Article | Beyond the Page, Author | Hua Hua
A robot has outrun Usain Bolt. Then what?
Running, jumping, doing somersaults, and dancing with professional performers. Today's humanoid robots are quite adept at showcasing their most impressive aspects to the world.
But what enterprises truly need is never just an impressive performance; they value a reliable employee they can trust.
There is still a long distance between these two points. And last weekend, Zhiyuan just happened to move from one end of that distance towards the other.
On the evening of August 22, it appeared at Beijing's Ice Ribbon; the next day, it was at Hengqin Chimelong, over two thousand kilometers away.
One was a competition venue, the other a work site. When viewed together, their significance is entirely different.
1. Robots Begin Competing in Two Types of Capabilities
Let's first discuss the competition on the field.
The value of the Second World Humanoid Robot Games is straightforward. Speed, strength, balance, perception, control—what a robot can truly achieve becomes visible when put to the test on the field.
The scale of this event is also expanding rapidly. The number of events increased from 26 in the first games to 51, participating teams grew from 280 to 666, robot numbers rose from about 500 to 2,056, and the schedule extended from 3 days to 5 days. Teams from 16 countries across six continents participated.
These numbers indicate that robots are increasingly moving out of laboratories and into open competition. Zhiyuan also became one of the most outstanding performers in this competition. Throughout the games, Zhiyuan secured 18 gold medals and 46 total medals, ranking first in both the gold medal and overall medal tallies.
However, competitive sports and tasks like a single lift in a factory or a service in a hotel test two different sets of capabilities.
This precisely constitutes the two dimensions of competition in embodied AI today. One is the technical ceiling, the other is commercial implementation. This is also the most noteworthy change in this edition of the games: the competitions are beginning to move out of standard venues and into real-world scenarios.
This year's event added 21 scenario-based competitions, covering four major fields: emergency rescue, hotel service, library management, and family life.
Among them, Zhiyuan won six gold medals in scenario events. This achievement itself is certainly important, but what is even more noteworthy are the scenarios where Zhiyuan appeared. Firefighting, libraries—these places don't have specially prepared standard environments for robots. Robots need to complete tasks within existing environments and will also encounter unforeseen situations.
What the competitions are beginning to assess has extended from what actions can be performed to whether a task can be completed.
On the evening of August 22, 24 Zhiyuan Expedition A3 robots performed alongside 20 professional dancers, followed by a human-robot table tennis match with world champion Ding Ning.
The next day, it was at Chimelong.
From the Ice Ribbon to Chimelong, separated by thousands of kilometers north and south, the robots also faced two completely different challenges. The former scenario demanded pushing its capabilities to the extreme, while the latter required embedding those capabilities into real work.
2. Being Good at Just One Thing Is No Longer Enough
If we place today's embodied AI on a coordinate system, two distinct paths become evident.
One path emphasizes commercialization. Hotel delivery, campus services, simple搬运—these are areas where robots can already find numerous orders. Their advantages are controllable costs, clear demand, and large-scale deployment, but their technical ceiling is relatively limited.
Another group of players emphasizes technical capability. More and more robot companies are using sports events as technology showcases, constantly刷新records in speed, strength, balance, and dexterity. Some even customize robots specifically for competitions, pushing a single machine's specific capability to its极限.
Frankly, both paths can tell their own stories. But today, the market is beginning to raise a third question:
Is there a robotics company that can both achieve industry-leading technical capabilities and truly deliver robots to commercial sites?
On August 19, Yushu landed on the科创板. With an issue price of 150.8 yuan, it opened as high as 1100 yuan at one point, reaching a market cap exceeding 444.9 billion yuan. It closed the day at 845 yuan, still with a market cap over 340 billion yuan. On the second trading day, the stock price fell 18.7%, bringing the market cap down to 277.9 billion yuan.
In just two trading days, the market played out both sentiment and rationality.
This actually also signifies that capital markets' evaluation of embodied AI is becoming more concrete. Having orders and scale can prove commercial value; being able to run, jump, and perform difficult actions can prove technical prowess.
But if a company can only achieve one of these two capabilities, the ceiling of its story becomes clear.
According to Counterpoint Research data, global humanoid robot shipments in the first half of 2026 exceeded 22,000 units, a year-on-year increase of nearly 300%. The Ministry of Industry and Information Technology estimates that China's annual整机output may exceed 100,000 units.
The number of robots is growing rapidly, but shipment growth is only the first step.
Once a robot truly enters a working state, the questions immediately land on the operational ledger. Runtime, task efficiency,故障recovery, maintenance costs—all eventually need to be calculated alongside the price the customer is willing to pay.
These numbers already give robots the potential to enter commercial scenarios. But upon真正entering operation, cost, efficiency, reliability, endurance, and maintenance simultaneously impact the final economic viability.
These questions rarely appear on the stage of product launches but will determine whether a customer finally places an order.
3. Another Competition Is at Chimelong
If the capital market is starting to question whether robots can create value, Chimelong provides a site for direct验证.
