China’s World Humanoid Olympics is framed as a live benchmark showing robotics leadership shifting toward China: humanoids won the 100m, 400m and 1500m, with 2,000 robots from over 16 countries and 96% Chinese participation. The core claim is that China’s scale, subsidies and manufacturing base are turning narrow robot demos into real physical capability faster than the US, where only a few humanoid efforts are named. The event is used to argue that embodied AI is now entering business use cases, from running to packaging, and that the next 1-3 years could expose whether Western humanoids can catch up. The episode also folds in a sponsored pitch for Ledger’s agent-security stack.
At China’s 5-day World Humanoid Olympics, a humanoid won the 1500m in 2:21, beating the human record of 3:26 that had stood since 1998.
The robots also posted a 9.39-second 100m versus Usain Bolt’s 9.58, and broke the 400m and standing high-jump events.
The event covered 55 Olympic-style events, drew 2,000 robots from over 16 countries, and robot entries were up 300% YoY while teams rose 200% YoY.
96% of the competitors were Chinese, supporting the claim that China is scaling humanoids far ahead of the US.
The winning robots were described as fully autonomous in flat events, with the new rules requiring zero human intervention in relays and other flat-surface races.
A robot ran 14.5 mph into a barricade, underscoring how raw and physical the competition still is despite the records.
Last year’s 100m-winning robot ran 21.5 seconds, so this year’s 9.39-second result was presented as less than a year of improvement and a collapse from more than 2x slower than Bolt to faster than Bolt.
The 400m winner, Tiang Gong, was said to have first been taught to run like a human, overheating its shoulder/rotator cuff and even breaking arms in tests before engineers switched to a reward objective that let it learn arm-swinging on its own.
That arm-flailing gait was described as an emergent balancing strategy, analogous to tails or counterbalances in fast animals.
In Beijing, Tiang Gong was named as the 100m record-holder, while Unitry was said to be runner-up across medals so far.
Unitry’s IPO was said to have come last week, with the stock up about 600% and a valuation cited at $55 billion after initially IPOing at $2 billion.
A separate robot-company IPO reference also said a stock was up 650%, still up roughly 4-500%, reinforcing the market re-rating around robotics.
China was said to be spending $20 billion in subsidies on robots over the last year and planning a 1 trillion yuan fund, with the mechanism described as subsidize until competitors win, then raise prices later.
China was also said to produce five times more robots than the US produced in all of 2025, and 70% of China’s AI tokens were claimed to be going to video models, much of that to robot models.
The US humanoid scene was described as thin — only two or three serious companies, named as Figure, Tesla and possibly Boston Dynamics — while China was described as having everyone building robots.
The Olympics format was framed as a physical benchmark that cannot be gamed like AI exams: if a robot can jump high enough or make a penalty/free throw, competence is visibly demonstrated.
The broader forecast was that a robot could be good enough to beat a human in tennis in 2-3 years, and possibly beat Carlos Alcaraz or Novak Djokovic within a year if the current pace holds.
Ledger’s agent stack was pitched as a security layer where the agent proposes, humans approve, and the ledger signer enforces, with open-source software that works across cloud code, codecs, Cursor and other AI integrations.