The robot king's confession: his decade-long mistake was not believing in AI enough

Wang Xingxing builds more humanoid robots than almost anyone on the planet. Today, at the opening of China's biggest industry fair, he offered a confession you don't hear from hardware founders. "The biggest mistake I made over the past decade was not believing in AI enough." That's the guy who ships the bodies talking about the brain — and he clearly means it. Unitree is already running GPT-6 Astra on its robots in task tests, and the results, in his words, are "actually quite good." His real claim is about what's next: if GPT-7 ships on schedule next year, it could directly shatter the core bottlenecks holding back embodied intelligence. His bet now: the tipping point lands within one to two years. For context, two months ago he was saying two to three years at the fastest, up to ten at the slowest. He's compressed his own timeline. Here's the part that stopped me: he thinks commercial service robots could hit scale before factory robots. Everyone else assumes the factories come first. The money agrees. DeepSeek is reportedly leading a seed round valuing robot-AI startup Winverse at about $400 million, and Shanghai is rolling out a plan to put humanoid robots to work across its factories by 2028. The bodies were never the hard part. The brain is. And the brain is on a one-year fuse. Sources: - BigGo Finance (Oct 12, 2026): https://finance.biggo.com/news/9938dcd3-1694-4918-8c8f-55d75d198ea8 - Robot Belt (Oct 12, 2026): https://robotbelt.com/unitrees-wang-xingxing-my-decade-long-mistake-was-under-beli-u02/ - Yonhap Infomax (Oct 12, 2026): https://en.infomaxai.com/news/articleView.html?idxno=142879 - Humanoids Daily: https://www.humanoidsdaily.com/news/unitree-wang-xingxing-gpt-6-embodied-ai-tipping-point - ChainCatcher/Bloomberg (DeepSeek/Winverse): https://www.chaincatcher.com/en/article/2295685 - ODaily (DeepSeek/Winverse, Oct 12, 2026): https://www.odaily.news/en/newsflash/524034