Dyna Robotics has launched DYNA-2, a robot foundation model trained on over 1 million hours of human egocentric video — equivalent to 170 years of continuous experience. The model demonstrates the first human-to-robot scaling law, proving robot performance improves predictably as more human video data is added. DYNA-2 bypasses traditional robotics data bottlenecks by training on video rather than scarce teleoperation data. In high-precision manufacturing tasks, it increased success rates from 20% to 80–90% through pre-training scale alone. The model can adapt to different robot platforms with just hours of fine-tuning. In real-world evaluations, DYNA-2 completed tasks 1.55 times more successfully than the company's previous model, achieving 87% pass rates at customer deployments. Dyna Robotics was founded by Lindon Gao, York Yang, and Jason Ma, backed by investors including CRV and First Round.

Dyna Robotics raised $120 million, boosting its valuation to over $600 million. The funding, involving backers like Nvidia and Amazon, highlights investor confidence in AI robotics. This event may influence AI-related cryptocurrencies like FET and RNDR, potentially increasing trading volumes and prices. Nvidia's involvement could also impact its stock, creating cross-market opportunities. Traders should monitor market indicators and diversify portfolios to manage tech sector risks.