On the evening of August 22, the 2nd World Humanoid Robot Games opened at Beijing's National Speed Skating Oval, the "Ice Ribbon." 666 teams from 16 countries across six continents brought 2,056 robots to compete across 51 events and 1,301 matches over five days. But the most talked-about numbers came the night before, during the 100m preliminary heats for the full-size category.
Faster than the fastest human
TianGong Ultra, fielded by Team Tianzhuo, ran 9.39 seconds to win its heat, setting a new Games record and beating Usain Bolt's 9.58s men's 100m world record by 0.19 seconds. In the same heat, Team Fenghuo Shandian ran 9.47s and Unitree's team 12.41s. In the standing vertical jump demonstration, the same TianGong Ultra cleared 2.8843 meters (recorded as 288.43 cm) — compared with the inaugural event's winning mark of 95.641 cm last year, a near-3x improvement in the same metric within 12 months, reflecting rapid hardware and control progress.
A generational leap, not just a stat anomaly
Multiple competitive events this year eliminated manual remote control entirely, requiring fully autonomous operation. That means a single robot had to localize and map its surroundings and perform high-speed running and jumping on unfamiliar ground without lane-marking assistance — not comparable to the "coach + racing shoes + track + starting blocks" support behind Bolt's records. Skeptics argue that cars run faster than humans, so a robot outrunning a human isn't a sporting breakthrough. But an industrial fact is hard to ignore: the same fully autonomous robot both sprinted 100m and jumped 2.88m, indicating that its underlying stack — large model + joint actuation + high-speed control — has become a transferable capability set rather than a single-purpose demo.
Upgrades on both the hardware and algorithm sides
The robot's body comes from the Beijing Humanoid Robot Innovation Center, developed by the TianGong team based in Beijing E-Town. Optimized joint configurations and a lightweight, aerodynamic design improved power output. The algorithms run on the Huisikaiwu platform with a high-speed running control framework adapted for sprinting, and the robot no longer depends on lane markings — it maintains a humanoid running gait using SLAM-style localization and mapping within the track. These concrete engineering details make the record more than marketing copy.
TianGong, Galbot, and Zhiyuan on the same stage
The opening ceremony also featured 36 anthropomorphic five-finger dexterous hands playing musical instruments, 80 T2 robots performing formation changes to form the Games emblem, 25 Galbot G1 robots from Galaxeon demonstrating long-horizon complex tasks in home and retail scenarios, and a Zhiyuan Expedition A3 playing table tennis against Olympic champion Ding Ning — turning the ceremony itself into a spectrum of embodied capabilities from "performance → training → work." Tennis star Zheng Jie commented that the robots "moved continuously and completely autonomously," which matches this edition's core upgrade: the generational shift from remote control to autonomy.
The next industry inflection point
An industry report released alongside the Games (China Economic Net, 8/22) puts the market structure at: 40,000 humanoid robots shipped in China in H1 2026 — 97% of global volume — with major industry revenue of RMB 165.5 billion, up 24.5% year-on-year. As the industry moves upstream from demos (banquet dancing, launch events) to supermarket and factory roles "working 14-hour shifts," the combination of "faster than a human + fully autonomous + same machine" can serve as an early indicator of whether general embodied intelligence has entered practical use. The semifinals are on August 25 and the finals on August 26 — 9.39s is not the endpoint.
One more detail worth noting
Around August 22, Schaeffler, the German precision engineering giant, announced it will begin mass production of reducers dedicated to humanoid robots in 2027. This is an external response from the European supply chain to a fully autonomous Games: Asia's "Ice Ribbon moment" and Europe's "positioning for mass production" landed in nearly the same week, suggesting that the leap in full-robot capability and the industrial scaling of core components are now happening in sync globally.