Five years ago, in the eyes of most investors, Yushu was definitely not a star target.
Its founder does not have a background from prestigious overseas universities and lacks the so-called “elite halo.” Yushu’s recently launched three products have just begun commercializing, and the four-legged robotic dog products on the market have yet to shed the label of being “big toys.” Compared to other four-legged robot companies, Yushu in 2020 also seems not to have proven itself as anything other than a consumer-grade company. However, due to its passion for technology and self-developed technological barriers, Wang Xingxing continues to attract a group of investors.
“Wang Xingxing talked about technology, talked about robots, his eyes lit up, he is very confident in technology, very persistent, and at the same time, Unitree’s product iteration ability is also very strong.” Zhao Nan, an early investor in Unitree and founder of SevenUp Capital, still remembers the scene when he first met Wang Xingxing.
In 2020, Zhao Nan represented an organization to lead the pre-A round financing of Yushu. In an interview, he revealed that behind the small-scale commercialization at that time, Yushu had already accumulated top users and laboratory users such as NVIDIA and Google around 2020, and the iteration of technology and engineering capabilities also allowed the initial shape of the enterprise ecosystem to emerge. Currently, Yushu has achieved profitability, and recently in discussions with Wang Xingxing, the other party responded that there are no plans for a new round of financing.
Interestingly, Wang Xingxing did not initially plan to create “humanoid robots” as he believed there was a “uncanny valley effect”. However, the wave of generative AI quickly transformed this young founder’s perspective, and this year, Yushu Robotics made an appearance on the Spring Festival Gala, sparking a nationwide robot frenzy.
“Compared to consumer-grade products, Yushu is more likely to become a hardware foundation for a robotic ecosystem.” Zhao Nan said that in the future, it has the opportunity to become an ecological integrator in the field of robotics, like Xiaomi.
The robotic dog is not a “big toy”.
“Yushu is not a company that fell from the sky,” said Zhao Nan. In April 2020, introduced by other investors, Zhao Nan flew to Hangzhou and met with Yushu’s founder and CTO (Chief Technology Officer) Wang Xingxing in the Binjiang District.
During that time, Unitree had released three products in over three years since its establishment: the quadruped robot dog Laikago, the industry-focused quadruped robot dog AlienGo, and the educational quadruped robot Unitree A1.
At that time, four-legged robotic dogs were defined by most investors in the capital market as “big toys”. When a company is regarded as a consumer goods company, its business model and valuation system will fall into the assessment framework of product fun, necessity, and pricing strategy. He revealed that Yushu had previously also contacted companies such as DJI and Cheetah Mobile, but ultimately failed to reach financing cooperation.
“In comparison to DJI, the ground ecosystem in the robotics sector is larger.” Zhao Nan recalled his judgment at that time. He believes that DJI solved the problem of aerial operations, but the complexity and practicality of ground scenarios are much higher. Quadruped robots, due to their bionic structure and adaptability, can become mobile carrier platforms, equipped with robotic arms for industrial, security, inspection, and other scenarios. Moreover, unlike Boston Dynamics’ high-priced approach, Zhao Nan believes that China’s supply chain and research and development cost advantages can significantly reduce the costs of quadruped robots.
Zhao Nan said that the key factors that impressed him about Yushu were, first, the accumulation of core technologies developed in-house by Yushu. Wang Xingxing led a team to research and develop the underlying technology for robots since his graduate studies at Shanghai University, holding patents in joint modules, main control boards, and more. Second, Yushu had the awareness and capability to build its own factory early on. By 2020, Yushu had already established a small workshop in Hangzhou, independently developing molds to produce non-standard parts and solving mass production issues; third, the initial form of commercial verification. At that time, 70% of Yushu’s clients were overseas laboratories and tech giants (such as NVIDIA, Google, Meta, etc.), used for secondary development or AI training, while about 20% of clients applied it in industrial scenarios like security inspections.