Cultural tourism scenarios naturally feature high concurrent游客, long operating hours, multi-language interaction, dynamic environments, and clear service standards. With over thirty years深耕in tourism, Chimelong has highly mature experience in visitor services, park operations, and on-site management. For a robot to enter such a scenario means it needs to face an already highly mature service system.
If the sports games answer what robots can do, Chimelong answers whether robots can turn these capabilities into sustained work.
According to the agreement between the two parties, Zhiyuan's cooperation with Chimelong will cover four global projects: the world's first immersive robot theme park, the world's首创large-scale robot circus show, the world's most distinctive cyber-themed parade, and the world's first robot-service hotel. Both sides also plan to jointly establish a文旅industry embodied AI joint laboratory, conducting technological攻关and scenario验证across multiple parks.
The scenes here have no standard laboratory environments whatsoever.
Visitors, language, tasks, and environments are constantly changing. The robot no longer faces a pre-written set of movements; it needs to understand people, understand the environment, complete movement, interaction, and task execution, and maintain stability during long-term operation.
More importantly, once robots truly enter the park, the real scenes will continuously generate feedback. Which actions are effective, which interactions fail, which tasks require human intervention—all will become input for the next round of product iteration. This is another kind of test that commercial scenarios pose to technical capabilities. The longer a project runs, the more it accumulates.
4. From the Body to Delivery, Each Layer Is a Hurdle
Entering the site is just the beginning. The real challenge is turning a single success into continuous delivery. Once in a real environment, a robot faces more than just本体performance.
There is one easily overlooked detail about these two events at the games and Chimelong: scale.
At the games, Zhiyuan's three major robot series and dexterous hands all had gold medal records, appearing from the opening ceremony performance to multiple competition events. In Zhuhai, Zhiyuan又assembled over a hundred robots投入into multiple Chimelong scenarios.
The two events were over two thousand kilometers apart, happening almost simultaneously.
This is no longer just a capability showcase of a single robot; it is essentially a stress test of Zhiyuan's delivery capabilities.
From manufacturing and调度to on-site support, every step must keep up. Daring to push such a large scale of robots into two high-intensity sites simultaneously is itself a vote of confidence in mass production capability and the delivery system.
This also means that the final competition in embodied AI will not be about just one machine. Extreme performance in competitions and continuous delivery in real scenarios, two seemingly different dimensions, ultimately, within a company, rely on比拼system capabilities. The body is responsible for entering the physical environment, the model for understanding tasks, software and data支撑continuous operation, and when problems arise, people are still needed for delivery and maintenance.
Zhiyuan's consistently emphasized "Three-Intelligence Integration," including运动intelligence,交互intelligence, and作业intelligence, now has concrete landing points. From this sports event to Chimelong, Zhiyuan appeared simultaneously in two completely different实战locations.
The sports games push technical capabilities to a higher上限, while Chimelong places these capabilities in a commercial environment.
Zhiyuan is doing both things at the same time.
5. The Real Competition Has Just Begun
For Zhiyuan today, the truly interesting aspect of these two events is that they place a question that was easily discussed separately in the past onto the same company. Robots must have impressive technology and viable business.
Zhiyuan's founder, chairman, and CEO, Deng Taihua, once predicted that 2026 is a key turning point for embodied AI moving from development state to deployment state. He described this path using the "XYZ Curve."
The X curve is the development尝鲜stage where robots can move like humans; Y is the deployment成长stage where robots can work like humans; Z is when, driven by large-scale real-world scenario data combined with continuous algorithm innovation, embodied intelligence moves from quantitative to qualitative change, towards broader application deployment and普及stage.
Looking at the industry, this XYZ curve corresponds precisely to the process of robots moving from being able to perform to being usable, and then towards规模化普及.
Mass production solves whether robots can be manufactured;规模化deployment faces another matter. As more and more robots enter different scenarios, can performance and stability be maintained?
Whether a robot can be continuously used by customers depends on hardware as the foundation. How much work it can ultimately complete also depends on the model's understanding of the environment and tasks, and the system's stability over long-term operation.
For enterprises, purchasing robots必然involves calculating an operational account: how many tasks can be completed per month, how much cheaper it is compared to labor, the故障rate, and maintenance costs. These questions rarely appear in tech launch events but will determine how far robots can ultimately go.
The two scenarios Zhiyuan engaged in this time恰好place the changes happening in the industry together. They also accelerate the competition in embodied AI from product competition among robotics companies into a competition of系统工程capabilities.
In the past, embodied AI players could gain attention through technology or secure orders through commercialization. Next, the market will increasingly value one thing:
This transition is happening faster than we anticipated. And Zhiyuan正好stands at the node where this change is occurring.
A Word from [Beyond the Page]:
A robot outrunning Usain Bolt only proves it runs fast.
What truly changes an industry is not only achieving a certain height in technology but also bringing those capabilities into production environments, leaving behind orders and customers, and truly running a replicable business model.
When a company begins to enter this cycle, embodied AI can be said to have truly stepped into the next phase.
