Robotics and physical AI startups raise $48B this year, racing to gather training data

Robotics and physical AI companies have raised approximately $48 billion year-to-date, according to PitchBook data, driven by a push to gather training data from real-world environments. The companies are deploying cameras and sensors in factories, offices and homes to capture video of people completing everyday tasks, filling a critical gap in the training datasets needed to build effective robots and physical AI systems.

This data collection strategy addresses a fundamental challenge in robotics: most existing training datasets are synthetic or lab-based, limiting how well systems perform in messy, unpredictable real-world settings. By capturing footage of actual human work and movement, developers aim to train robots that can handle the variability of actual deployment.