Wang Xingxing Moves Embodied AI's GPT Moment to 'Possibly Next Year'

Wang Xingxing has moved the timeline forward again. Speaking at the 2026 Private Economy Innovation and Development Conference in Wenzhou on September 9, he said the ChatGPT moment for embodied AI could arrive in two to three years at the fastest — "possibly even next year."

Half a year earlier, his phrasing was "probably another two to three years." In August, public reports quoted him giving a range: "2-3 years at the fastest, 5-10 years at the slowest." By September the 5-10 year tail had been cut off, leaving only "possibly even next year."

Within six months, the timetable narrowed from two to three years to possibly next year. The change in phrasing is itself a signal worth noting.

On March 29, at the 2026 China Network Media Forum in Zhengzhou, Wang first offered a quantifiable definition of embodied AI's GPT moment. Take a robot into a completely unfamiliar scene, he said, with no prior mapping and no pre-demonstration; given only a voice or text instruction, it completes 80% to 90% of tasks. His judgment at the time was that this was probably still two to three years away, adding that this year or next would bring very large technical progress.

The definition was later condensed into the "80/80 standard": 80% unfamiliar scenes, 80% task completion. It is becoming the industry's common yardstick for whether embodied AI has crossed the commercialization threshold.