Humanoid Robots Beat Usain Bolt's 100m Record at Beijing's 2026 World Robot Games

Posted on 26th Aug 2026 06:04:58 in Artificial Intelligence, Machine Learning

Tagged as: humanoid robots, robotics, AI, Beijing robot games, Tiangong Ultra

Usain Bolt's 100-metre world record of 9.58 seconds stood unchallenged since the 2009 World Athletics Championships in Berlin. On Saturday, August 22, 2026, it finally fell — not to a human sprinter, but to a two-legged machine. At the opening of the second World Humanoid Robot Games in Beijing, a Chinese humanoid robot crossed the finish line in 9.39 seconds, almost two tenths of a second faster than the fastest man in history. The five-day event, which wraps up on August 26, has become the clearest public benchmark yet of how fast humanoid robotics is advancing — and how far it still has to go.

The Games are an Olympics-style showcase organised by China Media Group and the Beijing municipal government. This year's edition is dramatically bigger than the 2025 debut: 2,056 robots, up 311 percent year over year, competing in 666 teams, up 138 percent, drawn from 16 countries and regions across six continents. Roughly 96 percent of the entries are Chinese, underlining how concentrated the humanoid industry currently is. Across 51 events — around 30 modelled on human sport and the rest on workplace scenarios — competitors faced off in everything from sprinting and table tennis to bricklaying and bottle-opening.

Records Fall: Faster Than Bolt, Higher Than Sotomayor

The headline numbers came early. Tiangong Ultra, developed by the Beijing Humanoid Robot Innovation Center, won its 100-metre heat in 9.39 seconds, breaking the 9.58-second mark Bolt set in 2009. It was pushed all the way by Lightning, a robot built by smartphone maker Honor, which finished in 9.47 seconds — also faster than Bolt's record. The result was a stunning leap from the inaugural Games in 2025, when Tiangong Ultra won the same distance in 21.50 seconds. In one year, the winning time improved by more than half.

Lightning's researchers had revealed their ambition even before the opening ceremony: in a preparatory test, the robot covered 100 metres in 9.32 seconds at a peak speed of 14.5 metres per second. The robot, which stands 169 centimetres tall, had its legs lengthened by 10 centimetres ahead of the Games, according to state broadcaster CCTV. Lightning also won this year's humanoid half-marathon in 50 minutes and 26 seconds, a pace faster than the elite men's human world record.

The records were not limited to the track. In the standing high jump, a humanoid cleared 2.88 metres — 43 centimetres above the 2.45-metre human world record set by Cuba's Javier Sotomayor in 1993, and roughly three times the 0.95-metre best achieved by a robot at last year's Games. Organisers and manufacturers touted the results as proof that Chinese humanoid robots have begun to surpass human athletic benchmarks, with hundreds of robots marching in synchronised formation at the opening ceremony as a demonstration of coordination at scale.

From Sport to Work: The Event List Is a Market Map

The medals are spectacle, but analysts argue the event schedule itself is the real story. Writing for Forbes before the Games opened, analyst John Koetsier called the programme "one of the most honest documents in robotics right now" — a market map of what China expects humanoid robots to be paid to do. The competition is split into two halves. The sporting events test dynamic motion: running, soccer, gymnastics, dance, boxing, table tennis and weightlifting. The scenario events simulate jobs: retail service, food preparation, assembly, warehousing, hospitality, household chores, firefighting and rescue.

The shift in emphasis is deliberate. Wang Peng of the Beijing Academy of Social Sciences summarised the change bluntly: "Last year it was about whether they could run to the finish line. This year it's about whether they can get the job done." Almost all of the event growth between 2025 and 2026 came in the work-scenario half. A new suite of eight dexterous-hand challenges is the clearest signal: robots must assemble power tools, weigh powder, stack blocks, fasten nails, open caps and lids, unpack boxes, pick up beans with tweezers, and connect cables. These are not sports; they are the fine-motor tasks that define real jobs on assembly lines and in warehouses.

The working-scenario events matter because China's humanoid industry faces a credibility problem. Manufacturers claim thousands of orders from government and private buyers, but experts note the technology is still used mostly for demonstrations and performances. The Games are an attempt to prove that prototypes can handle genuine work — and to give potential customers a public, comparable benchmark before they commit capital.

