Sonny Valindra
Less than three months after publicly unveiling its operations, XDOF, a pioneering startup focused on amassing real-world teleoperation data to train general-purpose robots, is reportedly in late-stage discussions to secure a Series B funding round at an impressive valuation of approximately $1.2 billion. This significant financial milestone, if finalized, would underscore the burgeoning investor confidence in XDOF’s innovative approach to solving a critical bottleneck in the advancement of robotics. The round is expected to be led by prominent venture capital firm 8VC, according to multiple individuals familiar with the ongoing negotiations.
The emergence of XDOF into the venture capital spotlight has been remarkably swift. Founded in 2024 by UC Berkeley researchers Philipp Wu, who serves as CEO, and Fred Shentu, the CTO, the company has rapidly established itself as a key player in the robotics data ecosystem. TechCrunch first reported on XDOF’s substantial $70 million Series A funding in June, a round that saw participation from a formidable lineup of investors including Thrive Capital, Andreessen Horowitz, Lux Capital, and Spark Capital. At the time, XDOF had not anticipated seeking further investment so soon after its initial substantial raise. However, the startup’s exceptional trajectory, characterized by a significant surge in growth with annualized revenue now approaching the $50 million mark, has evidently attracted considerable attention from venture capitalists eager to participate in its next phase of expansion.
While the precise total capital being raised in this Series B round remains undisclosed, and it is unclear whether the reported $1.2 billion valuation encompasses the new funding, the ongoing discussions indicate a strong appetite for XDOF’s vision. The terms of the deal are still subject to finalization and could potentially undergo adjustments before a definitive agreement is reached. Both XDOF and 8VC have not yet responded to requests for comment regarding the specifics of these negotiations.
Addressing the Data Deficit for General-Purpose Robotics
XDOF’s core mission is to construct the sophisticated data pipelines, collection tools, and annotation systems that are often beyond the immediate capabilities of frontier AI laboratories and robotics companies. In essence, the startup is positioning itself as an indispensable outsourced data supply chain, a crucial element for the maturation of the physical robotics industry. This strategic focus directly addresses a fundamental challenge that has long impeded the development of versatile, general-purpose robots.
The genesis of XDOF can be traced back to the academic research of its co-founders. Philipp Wu, during his doctoral studies at UC Berkeley, was deeply immersed in investigating how robots acquire knowledge from vast datasets. He identified a significant impediment to his research: the stark absence of "large-scale data to work with." This realization sparked a collaborative effort with Fred Shentu on a project named GELLO. GELLO, a cost-effective teleoperation system, was designed to enable human operators to remotely control robotic arms, thereby generating essential training data. The influential research stemming from this project laid the groundwork for XDOF, a venture that investors are now increasingly likening to data-labeling giants such as Scale AI or Mercor, companies that played a pivotal role in fueling the artificial intelligence boom.
Unlike large language models (LLMs), which initially benefited from the vast and readily available data of the internet, physical robots lack an equivalent, comprehensive real-world dataset. This disparity renders data collection a critical bottleneck, hindering the creation of truly general-purpose machines capable of performing a wide array of tasks in diverse environments. XDOF’s business model is built around alleviating this very constraint.
The ABC Dataset: A Landmark Initiative in Robotics Data
In a significant move to democratize access to high-quality robotics data, XDOF is collaborating with UC Berkeley’s Artificial Intelligence Research lab to release what is believed to be the most extensive collection of high-fidelity robot training data ever assembled. This ambitious project is known as ABC (Autonomous Behavior Collection). The initiative aims to provide researchers and developers with the foundational datasets necessary to accelerate innovation in robotics.
The methodology employed by XDOF to capture this data is multifaceted. It involves a sophisticated integration of remote robot teleoperation with the active participation of human collectors. These individuals are equipped with sensors to meticulously record everyday tasks, ranging from the mundane act of folding laundry to the more physical effort of flattening cardboard boxes. This dual approach ensures the collection of diverse datasets, encompassing both direct robotic manipulation and human-centric physical interactions that robots are intended to emulate or assist with.
A Global Network of Data Collectors
Looking ahead, XDOF has outlined ambitious plans to build and train a global workforce of data collectors. This workforce will be comprised of two primary groups: teleoperators, who will remotely steer robots to perform specific tasks, and "egocentric operators." The latter will wear advanced body sensors to capture detailed movement data from a first-person perspective, providing a rich source of kinematic and dynamic information crucial for training robots to understand and replicate human motion.
The company has previously indicated to TechCrunch that it is already engaged with a substantial customer base, including approximately 20 clients, among them several leading frontier AI laboratories. This existing clientele speaks to the immediate demand for XDOF’s services and the validation of its data collection methodologies within the industry’s most innovative circles.
The Competitive Landscape and Future Implications
XDOF operates within an increasingly competitive landscape. Other startups are also venturing into the realm of real-world data collection for robot training. Notable among these are Mecka AI, which is developing its own approaches to data acquisition. Furthermore, established human-data platforms that have historically focused on LLMs are now expanding their scope to encompass the physical world. Companies like Scale AI and Micro1, both significant players in the AI data annotation space, are also beginning to explore or have already entered the robotics data domain, signaling a broader industry recognition of this critical need.
The implications of XDOF’s success and its approach to data collection are far-reaching. A robust and accessible supply of high-quality, real-world data is paramount for overcoming the challenges associated with training robots for complex, dynamic, and unstructured environments. This, in turn, is essential for the widespread adoption of robots in sectors such as manufacturing, logistics, healthcare, and even domestic assistance.
The potential for XDOF to become the de facto data provider for the robotics industry is significant. By effectively solving the data bottleneck, the company could dramatically accelerate the development timeline for general-purpose robots, bringing us closer to a future where robots can seamlessly integrate into various aspects of human life and work. The valuation being discussed for its Series B round, if it materializes, would serve as a powerful indicator of the immense market potential and the strategic importance of XDOF’s mission in shaping the future of robotics.
The journey from academic research to a potential unicorn valuation in less than a year highlights the immense pressure and opportunity within the AI and robotics sectors. Investors are clearly betting on XDOF’s ability to scale its data collection operations and provide the foundational intelligence that will power the next generation of autonomous machines. The success of this funding round will not only benefit XDOF but could also catalyze further investment and innovation across the entire robotics ecosystem.
