XDOF Emerges from Stealth, Nearing Series B Funding Round at $1.2B Valuation

XDOF: The Rising Startup in Robotics Data Collection

XDOF, a startup founded by UC Berkeley researchers Philipp Wu and Fred Shentu, is making waves in the world of robotics data collection. In just a few months since coming out of stealth mode, XDOF is already in talks for a Series B funding round led by 8VC at a valuation of approximately $1.2 billion. This rapid growth is a testament to the company’s innovative approach to training general-purpose robots using real-world teleoperation data.

The startup’s journey began in 2024 when Wu and Shentu founded XDOF. TechCrunch reported on their successful $70 million Series A funding round, which attracted investments from Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. Despite not planning to raise funds so soon after their Series A, XDOF’s impressive growth, with annualized revenue nearing $50 million, caught the attention of venture capitalists, leading to discussions about a new funding round.

While the exact details of the Series B funding round are yet to be finalized, the interest from investors reflects XDOF’s potential in the robotics industry. The startup aims to provide data pipelines, collection tools, and annotation systems that are essential for AI labs and robotics companies to train their machines effectively. XDOF essentially acts as an outsourced data-supply chain for the robotics industry, addressing a crucial need for high-quality training data.

Wu’s background as a PhD student studying robot learning from large datasets inspired the creation of XDOF. His collaboration with Shentu on the GELLO project, a teleoperation system for generating training data by remotely controlling robotic arms, laid the groundwork for the startup. This innovative approach led to significant advancements in robotics research and paved the way for XDOF’s establishment as a key player in the industry.

Investors now liken XDOF to Scale AI or Mercor for physical robotics, highlighting its role in providing essential data for training general-purpose machines. Unlike language models that can draw from vast internet datasets, physical robots lack access to equivalent real-world data, making data collection a critical bottleneck in their development. XDOF’s partnership with UC Berkeley’s AI Research lab to release the ABC dataset, the largest collection of high-quality robot training data, underscores the startup’s commitment to advancing the field of robotics.

By combining remote robot teleoperation with human collectors equipped with sensors, XDOF captures a wide range of everyday tasks to enhance its training data. The startup plans to expand its team of data collectors globally, including teleoperators for remote robot control and egocentric operators for capturing movement data. With 20 customers already on board, including leading AI labs, XDOF is well-positioned to drive innovation in the robotics industry.

While XDOF faces competition from other startups in the real-world data collection space, its unique approach and strategic partnerships set it apart. The company’s dedication to providing high-quality training data for robot development positions it as a key player in shaping the future of robotics.

As XDOF continues to grow and expand its capabilities, it remains at the forefront of driving innovation in the robotics industry. The startup’s commitment to excellence and collaboration with industry leaders solidify its position as a leading provider of essential data for training next-generation robots.