to comment.
The startup positions itself as an outsourced data supply chain for the robotics industry, providing the pipelines, collection tools, and annotation systems that frontier AI labs often lack the infrastructure to build internally. Wu, drawing from his doctoral research on robot learning, identified a critical shortage of large-scale datasets as a major obstacle. To address this, he and Shentu developed GELLO, a low-cost teleoperation system allowing humans to control robotic arms remotely to generate training data.
Investors have likened XDOF’s role in physical robotics to that of Scale AI or Mercor in the generative AI boom. While large language models were initially trained on vast internet corpora, physical robots lack an equivalent real-world dataset, making data collection a primary bottleneck for developing versatile machines.
To capture this essential data, XDOF combines remote teleoperation with human collectors wearing sensors who record everyday tasks such as folding laundry and flattening boxes. The company is currently collaborating with UC Berkeley’s AI Research lab to release ABC, described as the largest collection of high-quality robot training data ever assembled.
XDOF plans to recruit and train global teams comprising both teleoperators and egocentric operators. The startup reports it is already working with 20 customers, including several leading AI laboratories. Other firms competing in the real-world robotics data space include Mecka AI, alongside data platforms like Scale AI and Micro1 that are expanding their offerings beyond large language models.
Leave a Reply