Rhoda AI Raises $450M to Build Foundational Robotics Models that Learn From Internet Videos

Why It Matters

Rhoda’s video‑driven training could dramatically reduce data bottlenecks, accelerating deployment of adaptable robots across manufacturing, logistics and service sectors.

Key Takeaways

• Raised $450M Series A, valuation $1.7B.
• Uses internet videos to train robotics models.
• Direct Video Action improves generalization over teleoperation.
• Backers include Premji Invest, Khosla, Temasek, John Doerr.
• Plans for humanoid robots and hardware licensing.

Pulse Analysis

Robotics has long wrestled with the scarcity of high‑quality training data. Traditional teleoperation relies on specialized gloves and sensors to capture a narrow set of motions, forcing developers to generate synthetic datasets that often miss real‑world variability. This data bottleneck limits a robot’s ability to adapt when faced with novel object orientations or unexpected failure modes, slowing adoption in dynamic environments such as warehouses or assembly lines.

Rhoda AI’s Direct Video Action approach flips the script by mining the internet’s massive repository of human activity videos. By aligning visual cues from everyday tasks with robot telemetry, the model learns a richer repertoire of motions, contexts, and edge cases without costly data collection campaigns. Early tests with off‑the‑shelf components in partnership with an automotive firm demonstrate that the system can handle diverse object orientations and recover from errors that would stump conventional teleoperated models. This video‑centric paradigm promises faster iteration cycles and broader applicability across industries that need flexible, perception‑driven automation.

The $450 million funding underscores investor confidence in physical AI as a growth frontier. With backing from Premji Invest, Khosla Ventures, Temasek and John Doerr, Rhoda is positioned to compete with established robotics players while expanding into humanoid platforms and proprietary hardware. As enterprises seek to blend digital intelligence with tangible actions—ranging from autonomous vehicles to smart factories—the ability to train robots at scale using publicly available video could become a decisive competitive advantage, reshaping the economics of robot deployment worldwide.