HomeArticle

No pressing need found: Is the humanoid robot business a good one?

第一新声2025-09-19 19:57
In the future, victory or defeat will no longer depend on who is more like a human, but on who can better solve human problems.

Skilled in dancing, performing, and fighting, today's humanoid robots are no longer confined to science - fiction movies but have truly emerged in real life.

The humanoid robot industry is undoubtedly the hottest track in 2025. In the first half of 2025, the global financing amount in the humanoid robot track soared to 14 billion yuan, among which the domestic market contributed 8.4 billion yuan. The proportion of Pre - IPO round financing in the global financing amount jumped to 30%.

At the upcoming Yunqi Conference, the latest service scenarios of humanoid robots will be exhibited, new ways of human - machine collaboration will be demonstrated, and an "athletics meet" for embodied intelligence will be held to showcase the achievements of embodied intelligence through multiple channels.

Driven by this boom, the market scale of humanoid robots is also constantly expanding. According to the prediction of GGII, the global sales volume of humanoid robots is expected to reach 12,400 units in 2025, with a market scale of 6.339 billion yuan; by 2030, the sales volume will approach 340,000 units, and the market scale will exceed 64 billion yuan.

Meanwhile, another voice has quietly emerged. Humanoid robots face challenges such as lack of profitability and limited application scenarios. There is still a long way to go before they can be truly put into production and daily use.

What is the real current situation of the humanoid robot industry? How can humanoid robots break through barriers such as dependence on imported core components and limited application scenarios in the future and find a sustainable commercial development path? In this issue, First New Voice attempts to clarify the core capabilities and existing pain points accumulated in the development process of humanoid robots from the front - line perspective of the industry.

01  Four Echelons Have Taken Shape 

Currently, the player camps in the humanoid robot track have gradually become clear. From the perspective of the competitive landscape, the global humanoid robot market has currently formed four echelons. Different manufacturers are exploring differentiated paths in the balance between human - like form and practical value based on their technological accumulations.

Industry leaders such as Boston Dynamics and Honda firmly occupy the first echelon. They have become industry pioneers with their profound technological accumulations, strong R & D capabilities, and remarkable industrialization progress.

Companies such as Ubtech and Agility Robotics are firmly in the second echelon. They have stood out with their unique technological routes or precise positioning of application scenarios.

Companies such as Unitree and Zhipu Robotics form the third echelon. Although they entered the industry later than the former, they have shown amazing development speed and late - comer potential.

Numerous startups such as ZeroPower Robotics and Dongyi Technology, as well as cross - border companies, form the fourth echelon. They bring new vitality and innovative thinking to the industry. Although they are temporarily behind in terms of technological strength and market share, their exploration in niche fields may give rise to new business models and technological breakthroughs.

Specifically for each manufacturer, in the North American track, Tesla is undoubtedly the most - watched game - changer. Since the release of the Optimus prototype in 2022, every technological iteration has affected the nerves of the industry. At Tesla AI Day in 2024, the second - generation Optimus prototype showed significant progress. It is equipped with a self - developed FSD chip and an end - to - end control algorithm, achieving smoother joint movements. The completion degree of actions such as single - leg squats and precise grasping of screwdrivers is close to the human level.

Boston Dynamics is also a benchmark in the technological field. Its Atlas robot has achieved high - difficulty actions such as parkour and backflips through the in - depth integration of hydraulic drive and dynamic balance algorithms and has long been regarded as the ceiling of industry movement. In 2024, the company announced that it would upgrade it to a fully electric system, further improving joint flexibility and mute performance.

Also worthy of attention is Agility Robotics. Its bipedal wheeled robot Digit has been deployed in 15 Amazon warehousing centers, with a global installed capacity of more than 120,000 units, becoming the first humanoid robot to achieve large - scale implementation in the logistics scenario.

In the Asian market, Japanese manufacturers are characterized by precision manufacturing, forming unique competitive advantages. As a leading enterprise in the industrial robot field, FANUC launched the Humanoid Robot CR - 5iA in 2023, which is more suitable for industrial scenarios rather than extreme human - likeness. This product adopts a modular design and can quickly replace end - effectors, suitable for precision industrial scenarios such as electronic component assembly and food packaging.

Chinese local players are characterized by rapid iteration and breakthrough in scenarios. Unitree is one of the first companies to achieve product commercialization. As a player extending from the quadruped robot track to the humanoid field, Unitree launched the H1 humanoid robot in 2023, continuing its advantages in the field of motion control. It uses a motor - driven solution, with a maximum load of 8 kilograms and a battery life of about 2 hours. It is the first mass - produced and delivered humanoid robot product in China.

