Robotics data startup XDOF seeks $1.2B Series B just months after launch

Fresh from its public debut less than three months ago, XDOF is pursuing a Series B round that would value the robotics data company at roughly $1.2 billion, according to sources familiar with the negotiations. The round is being led by 8VC.

The swift move to a new funding stage reflects the startup's explosive early traction. Annual revenue is approaching $50 million, and investor interest has reportedly been strong enough that venture backers approached XDOF unprompted about the new raise.

Building the data foundation for robotics#

XDOF addresses a fundamental challenge in robotics development: the shortage of large, high-quality datasets needed to train general-purpose machines. While large language models could initially learn from vast amounts of internet text, physical robots lack an equivalent real-world resource, creating a critical gap in AI training infrastructure.

Co-founders Philipp Wu (CEO) and Fred Shentu (CTO)—both UC Berkeley researchers—built XDOF around a technology they developed as PhD students called GELLO, a low-cost remote control system that lets human operators guide robotic arms to generate training data. Their research produced an influential paper in the robotics field.

How it collects robot training data#

The company blends multiple data-collection methods to build its datasets:

• Remote teleoperation teams who control robots from a distance
• Egocentric operators wearing body sensors to capture natural human movements
• Collaborative work with UC Berkeley's AI Research lab on a dataset called ABC, which XDOF claims is the largest robot training data collection ever assembled

XDOF plans to expand globally, recruiting and training worldwide teams across both teleoperation and sensor-based data capture roles.

Market positioning and competition#

Investors increasingly liken XDOF to data giants like Scale AI and Mercor—companies that became critical infrastructure for the AI boom by providing labeled and curated datasets. The robotics domain, however, presents different technical challenges and represents a relatively untapped market.

The startup is already collaborating with about 20 customers, including several frontier AI labs. Other contenders in the space include Mecka AI, as well as platforms such as Scale AI and as they expand beyond language-model data collection.