With the delivery of the first batch of K15, Shen Mingda, General Manager of Product Solutions at Wujie Power, temporarily breathed a sigh of relief. For several months, his schedule has been densely packed: Jiangyin in Jiangsu, Cangzhou in Hebei, Jinchang in Gansu... He led his team to run point by point, selecting industrial scenes where robots could land, and repeatedly verifying the concept of overseas operations. In order to obtain the industrial grade global CE certification known as the "European Industrial Visa", Wujie Power spent 8 months and millions of yuan. During these days, Shen Mingda has been worrying from beginning to end.

Now, the arrival of the first batch of K15 in France does not mean everything is fine. Our robots are fully capable of safe operations within a certain range, but there are still some issues if we need to cross workshops over long distances in the future. For example, when encountering people on the road, should the robot slow down or stop? To solve these problems, we need to study EU regulations, "said Shen Mingda.

Breaking through, upgrading, and overcoming again is a true reflection of China's embodied intelligence enterprises going global.

In 2026, it will become an important node for Chinese embodied intelligent enterprises to go global. The first quarter data shows that China's exports of humanoid robots increased by 210% year-on-year. Not long ago, the humanoid robot factory jointly built by Serbia and Chinese enterprises officially started mass production.

However, under the spotlight, embodied intelligent robots still face challenges in terms of technological iteration and commercialization routes.

Unlike the path of "first cultivating locally and then expanding overseas" for new energy vehicles, China's embodied intelligent robots began to enter the international market in the early stages. What should we do on this path of facing both opportunities and challenges in going abroad?

The consensus in the new energy vehicle industry that 'if you don't go abroad, you'll be eliminated' seems to have come earlier in the field of embodied intelligent robots.

The domestic market of this emerging track is gradually becoming crowded. According to data, by 2025, there will be over 140 domestic humanoid robot manufacturers and more than 330 humanoid robot products released. According to a report from market research firm SAG, in the first half of 2026, Chinese humanoid robot manufacturers accounted for over 97% of global shipments.

Among them, Yushu Technology and Zhiyuan split about 75% of the market share, and the vast majority of enterprises can only seize the remaining small market share. Due to the fact that the commercial scenario has not yet been fully implemented, the vast majority of embodied intelligence enterprises have not yet made profits, and many rely on financing to 'extend their lives'. Especially for companies that enter the market later, the survival pressure is very high. Unlike the domestic market, the overseas market is in the' trial and error 'stage for robots, and even in the entertainment consumption field, there is still considerable market space, "said an industry insider.

Lu Xizhe, the head of marketing and public relations at Youliqi Robot Technology (Suzhou) Co., Ltd., said, "Going abroad is not about chasing hot topics, but about calculating accounts. Some countries have difficulty recruiting workers, high labor costs, and a more urgent need for labor substitution. Customers have a higher willingness to pay

An industry insider revealed to reporters, "A humanoid robot can be sold for 200000 yuan in China, and in some countries it can be sold for 400000 yuan or even 800000 yuan

Early overseas expansion is not only related to short-term high premium orders. At present, the global embodied intelligent robot track has not yet formed a monopolistic international top brand, and the international market is in a brand blank period. If the early window period is missed, overseas technology giants may rely on computing power resources, algorithm advantages, and local scenarios to complete the global ecological loop first, which will suppress the generation gap in China's industries. Pei Xuan, a researcher at the Business Ethics and Social Responsibility Research Center of the University of International Business and Economics, believes that early layout of overseas distributors, experience centers, and benchmark projects can lock in brand awareness, build localized service networks, and form a first mover channel advantage.

In recent years, a group of embodied intelligent enterprises such as Yushu Technology, Zhiyuan, Yuejiang Technology, Fuliye, Accelerated Evolution, and Xingdong Era have actively positioned themselves in the global market. The overseas business contribution rate of companies such as Boundless Power has exceeded 50%.

There are two "trump cards" behind the collective overseas expansion of Chinese embodied intelligent enterprises: supply chain advantage and scenario resource advantage.

The advantage of China's supply chain is a 'flash sale' level, "said Xu Chunshan, founder and chairman of Zhejiang Smart Intelligent Technology Co., Ltd. This sentiment is not unfounded. This young enterprise, founded in 2024, created the first domestically produced high degree of freedom five finger dexterous hand in less than 10 months and won an overseas cooperation project. Putting thousands of components into one palm is impossible without industrial ecological collaboration. This requires joint efforts from enterprises in various links of the industrial chain, such as motors, tendons, sensors, etc., to tackle the goal of a lifespan of one million times, "said Xu Chunshan.

Multiple interviewees told reporters that in the field of humanoid robots, China has formed a complete manufacturing system from motors, reducers, sensors, controllers to complete machines, which can support rapid product iteration. Moreover, the underlying technology of new energy vehicles and humanoid robots is highly homologous, and many suppliers of new energy vehicle components are "crossing over" to supply humanoid robot components. Chinese robots used to rely on Japanese and other foreign suppliers to provide components such as sensors and joints. But now, these components are also manufactured in China. "The New York Times reported that" it is almost impossible to manufacture a humanoid robot without using Chinese made components.

Scene resources are equally crucial. From the landing of a batch of national artificial intelligence application pilot bases to the launch of the 2026 special action plan for real-life training of humanoid robots and embodied intelligence, a large number of innovative scenarios released by China are driving embodied intelligence robots to accelerate the start of their "operation mode". The massive industrial scenes in China have brought advantages in data iteration and mature mass production capabilities, and China's embodied intelligent robots as a whole have the foundation to seize the opportunity in the international market, "said Pei Xuan.