Zhipu Robotics is a technological chaser in the track. The Yuanzheng A1 humanoid robot it launched in 2024 targets the mid - to - high - end industrial market. The Yuanzheng A1 is equipped with a self - developed Zhipu large - scale model, supporting natural language interaction and autonomous path planning. It can complete the whole process of identifying shelves, grabbing boxes, and placing them on the conveyor belt without pre - set programs.

In addition to the above leading players, a number of specialized enterprises focusing on niche scenarios have emerged in the domestic track. Deep Robotics entered from the bionic form. The Walker X humanoid robot it launched in 2024 is more human - like in appearance design. It is equipped with multi - modal sensors and can simulate facial expressions and gesture interactions. It is mainly used in performance scenarios such as science and technology museums and commercial exhibitions, providing a sample for the exploration of consumer - end scenarios.

It is not difficult to see that the current humanoid robot field has formed a competitive situation with multiple technological routes advancing simultaneously and multiple scenarios for trial - and - error. Foreign manufacturers have advantages in high - end technology and cost control with their large AI models and supply - chain integration capabilities. Chinese players, on the other hand, rely on the demand dividends of local industrial scenarios to rapidly promote product piloting and iteration.

However, all types of players face common challenges. In the future, victory may no longer depend on who is more human - like, but on who can better solve human problems.

02  Is the Humanoid Robot Business Really Profitable? 

The humanoid robot industry is experiencing the fastest market expansion in history.

Currently, humanoid robots have moved from the Demo display stage to the early stage of commercial implementation. The distribution of orders and procurement scenarios has initially shown the outline of industry applications, presenting a hierarchical pattern of large orders from leading enterprises and small - to - medium - scale orders.

In terms of order scale, there are significant differences among leading enterprises, ranging from hundreds to thousands of units. For example, Zhipu Robotics has mass - produced more than 2,000 units offline and is expected to mass - produce tens of thousands of units in 2026. Although Ubtech's contract amount reaches 400 million yuan, calculated at a unit price of about 600,000 yuan per industrial humanoid robot, the scale is less than 700 units.

In addition to large orders from leading enterprises, orders with a scale of less than 10 million yuan are still the mainstream in the current industry. These orders cover a large number and wide range of enterprises. The order volume of faster - developing enterprises can reach thousands of units, while more are in the scale of hundreds or even smaller. They together form an important foundation for the commercial implementation of the industry and show obvious characteristics among different enterprises.

Robosen and Xingdong Jiyuan belong to the high - end resource - based type. They not only have orders worth more than 10 million yuan or reserve orders in the scale of hundreds of units, but their customers also include well - known global technology giants, large state - owned enterprises, as well as high - end overseas customers such as Middle Eastern royal families and German factories. They lay the market foundation through a diversified high - end cooperation network.

Zhongqing Robotics and Jiasu Jinhua are of the scientific research and regional demand type. The order scale reaches hundreds of units, but the purchasers are mainly universities, scientific research institutions, and local state - owned enterprises. The demand focuses on technological R & D and regional industrial supporting.

Kepler, Kuawei Intelligence, Magic Atom, and Youliqi are of the regular commercial service type. Their order customers are concentrated in fields such as ordinary industrial enterprises, hotel properties, and commercial service institutions. The demand focuses on daily operation scenarios and is closer to the basic service needs of the market.

Looking at the background of purchasers, the purchasers of billion - level orders include industrial manufacturing enterprises such as those in the automotive and electronics industries, large state - owned enterprises, local state - owned platforms, universities and scientific research institutions, technology enterprises, and retail service providers, reflecting the potential for cross - industry and multi - scenario applications.

Among them, industrial manufacturing and warehousing and logistics scenarios have become the fields with the strongest implementation ability at present due to clear demand and mature budgets. Ubtech relies on the support of automotive industry chain enterprises and state - owned platforms, and its orders in core industrial scenarios are stable. Zhipu Robotics has entered core fields such as industrial intelligent manufacturing and the construction of national - level platforms and has a stable customer source.

The humanoid robot industry is not only booming in terms of order quantity. Data shows that in the first half of 2025, there were more than 80 global financing events in the humanoid robot track, with the amount exceeding 14 billion yuan, breaking the record of the whole last year. However, many institutions interpret it as a coexistence of high - heat and cold reality. Although money is flowing in at an accelerating pace, the mass - production ability is still scarce.

Currently, the capital enthusiasm is still rising, and giants and startups are competing on the same stage, but the frictions in the industrial chain are also clearly visible. The tension among computing power, algorithms, costs, and scenarios is determining whether humanoid robots can cross the key threshold from being able to move to being usable and being able to be implemented on a large scale.

