Robotics Data Startup XDOF in Talks for Series B at $1.2 Billion Valuation
XDOF, a robotics data startup that emerged from stealth less than three months ago, is in late-stage negotiations to secure a Series B funding round at a valuation of approximately $1.2 billion, according to individuals familiar with the matter. The investment round is reportedly being led by venture capital firm 8VC.
The discussions come shortly after the company raised a $70 million Series A round in June, which saw participation from prominent venture firms including Thrive Capital, Andreessen Horowitz, Lux, and Spark Capital. While XDOF did not initially plan to raise additional capital so quickly, venture capitalists approached the startup following its rapid financial growth, with annualized revenue now approaching $50 million.
The final terms of the deal have not been finalized and remain subject to change. The total amount of capital being raised in this round, and whether the $1.2 billion valuation includes the new funding, could not be confirmed. Both XDOF and 8VC did not respond to requests for comment regarding the negotiations.
Addressing the Robotics Data Bottleneck
Co-founded in 2024 by UC Berkeley researchers Philipp Wu, who serves as chief executive officer, and Fred Shentu, the chief technology officer, XDOF focuses on collecting real-world teleoperation data to train general-purpose robots. The startup aims to establish the data pipelines, collection tools, and annotation systems required by frontier artificial intelligence labs and robotics developers, positioning itself as an outsourced data-supply chain for the industry.
Unlike large language models that train on vast amounts of text available on the internet, physical robots lack a comparable pre-existing real-world dataset. This shortage of high-quality physical data has historically acted as a significant bottleneck for the development of general-purpose machines. Investors have compared XDOF's role in the physical robotics sector to that of Scale AI or Mercor in the text and software domains.
The concept for XDOF originated during Wu's doctoral research on how robots learn from large datasets, where he identified a critical shortage of large-scale data. To address this, Wu and Shentu developed GELLO, a low-cost teleoperation system enabling human operators to control robotic arms remotely to generate training data. Their research paper on the system laid the technical foundation for the startup.
Data Collection and Industry Partnerships
To build its data library, XDOF utilizes a combination of remote robot teleoperation and human collectors equipped with body sensors. These collectors record physical movements during everyday tasks, such as folding clothes and flattening boxes. The company plans to expand its operations by hiring and training global teams of data collectors, including teleoperators and egocentric operators.
XDOF is also partnering with the UC Berkeley AI Research lab to release a dataset named ABC, which it describes as the largest collection of high-quality robot training data assembled to date. The startup currently services 20 customers, a group that includes several frontier AI laboratories.
As the demand for physical robotics data grows, other companies are also entering the space. Competitors attempting to collect real-world data for robot training include Mecka AI, alongside established human-data platforms like Scale AI and Micro1, which are expanding their operations beyond language models.
