Decoding the Differentiation and Breakthrough Strategy of Chinese Humanoid Robots from UBTECH and Yushu Technology Financial Reports

Recently, Hong Kong Exchanges and Clearing Limited’s (“HKEX”) “first stock of humanoid robots” story—Ubtech (09880.HK)—released its 2025 annual performance report, drawing attention from the capital markets.

The next day at market open, Ubtech’s share price surged with a gap-up and opened significantly higher. During the trading session, the gains at one point nearly approached 20%, setting the largest single-day jump in nearly a year. Subsequently, Citigroup raised Ubtech’s target price by 23% to HKD 190 and also increased its forecast for the company’s annual revenue.

Previously, Ubtech officially announced that it had reached a comprehensive strategic partnership with Siemens, a global industrial automation powerhouse. The two sides will collaborate end-to-end in industrial digitalization, intelligent manufacturing, and the deep integration of humanoid robots, jointly building industrial humanoid robot solutions for the global market.

Against the backdrop that most enterprises are still in the technology validation and scenario exploration stage, Ubtech’s business scale has shown a significant increase.

According to the financial report, in 2025 Ubtech achieved revenue of RMB 2B, up 53.3% year over year; overall gross profit was RMB 750 million, up 101.5% year over year. Among them, the full-size embodied intelligence humanoid robot business grew more notably: full-year revenue reached RMB 820 million, up more than 20-fold year over year; unit sales were 1,079 units; gross margin was 54.6%. It became the company’s largest revenue source, accounting for 41.1%.

In terms of delivery scale, Ubtech became the first company globally to deliver industrial humanoid robots at the thousand-unit level.

This development is not only an important milestone in Ubtech’s growth process, but also to a certain extent reflects that humanoid robots are starting to move from technology validation to application exploration. Recently, Ubtech has been collaborating with the aforementioned Siemens, Airbus, and Texas Instruments—major Western high-end manufacturing players—showing the ability of Chinese manufacturers in product engineering and delivery pacing, and is gradually gaining attention from international industrial systems.

From an industrial foundation perspective, China has already formed a relatively complete manufacturing and supply-chain system in the humanoid robot sector, with certain advantages in engineering capabilities, cost control, and the pace of scenario rollout. Against this backdrop, different enterprises have also formed differentiated development paths. Ubtech mainly focuses on industrial scenarios, promoting real applications of humanoid robots in production processes; Unitree Technology relies on its strengths in motion control and body design to build a general-purpose platform for research, education, and developers.

These two approaches correspond to different stages and priorities of industrial development, together forming the current landscape of diverse exploration in the humanoid robot industry.

Two approaches

During the development of the humanoid robot industry, companies have shown clear differentiation in technology routes and resource allocation.

Represented by Ubtech and Unitree Technology, two types of companies lay out their strategies around different technical emphases and application directions, gradually forming differentiated product systems and business structures.

Among them, Ubtech continues to increase R&D investment, focusing on building a technical framework related to embodied intelligence. In 2025, the company’s R&D spending reached RMB 507 million, accounting for about 25.4% of revenue. In the past four years, cumulative investment nearly reached RMB 1.9 billion. The size of the company’s R&D team is 942 people, of which the proportion of master’s and PhD personnel is close to half.

On this basis, Ubtech has structured its strategy around embodied intelligence large models, vision-language-action (VLA) models, and world models, aiming to enhance humanoid robots’ capabilities for autonomous decision-making and coordinated work in complex scenarios. The company’s self-developed Thinker embodied intelligence large model, Thinker-VLA vision-language-action model, and Thinker-WM world model for industrial embodied intelligence humanoid robots together build a complete embodied intelligence technology stack. Its industrial version robot also completes the upgrade of the Brain-in-a-Group network 2.0 and Co-Agent industrial collaboration intelligent agents, driving humanoid robots to evolve from single-unit autonomy to group intelligence, achieving unified scheduling and coordinated operations.

As of now, Ubtech has accumulated more than 2,900 authorized patents, including over 1,700 invention patents.

Unitree Technology, by contrast, adopted a more focused R&D strategy, concentrating mainly on core areas such as motion control, body design, and hardware systems. In the first three quarters of 2025, the company’s R&D expenses were about RMB 90.2 million, accounting for about 7.7% of revenue. Full-year R&D investment is about RMB 90 million to RMB 120 million. The size of its R&D team is about 175 people.

