Unitree Robotics CEO Wang Xingxing is still trying to cool expectations after the company’s IPO. At the 2026 World Robot Conference, he said humanoids are not yet ready for large-scale factory or home use because they remain inefficient and must be retrained for new tasks. His benchmark for a “ChatGPT moment” in embodied AI is demanding: robots should complete about 80% of tasks in 80% of unfamiliar environments from simple voice or text commands. He expects that milestone in 3–5 years, or within 10 at most. Unitree is also developing “Physical AI Robot Self-Evolution V1.0,” a system where AI models generate robot-control code, train and validate it in simulation, then deploy it to real robots for feedback and further improvement. Unitree earned about $41 million on $252 million in 2025 revenue. At IPO pricing, it trades at 219x earnings—leaving Wang little room to disappoint investors.