“The existence of these customers proves that Yushu is not a toy company,” said Zhao Nan.
“The advantages and disadvantages of Yushu are very obvious.” Zhao Nan said that compared to other star projects, Wang Xingxing does not have the so-called “elite halo,” and Wang Xingxing himself admits to being “skewed in expertise.” The commercialization of the team is still in its early stages and lacks sufficient validation. But he is passionate about technology; “When he talks about technology, his eyes light up.”
From the perspective of industrial ecology, four-legged robots are a stable mobile platform. Zhao Nan positions the Yushu four-legged products as a mobile chassis that can be applied in industrial scenarios, equipped with intelligent sensors, robotic arms, and other secondary development structures for commercialization. For companies to achieve a certain market scale and valuation, platform-level companies have a greater opportunity. “Companies cannot integrate the upstream and downstream of the industry chain relying on a single product,” Zhao Nan said.
He also mentioned the name UNITREE, which actually reveals Wang Xingxing’s “ambition” — “Yu” means universe, and “Shu” represents the growth and extension of the technology tree. He hopes UNITREE will become a global technology company, rooted in core technology, with branches connecting ecological partners, ultimately growing into a “forest” that covers numerous ecological scenarios.
“Profitable every year after 2020”
Zhao Nan revealed that since 2020, Yushu’s financial statements have maintained a profitable status every year.
The secret to profitability lies in Yushu’s technology and cost accounting. At the beginning of Yushu’s commercialization, the well-established American robotics company Boston Dynamics was also testing the waters for commercialization at a price exceeding $70,000. “The problem with Boston Dynamics is that the hardware manufacturing cost is too high, but it already has some stable clients, which proves the market potential for this product,” Zhao Nan said.
Yushu’s advantage lies in its systematic opportunities backed by the Chinese industrial chain. The self-research of core components has transformed Yushu’s upstream suppliers directly from component suppliers into material suppliers, stripping away a layer of premium. In the non-standard component research and development phase, Yushu insisted on using its own factory for assembly and processing from the very beginning. Zhao Nan revealed that he visited Yushu’s own factory in 2022, “right near Yushu company, the team purchased relevant development molds themselves to enable some flexible supply chain production and develop core components.”
This directly enhances the efficiency of training performance for the Yushu training robot. High torque density joint motors and joint force sensors are key to achieving advanced motion capabilities such as flips. If the core components of the robot are damaged during the training process, it will take a long time to contact upstream component suppliers for repairs and parameter tuning. However, if the core components are developed in-house, the technical team will be familiar with the design, structure, and performance parameters of these components, enabling them to quickly determine the cause of the failure and begin repairs, allowing the robot to return to the training field.
Therefore, Yushu’s product iteration and cost control capabilities have been gradually stimulated through repeated training. In 2020, the price of the educational quadruped robot A1 released by Yushu was the first to drop to 100,000 yuan. A year later, the price of the robotic dog Go1 released for the C-end consumer market dropped to 16,000 yuan, and the price of Go2 released in 2023 fell to below 10,000 yuan.
The cost wheel of quadruped robots was quickly applied by Yushu to biped humanoid robots. Interestingly, Wang Xingxing did not plan to develop humanoid robots at the beginning of Yushu’s establishment. Zhao Nan revealed to the First Financial News reporter, “At that time, he felt that humanoid robots had the ‘uncanny valley effect’, and the technology and commercialization were not mature enough for development.” Zhao Nan said that it wasn’t until OpenAI released ChatGPT in 2022 that Wang Xingxing had a new judgment: the technology for intelligent learning in robots had new development opportunities.
This turning point also reflects Wang Xingxing’s pragmatism and keen insight.
“At this point, Yushu’s quadrupedal technology has gradually matured through several iterations, and there are still some overlapping parts in the underlying logic of joint sensors, controllers, and actuators between quadrupedal and bipedal systems.” Zhao Nan believes that although Yushu was not the earliest to develop bipedal robots, entering the field of bipedal humanoid robots is a natural expansion of Yushu’s business, thanks to the accumulation of quadrupedal technology.
