Unitree CEO: Embodied AI ChatGPT Moment May Take 2-10 Years

The wide 2–10 year window is itself the story: it signals that embodied AI still lacks a consensus path to general capability, unlike software LLMs where scaling laws offered a clearer roadmap. The bottleneck Wang identifies is dual — models and hardware — which matters for anyone tracking where robotics value accrues, since progress in either alone won't unlock general-purpose systems. Whether that window narrows or holds will depend on whether model breakthroughs and hardware advances start reinforcing each other, and that convergence is the thing to watch.