The Autonomy Fine Print: What the Rules Admit Is Still Hard

The competition rules are an unusually candid disclosure of what is solved and what is not. In the sporting half, flat-ground running races — including the 400 metres, 1,500 metres and the 4x100 relay — are now "fully autonomous" categories: robots must run without any remote control. Soccer, gymnastics, dance, boxing, Tai Chi and table tennis likewise require hands-off autonomy. But hurdles, jumping events, maximum-load weightlifting and kickboxing do not. In those events, a human operator may still be in the loop.

The distinction maps cleanly onto known technical frontiers. Closed-loop perception at speed — seeing the track, staying balanced and correcting course without a human pilot — is now considered a solved, commodity capability. Explosive dynamics with hard landings, adversarial contact and reactive decision-making under physical stress are not. In the work-scenario events, the scoring system encodes the same message: full points for completing a task autonomously, half points if a human teleoperator is involved.

What is missing from the schedule is just as revealing. There are no stairs or ladders, no unstructured terrain, no endurance or battery-life event, no test of human-robot physical collaboration, and no metric for cost, energy per task, reliability across a full shift, or mean time between failures. There is also no standalone test of instruction-following — giving a robot verbal orders and checking whether it can complete the task. Forbes' verdict is blunt: "You've come a long way, baby. But there's a really, really long way to go yet."

Crashes, Fires and Stretchers: The Other Side of the Spectacle

The failures at the Games were as public as the records. Footage showed sprinters slamming into padded barriers at the finish line because they still cannot stop themselves safely at speed, before being carried off on stretchers. Other robots tripped, fell apart mid-competition, or in at least one case caught fire. NBC News captured the paradox in its headline: "Robots can outrun humans, but can they plug in a cable?" TechRadar noted that the biggest hurdle for a record-breaking sprinter is not the race itself but stopping safely at the end of it.

These failures are not embarrassing footnotes; they are the honest state of the technology. A robot that can run 100 metres in 9.39 seconds but cannot decelerate without crashing is not a product, it is a prototype. The gap between athletic peak performance and mundane reliability — walking on stairs, plugging in a cable, working a full shift without falling — is precisely the gap between the Games' sporting half and its work-scenario half. The organisers' own rules, which still permit teleoperation for anything involving hard impacts or fine manipulation, acknowledge that gap openly.

What the Games Mean for the US-China Robotics Race

The timing of the Games is politically loaded. In July 2026, the US Federal Communications Commission banned imports of new foreign-made humanoid robots, citing national security, and the Pentagon added Unitree — one of China's leading humanoid makers — to its list of companies deemed to have ties to the Chinese military. Beijing has rejected the accusations, but the message from Washington is clear: humanoid robots are now treated as strategic infrastructure, not consumer electronics.

China makes the majority of the world's humanoid robots, and the Games are a deliberate national showcase for a sector President Xi Jinping has singled out as a strategic emerging industry. State media coverage is celebratory, and the spectacle — synchronised formations, record-breaking sprints, table tennis rallies — is designed to attract investors and customers as much as viewers. The subtext of the 2026 Games is that China intends to industrialise humanoids on its own terms: at scale, at speed, and with a public scoreboard to prove progress.

The honest read of Beijing's Robot Olympics is therefore a dual one. On the track, the progress is real and accelerating: a 21.50-second 100 metres became a 9.39-second one in a single year. In the workplace, the verdict is still open. The event list proves that running is a commodity, but cable-plugging is not. Whoever closes that gap first — turning a 9.39-second sprinter into a robot that can stop safely, climb stairs and finish a shift — will win a market measured in trillions, not medals.

Sources

• AP News — Chinese humanoid robots break human records at Beijing's robot games
• ABC News — Chinese humanoid robot beats Usain Bolt's 100 metres world record
• CBS News — Chinese robot beats Usain Bolt's 100-meter sprint world record in humanoid bot games
• Forbes — The Humanoid Robot Games Events Are A Market Map
• Wikipedia — World Humanoid Robot Games
• NBC News — Robots can outrun humans, but can they plug in a cable?