In February 2023, the humanoid robot project was internally initiated at Yushu. Just about 7 months later, Yushu’s first humanoid robot product, H1, has already been unveiled to the public.
Embodied intelligence is expected to give rise to platform-level companies.
In the continuous iteration, the reliability and stability of hardware technology have become a new asset for Yushu. In a recent conversation with Wang Xingxing, Zhao Nan asked Wang Xingxing whether Yushu needed to conduct a new round of financing, to which Wang Xingxing replied negatively.
“Yushu is a platform-level company with ecological chain and collaborative capabilities. If we consider Yushu as a technological foundation, then its ceiling is very high,” said Zhao Nan. Yushu staff told reporters from Yicai at this year’s CES that Yushu’s quadruped robot has a global market share of 60%-70%.
In Zhao Nan’s vision, Yushu could potentially become a technological base in the robotics field, similar to the Android or Apple systems. “Yushu has the opportunity to become an ecological company like Xiaomi.” Just as Xiaomi has built a vast ecological system covering multiple fields such as mobile phones and smart home devices, Yushu is also relying on its bipedal and quadrupedal systems to carry out ecological development.
As a product that combines both hardware and software, the evolution of robots is destined to be somewhat slower. “The development of robots goes through two stages. The first stage involves the hardware itself, which is the foundational capability of the robot. The second stage is the integration with AI, enabling smarter learning and operations,” said Zhao Nan. He noted that software will experience explosive progress, but hardware cannot improve explosively; it can only change linearly.
“Perhaps Yushu will launch an AI OS system for robots based on the current carrier in the future, and based on this operating system, Yushu will open more interfaces to companies in the ecosystem.” In Zhao Nan’s vision, vertical AI reasoning applications developed by different enterprises will form a data-sharing and feedback mechanism within the robot’s hardware platform, thereby feeding back into the optimization of large models, continuously enhancing the robot’s understanding and response capabilities in complex scenarios.
He explained to reporters that for embodied intelligence companies, developing an embodied intelligence large model suitable for robots cannot be accomplished solely by an AI team, as it requires a deep understanding of the operational mechanisms of hardware during training, such as joint sensors, controllers, and drive components.
In 2025, some manufacturers of embodied intelligent entities also began to launch large models. In February of this year, the humanoid robot company Figure announced the launch of its self-developed universal VLA (Vision-Language-Action) large model Helix. Robots equipped with Helix can collaborate in multiple machines through a dual system architecture with low computational power requirements. The manufacturer Galaxy General also released an end-to-end embodied grasping foundational large model GraspVLA at the beginning of the year, demonstrating the robot’s action generalization capability under object interference. Star Motion Era also released an end-to-end native robot large model ERA-42, enhancing the robot’s operational capabilities. These companies are all starting from the robotic entities and gradually equipping them with richer AI capabilities.
Currently, China’s robotic dogs and humanoid robots are gradually being implemented in practical scenarios. On March 25, Audi FAW introduced the UBTECH industrial humanoid robot Walker S1 at its production base for the first time, conducting a pilot project for quality processes in the automotive production process, performing air conditioning leak detection tasks. In June last year, Leju Robotics signed a cooperation agreement with Jiangsu Hengtong to explore the large-scale application of humanoid robots in industrial scenarios; by the end of 2024, Zhaimi also announced a partnership with the humanoid robot company Magic Atom, providing factory environments and actual procedural scenarios for humanoid robots to undergo practical operation training.
As a semi-structured scenario, industrial environments are more “friendly” for current robots. “AI combined with embodied intelligence hasn’t reached the stage of showing off its muscles yet, but I believe that maybe in 1-2 years, there will be relevant breakthroughs in the industry,” Zhao Nan said.
(Source: Yicai)
Source: East Money
Author: Yicai
