Embodied AI Surges: New Models, Datasets, and Billions in Funding

Today's embodied AI landscape is defined by breakthroughs in data and learning. Skild AI's S1 foundation model learns tasks from a single video, rivaling VLA models trained on 100x more data. Figure's Index, the largest robot dataset with 16M videos, tackles the data bottleneck head-on, backed by $1B. In funding, XPENG Robot raised $900M, valuing it at $6.3B, a record for China's sector.

Policy and industry momentum follow: MIIT designates intelligent robots as pillar industries, while AGIBOT dominates the World Humanoid Robot Games and teams up with Chimelong for an embodied AI theme park. JD.com commits $1.4B to robot deployment, and Ubtech partners with Singtel for enterprise adoption. On the tech front, NVIDIA's Jetson Orin Nano 2 doubles edge inference, and a Chinese humanoid robot sprints 100m in 8.86 seconds.

These developments signal a rapid convergence of research, investment, and real-world applications, propelling embodied AI toward mainstream deployment.