U.S-based embodied AI infrastructure startup XDOF has officially emerged from stealth, announcing a $70 million funding round backed by investors including Thrive Capital, Spark Capital, Andreessen Horowitz (a16z), Lux Capital, and WndrCo.
Founded in 2024 by former researchers and engineers from UC Berkeley, Tesla, Meta, and Covariant, XDOF focuses on building the underlying data infrastructure required for robotics foundation models and embodied intelligence systems. Its platform is designed to support large-scale robot data collection, teleoperation, annotation, and training pipelines for next-generation robotic systems.
Alongside the funding announcement, XDOF unveiled ABC-130K, which it describes as the world’s largest open-source teleoperation dataset for robot manipulation. Developed in collaboration with researchers from UC Berkeley, Carnegie Mellon University, MIT, and Amazon, the dataset contains approximately 130,000 robot manipulation trajectories, aiming to accelerate robotics foundation model development across academia and industry.
According to CEO Philipp Wu, XDOF is already working with around 20 customers, including several leading AI laboratories, to address one of the biggest challenges in robotics: the lack of large-scale, high-quality real-world training data.
The funding will be used to expand XDOF’s global robot data collection infrastructure, annotation systems, and operational capabilities as the race to develop general-purpose robots continues to accelerate worldwide.