In terms of its technology route, its products emphasize high-performance motion capabilities and modular design, and promote algorithm iteration and application expansion through an open-source ecosystem and developer community.

In terms of R&D structure, the two types of companies have accumulated strengths in different dimensions such as “system capabilities” and “body capabilities.”

Different technical routes also correspond to different product forms. Ubtech mainly focuses on full-size industrial humanoid robots, emphasizing autonomous working ability and scenario adaptability. Unitree Technology mainly uses mid- to small-size humanoid robots, emphasizing flexibility, cost control, and platform generality.

At the application level, Ubtech mainly serves industrial scenarios, with related business accounting for over 80%. Its products have begun application exploration in areas such as manufacturing and logistics. In 2025, Ubtech’s full-size embodied intelligence humanoid robots achieved mass production, delivery, and acceptance at the thousand-unit level. They were used across multiple industrial fields including automobile manufacturing, smart logistics, 3C electronics manufacturing, semiconductor manufacturing, aerospace manufacturing, and industrial data acquisition. On January 18, Ubtech signed a services agreement with European aviation giant Airbus to achieve the first landing of humanoid robots in an aviation manufacturing scenario. At the same time, the company has also partnered with Fortune 500 enterprises such as Texas Instruments, Audi FAW, BYD, and Foxconn, enabling large-scale applications of its robot products in processes including handling, sorting, and quality inspection.

Unitree Technology mainly serves the R&D, education, and developer market. Its products are widely used in universities, research institutions, and various exhibition scenarios. In 2025, the shipment volume of its non-full-size humanoid robots exceeded 5,500 units, and about 70% of its products were sold to major universities and research institutions worldwide, used for algorithm research and teaching demonstrations. The remaining products are widely applied in scenarios such as cultural tourism performances, live-stream e-commerce, and home entertainment. The company’s robots have appeared multiple times on CCTV’s Spring Festival Gala stage, and through performance formats such as dance and martial arts, effectively enhancing brand awareness in the market.

Overall, the two types of companies drive the development of the humanoid robot industry in different dimensions: one side focuses on scenario deployment and system integration capability, while the other drives technology dissemination and ecosystem building.

What kind of humanoid robots does the industry need?

In early 2026, operating data disclosed by multiple humanoid robot companies reflected that the industry is advancing commercialization and exploration along different paths.

In the early stage, products with strong motion capabilities and demonstration effects help improve public recognition and market attention. At the same time, they also provide a foundational platform for research and developers, promoting technology iteration and application validation.

Data show that in 2025, the number of complete-hardware humanoid robot enterprises in China exceeded 140, and more than 330 products were released. The total industry financing scale for the year reached USD 2.65 billion, exceeding the sum of the previous several years, as capital and industry resources have accelerated their inflow.

As the industry advances, the focus of attention is also changing. Compared with “whether it can complete complex actions,” the market has started to pay more attention to a robot’s ability to perform sustained tasks in specific scenarios—i.e., whether it can “complete tasks reliably.”

IDC pointed out that the focus of competition in the industry is shifting from hardware performance to technical depth, service capabilities, and ecosystem building. By the end of 2025, the Ministry of Industry and Information Technology officially established the Standardization Technical Committee for Humanoid Robots and Embodied Intelligence. In early 2026, it released China’s first Standard System for Humanoid Robots and Embodied Intelligence (Version 2026), covering the entire industrial chain and full lifecycle. This marks that the industry has entered a new stage of standardized, systematic development. At this stage, it is the “technical depth,” “industrial deployment capability,” and “independent development of core components” that are the key to determining who can finish the entire course.

As a productive tool, humanoid robots are increasingly becoming a core proposition in technology competition among major countries.

IDC data show that in 2025, global shipments of humanoid robots were about 18k units, up 508% year over year, with Chinese manufacturers taking a dominant position. Research comparing the innovation capabilities of humanoid robot companies in China and the United States across four dimensions—product development, technical innovation, financing situation, and market applications—found that Chinese companies are better in exploring a broader range of product application scenarios, while U.S. companies have an advantage in breadth of patent portfolio development.

This structure implies that the industry is still in a parallel stage of “application advancement” and “technical accumulation.” If companies overemphasize front-end demonstrations and neglect building underlying capabilities, their long-term competitiveness will still be constrained.

Against this backdrop, different types of companies play different roles in the industry: one type drives technology diffusion and application exploration through general-purpose platforms and a developer ecosystem; the other focuses on specific scenarios and promotes real-world applications of robots within production systems.

