Robotics Data Startup XDOF Nears $1.2 Billion Valuation in Fast-Tracked Series B

XDOF is in advanced discussions to raise a Series B funding round at a valuation of approximately $1.2 billion, according to a report from TechCrunch. The round is slated to be led by venture firm 8VC, marking a rapid escalation in value for a company that emerged from stealth less than three months ago.

The startup had not planned to return to the fundraising circuit so quickly after securing a $70 million Series A in June backed by Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. However, aggressive inbound interest from investors followed a sharp acceleration in commercial traction, with annualized revenue now closing in on $50 million across roughly 20 enterprise customers, including prominent frontier artificial intelligence labs.

Founded in 2024 by Philipp Wu, chief executive officer, and Fred Shentu, chief technology officer, XDOF originated from research at the University of California, Berkeley. During their academic work, the founders developed GELLO, a low-cost teleoperation mechanism designed to record precise human manipulation for robotic training. That research formed the architectural foundation for XDOF, which operates as a dedicated data-supply chain providing data collection pipelines and labeling infrastructure for physical machines.

To generate training datasets for tasks such as folding laundry or packing materials, XDOF relies on a network of global workers. These contributors provide demonstrations through remote teleoperation rigs and wearable body sensors. The company is also collaborating with the UC Berkeley AI Research lab to publish ABC, an initiative aimed at assembling one of the largest public corpuses of physical demonstration data.

The Physical Data Bottleneck

While first-generation large language models advanced by ingesting vast archives of public internet text, embodied AI systems face a severe physical constraint. Real-world mechanical interactions cannot simply be web-scraped; they require high-fidelity, multimodal sensor recordings of forces, angles, and spatial trajectories. As robotics developers race to build general-purpose systems, the market is placing premium valuations on providers that can systematically generate physical ground truth at industrial scale.

The escalating demand has created a competitive race among specialized data suppliers. Alongside XDOF, companies such as Mecka AI and general-purpose annotation vendors like Scale AI and Micro1 are expanding their operational footprints to capture the physical workflows needed to train the next wave of robotic foundation models.