Looking at the industrial development, when humanoid robots face more critical commercial implementation, the current pain points are concretized, mainly concentrated in four aspects:

First, they can be sold but cannot make money. From the overall market situation, university purchases account for a relatively high proportion in humanoid robot orders. According to the data of hundreds of orders disclosed by Unitree, although there are many university purchases, the single - order amount mostly hovers in the range of 100,000 - 500,000 yuan, and they are mostly used for scientific research verification and teaching demonstrations, making it difficult to form large - scale and high - profit commercial application scenarios.

Looking at the cost and selling price, core components such as lead screws and sensors still account for more than 40% of the production cost. Currently, the domestic supply chain has not formed a sufficient scale, and the scattered production capacity further drives up the unit cost.

According to Qichacha data, there are currently more than 800 domestic enterprises with humanoid robots in their business scope, but 70% of them mainly assemble imported components and lack independent core technologies. In contrast, Tesla's Optimus production line uses a modular production model, and its production capacity reached 500 units in Q2 this year. Large - scale production has reduced the unit cost from 150,000 US dollars in 2024 to 120,000 US dollars.

Fourth, there is insufficient technical collaboration. The problem of insufficient technical collaboration in humanoid robots essentially means that although the performance of a single module is outstanding, the mismatch between modules will still lead to a discount in the overall performance.

For example, in rainy and foggy environments, there are often delays in object recognition, and there are perception blind spots in the scenario of crowded shopping malls. This defect in the environmental adaptability of the perception system directly leads to a lag in the decision - making of the control system. Even if equipped with advanced algorithms, insufficient hardware performance will still expose the gap between the intelligence of large - scale models and the hardware execution ability.

The key to breaking the deadlock lies in overcoming these pain points. For industry participants, the current focus is to make robots continuously and reliably perform small tasks well in the real world. Whoever can enhance the "stability + reuse rate" in a vertical scenario will be able to enjoy the cost curve and scale dividends in the next stage.

03  The Way to Break the Deadlock is to Find the Rigid - Demand Scenarios 

Humanoid robots are not just about "looking like humans".

The core technical system of humanoid robots mainly includes five modules: mechanical structure, drive system, perception system, control system, and energy system. The technological breakthroughs and collaborative innovation of each module constitute the core driving force for the development of humanoid robots.

Currently, humanoid robots are at a critical turning point towards industrial implementation. This track is facing multiple tests. The way to break the deadlock does not depend on a single technological leap but on finding rigid - demand scenarios.

Whether it is the repetitive work in high - risk environments in the industrial field or the care needs for the aging society in the service field, only by anchoring real and unmet rigid demands can technological breakthroughs have a clear implementation direction and make the solution output more targeted. Only in this way can humanoid robots be truly promoted to large - scale industrial applications and achieve substantial breakthroughs in industry pain points.

First of all, to break the deadlock, it is necessary to clarify the service scenarios and customers and connect the capabilities of all ends.

Scenarios are the most crucial issue for product implementation. The breakthrough in service scenarios follows the path of precise entry. The intelligent pharmacy solution jointly launched by Galaxy Universal and Meituan enables the Galbot robot to complete the whole process of receiving orders, finding medicines, and bagging. In the actual operation of nearly ten stores in Beijing, the medicine - taking accuracy rate has increased from 85% at the initial stage to 98.7%.

The medical delivery robot jointly developed by Yuejiang Technology and Yaoshibang solves the refined demand for medicine sorting through visual recognition and semantic understanding technologies, increasing the sorting efficiency by 40%. Service robots can achieve a qualitative change from being able to do to doing well through data accumulation.

The scenario expansion in the industry is also presenting a clear gradient evolution strategy. The industry generally follows a gradual path from industrial workshops, commercial services, to home environments. Ubtech plans to deploy 500 simulated humanoid robots in the retail stores of Juranzhijia before the end of 2025. Through the interaction data accumulated in commercial scenarios, it will feed back the complex environment adaptability required for home scenarios.

The self - developed wheeled dual - arm humanoid - like robot Quantum 2 launched by Zibianliang Robotics has reached cooperation with leading service and industrial customers and has been put into use in multiple scenarios. This cross - scenario verification accelerates the improvement of technological maturity.

Secondly, it is necessary to build standardized service capabilities to match scenario requirements.

It is generally believed in the industry that the core of humanoid robot technology lies in the in - depth integration of large - scale models and embodied intelligence, further realizing the robot's autonomy and improving service capabilities.

Traditional industrial robots can complete tasks with high speed and precision in specific scenarios, but they highly rely on pre - set systems and are difficult to adapt to environmental changes. The introduction of the embodied intelligent large - scale model has broken this situation. This means that in a unified representation space, the model simultaneously processes perception, reasoning, and action, directly conducts cross - modal causal reasoning and action decision - making, and finally enables the robot to think and work like a human.

End - to - end large - scale model reasoning has become the key to technological breakthrough