Only in this way can Chinese humanoid robots truly lead change in the new round of the global industrial revolution—turning mass-production advantages into technological advantages, scaling advantages into standard advantages, scenario advantages into ecosystem advantages—and ultimately securing a first-mover position in the global landscape of scientific and technological competition.

Embodied intelligence enters the first year of mass production

In 2025, the global technology community widely recognized it as the “first year of mass production for humanoid robots.” The latest data from international research institution Omdia show that in that year, global shipments of humanoid robots were about 13k units, with Chinese companies accounting for 90% of market share and taking the top six spots on the shipment ranking. This figure breaks the longstanding perception that “the United States leads in technology while China follows with imitation.” China and the U.S. have already taken entirely different development paths in the embodied intelligence track.

The U.S. camp, represented by Tesla’s Optimus and Figure AI, places more emphasis on general capabilities and frontier technology exploration in the humanoid robot field.

In terms of progress, related products are still mainly in testing and early validation stages. Data newly disclosed by Tesla show that in 2025, Optimus had a production volume of about 150 units, mainly for internal testing. This mass-production figure is also far below the 5,000-unit target previously set by Musk. Although Figure AI has received investment from multiple technology giants including Microsoft, Nvidia, and Amazon, its delivery scale is still limited: fewer than 200 units for the whole year, mainly serving laboratories and early customers.

Overall, U.S. companies have an accumulation advantage in foundational technologies such as algorithms and large models, but they are still in the process of continuously pushing forward in areas including industrial-chain support, cost control, and scenario deployment.

U.S. companies’ strengths lie in foundational algorithms and large-model technology accumulation. However, their shortcomings are equally obvious: industrial-chain support is incomplete, and core components are highly dependent on external procurement, leading to persistently high production costs. Currently, the price of comparable humanoid robots in the U.S. is generally around 300k USD, which is 5–10 times that of China’s industrial-grade products, making large-scale commercialization deployment difficult. More importantly, U.S. companies generally lack experience in manufacturing industry scenarios. Most of their products remain at the stage of “can walk and can move,” and there is still a long road ahead before true industrial applications.

By comparison, Chinese companies focus more on specific application scenarios, driving validation and iteration of technology in real environments.

In this process, a batch of representative companies have gradually emerged domestically. Taking Ubtech and Unitree Technology as examples, the former focuses on industrial scenarios and promotes application deployment of humanoid robots in production systems; the latter leverages strengths in motion control and body design to build a general-purpose platform for research, education, and developers. Through exploration along different dimensions, the two types of companies correspond to different emphases on scenario deployment and technology diffusion, jointly promoting the development of the humanoid robot industry.

China’s leading advantages are not accidental; they are built on three major non-replicable foundations:

First is the most complete global industrial chain support system. China is the only country in the world that has all industrial categories in the United Nations Industrial Classification system. In the humanoid robot field, China has achieved full-chain coverage—from reducers, servo motors, and sensors to final assembly of complete units. The localization rate of core components exceeds 80%. Places such as Shenzhen and Dongguan have formed industrial clusters where “moving up and down floors means upstream and downstream,” and Ubtech can assemble a humanoid robot from zero components, test the complete unit, and get it out of the factory in about one hour—far faster than the 15–20 day cycle of European and American companies.

Second is extreme cost-control capability. Relying on the scale-based production experience accumulated in consumer electronics and new-energy vehicle industries, Chinese companies have reduced the cost of the humanoid robot’s core components by an order of magnitude. The price of the new generation of joint modules has moved from the thousands of RMB to the hundreds-of-RMB range. Complete unit costs have dropped from several million RMB to tens of thousands of RMB, laying the foundation for large-scale commercialization.

Finally is the richest global application scenario resource. China has the world’s largest manufacturing market, providing a broad testing ground for industrial robots. From automobile manufacturing to 3C electronics, from semiconductors to aerospace, Chinese companies can quickly deploy products into real scenarios for verification and iteration, forming a positive loop of “scenario–data–technology.”

Industry experts generally believe that the “Scaling Law” in the field of embodied intelligence has already become apparent: R&D demand is rising exponentially, and only top-tier companies can bear the high R&D costs and trial-and-error risks. In the next 3–5 years, the industry’s landscape will accelerate toward consolidation. Capital, technology, and scenario resources will concentrate on leading companies, ultimately forming a stable situation of “two heroes coexisting and differentiated competition.”

(Source: Caixin Global)