Wang Xingxing on the ChatGPT Moment for Embodied Intelligence: 80% Household Robots Could Enter Homes in 2–3 Years (Fast) or 5–10 Years (Slow)

The 2026 World Robot Conference (WRC) is currently underway in Beijing, where Wang Xingxing, founder and chairman of Unitree Technology—fresh off its STAR Market IPO—delivered a keynote speech, offering a clear technical benchmark for when 'general-purpose robots will enter homes.'

Wang explained that the current global bottleneck in embodied intelligence is not motion control, but insufficient generalization capability. He pointed out that while most AI models can achieve near 100% success rates in fixed scenarios after sufficient data collection and project-specific training, their performance plummets dramatically when the operating object changes or the environment slightly shifts. This, he said, is the primary reason humanoid robots have yet to be widely deployed in homes and factories.

He believes that substantial breakthroughs in embodied intelligence could come in as little as two to three years, or as long as five to ten years, and expressed his hope to bring robots capable of performing 80% of household chores into ordinary households within a few years.

These remarks come just as Unitree went public on the STAR Market on Wednesday (August 19), with its opening market cap soaring to 440 billion yuan. The market is pricing in 'long-term discounted' expectations, but the industry remains stuck in a 'cold start' phase where robots fail as soon as the environment changes.

Wang's 'Dual 80%' criterion—80% task success in unfamiliar environments via voice commands—has thus become a hard metric for measuring the 'ChatGPT moment' for robots. Without meeting this threshold, no matter how acrobatic the hardware, the robot remains just a demo unit. But once this line is crossed, general-purpose robots can truly move from exhibition halls into living rooms.

Over 300 companies and 150 new product launches were showcased at WRC, but between 'showing moves' and 'doing chores,' what's missing isn't motors or computing power, but the layer of intelligence that allows models to make autonomous decisions in unseen kitchens, bedrooms, or in front of unfamiliar outlets. According to Wang's timeline, real-world validation data from actual homes within the next two to three years will determine whether the capital narrative can translate into appliance-level repeat purchases.