2025.08.09
Word count: 1389, estimated reading time: about 2 minutes
Author | First Financial Liu Jia
With protective gear, the robot swiftly punches, dodges, and high-kicks to knock out its opponent—at the Yushun booth during the World Robot Conference, the combat arena is surrounded by crowds. On the large screen in the background, scenes from the Spring Festival Gala show robots dressed in Northeast floral jackets performing traditional dances.
Behind the various “tricks” performed by the robots, during today’s speech, First Financial reporters noted that Yushun CEO Wang Xingxing repeatedly mentioned the desire for robots to truly perform tasks.
“It may sound unbelievable, but the people in our company who are actually making robots work, performing hand operations, and doing full-body tasks with AI are probably the most in number in our company,” he said. Yushun’s goal is to create a versatile humanoid robot capable of performing various tasks, including serving tea, working in factories, and performing in shows, not just in a single scenario, but so far, the results have not been particularly ideal.
This founder, who once firmly opposed humanoid robots, described a dramatic turnaround: Yushun had not developed humanoid robots at all until 2022, but significant advancements in AI technology represented by ChatGPT, along with pushes from figures like Musk, ignited market enthusiasm. At that time, customers approached Yushun, expressing their willingness to place orders directly, leading Yushun to start developing humanoid robots in early 2023.
This year, Yushun’s first-generation model made a highlight appearance on the Spring Festival Gala stage; the Unitree G1 from Yushun, starting at a low price of 99,000 yuan, ranks among the top in global shipments; the latest R1 robot’s price has been reduced to 39,900 yuan, with customizable appearances now available.
Wang Xingxing expressed that in the past six months, the advancement of AI technology in global robotics has been rapid. For example, in January this year, Yushun’s robots ran relatively stiffly, but after recently opening OTA updates to customers, the robots not only run more smoothly but can also navigate complex terrains.
The demand side is also experiencing an explosion. Wang Xingxing observed that in the first half of 2025, nearly every company in the robotics industry saw growth rates of almost 50%-100%, “the growth is quite astonishing.” From an international perspective, both Tesla and tech giants like Apple and Meta are increasing their investments in the robotics industry.
Behind the industry explosion, Wang Xingxing discussed common misconceptions about robots, such as why there hasn’t been large-scale application—is it due to insufficient hardware or high costs?
He stated that the current hardware, even the dexterous hand systems, is completely sufficient, even if not perfect. The biggest challenge currently is that embodied intelligence AI is still inadequate.
In simple terms, it feels like the 1-3 years before ChatGPT was released; the industry has already identified similar directions and technical routes, but no one has managed to implement them yet.
Is the lack of ideal results due to model issues or data issues? Wang Xingxing said that globally, there is too much focus on data issues, while the biggest problem is actually with the models. The current architectures for embodied intelligence models are not good enough or standardized, making it difficult to effectively utilize vast amounts of data. In the field of embodied intelligence, there are many cases where data is available, but it cannot be effectively used.
He also discussed the potential impact of video generation on the robotics industry, including Yushun’s plans to use AI-generated videos to guide the physical actions of robots starting in 2024. However, the industry is currently too focused on the quality of video generation, leading to excessive GPU consumption, while robots do not require high-precision video generation quality; they just need to be driven to work.
Wang Xingxing predicts that in the next 2-5 years, the focus of intelligent robot technology will include unified end-to-end intelligent robot models; lower-cost, longer-lasting hardware; and ultra-large-scale manufacturing; as well as low-cost, large-scale computing power (distributed).
Regarding when the “ChatGPT moment” for robots will arrive, Wang Xingxing painted a scenario of a critical point: if one day a humanoid robot is brought to a venue it has never seen before and is casually asked to hand a bottle of water to a certain audience member, and the robot can smoothly walk over and complete the task, that will be the moment for robots’ “ChatGPT moment.”
“If things go quickly, it could happen in the next 1-2 years or 2-3 years, with the slowest estimate being 3-5 years,” Wang Xingxing predicted.
WeChat Editor | Qi San
First Financial continues to track financial hotspots. If you have valuable insights on company dynamics, industry trends, financial events, etc., please feel free to share.Dedicated email:[email protected]
(Note: We will verify the tips. Your privacy will be strictly protected.)
Recommended ReadingDongfeng Liuzhou Auto Personnel: First Restoration of “Car Head Drop” Doubts and Details