However, amidst the lively trend of going global, hidden concerns about the imbalance of market structure have emerged.

According to statistics from the international data company IDC, the actual shipment of humanoid robots worldwide in 2025 will only be about 18000 units, with over 85% going to scenarios such as entertainment performances, education and research, and guided tours that focus on technology verification and interactive display. Industrial manufacturing and warehousing logistics are still mainly in the pilot stage.

For example, according to Yushu Technology's prospectus, its revenue from overseas accounts for over 40%. Public information shows that Yushu Technology's overseas sales mainly flow to universities, research institutions, etc. At the 2026 World Robot Conference, an overseas agent of Yushu Technology told reporters that in Indonesia, currently more than 90% of customer inquiries come from fields such as entertainment and performance, education and research. A Swedish industry insider said in an interview with reporters that China's embodied intelligent robots have made rapid progress, but many still remain at the level of "showing off skills" and are still far from truly "working" on the production line.

The overseas research and education market is a 'low hanging fruit', but it is difficult to support true industrial level growth. If companies only sell robot bodies to universities and overseas laboratories, they will definitely encounter a growth ceiling in the future, "Pei Xuan analyzed. The way out is to quickly shift to B-end scenarios such as industrial manufacturing and special operations, and shift from selling prototypes to selling productivity.

However, there is a huge gap between future technological imagination and industrial reality: currently, humanoid robots have a task success rate of nearly 100% after sufficient training in fixed scenarios; Once the operating items are changed or the environment is slightly altered, the success rate significantly decreases. This means that the challenge of humanoid robots engaging in long-term work in complex environments remains unresolved. In the market for embodied intelligent robots, there is no shortage of demand or customers, but what is lacking is technology. Currently, overall, embodied intelligent robots lack stability, speed, and generalization ability in industrial scenarios, and technology is still lacking

This' qi 'is mainly lacking in the' brain '.

China's embodied intelligent robots have made significant progress in the "cerebellum" and "ontology" levels, but their "brain" capabilities still face many challenges. Pei Xuan believes that the hardware manufacturing capability determines the foundation for the mass production and landing of robots, while the core technology of the "brain" is the key to determining the commercial value, large-scale commercialization, and globalization of products. If substantial breakthroughs in 'brain' technology cannot be achieved quickly, China's strong hardware production capacity advantage will be difficult to fully unleash, and it is highly likely to fall into the dilemma of 'large export volume of prototypes and weak commercial landing', "Pei Xuan analyzed.

There is as much intelligence as there is data. The intelligence of the brain ultimately requires data. According to data, as of early 2026, the total amount of real physical interaction data available for global compliance is only about 500000 hours, while training a general embodied large model requires at least 10 million hour.

How can such a large gap be filled? Deploying robots at customer sites to collect scene data is crucial. Pei Xuan believes that embodied intelligent robots belong to highly non-standard products, and domestic scenarios cannot cover the differentiated working conditions of global manufacturing and service industries. Real deployment in multiple regions and industries overseas enables rapid accumulation of heterogeneous physical world data, reverse optimization of multimodal perception and general task planning algorithms, and solves industry pain points where laboratory prototypes are disconnected from real working conditions. Compared to companies that rely solely on domestic scenario iteration, manufacturers that go global in advance can complete reliability, load, and human-computer interaction capability verification faster, shortening the cycle from prototype to mass production models, "said Pei Xuan.

Shen Mingda deeply shares the same feeling. In his opinion, the most important thing is not to pursue the ultimate performance of the model before landing, but to "move forward in the mud of real scenes". A beautiful prototype demonstration is far from a commercial closed loop. It is important to 'sleep in the factory' and work with customers to solve some down-to-earth problems. It is difficult to deliver a perfect robot from the beginning, "Shen Mingda said for example." Our K15 robot initially did not meet customer requirements in terms of production pace and operational accuracy. The team and customers worked together on the real production line to achieve rapid performance ramp up. Only through repeated scenario verification, hardware modification, and algorithm adaptation can we promote the co evolution of the 'brain' and the 'ontology'. ”

In the process of going abroad, it's not about immediately stocking up on a large scale, but about first running the model scene. "Lu Xizhe believes that the work of supermarket tallying, hotel cleaning, building inspection, pre warehouse sorting, reception and delivery is highly repetitive, the process is relatively fixed, and the labor cost is high. For overseas customers, as long as embodied intelligent robots can stably complete some tasks, they have practical value. He suggested promoting embodied intelligent robot companies to "go global" together with other Chinese enterprises. For example, in the fields of hotels, property management, retail, and park operations, many Chinese companies have projects overseas, and embodied intelligent enterprises can land together with real scenarios.

Meanwhile, relying on simulation to create a "training ground" for embodied intelligent robots is crucial for breakthroughs in "brain" technology. However, a practical dilemma is that many embodied intelligence enterprises in China highly rely on overseas simulation systems, making it difficult to form independent and controllable research and development loops, which hinders the breakthrough progress of their "brain" capabilities.

This issue cannot be taken lightly. "Yang Haibo, co-founder and president of Lightwheel Intelligence (Beijing) Technology Co., Ltd., believes that simulation dependence is more covert and underlying than computing power dependence. Computing power determines whether the model can run, while simulation determines where the robot learns, follows what physical laws, and is evaluated using what standards. In the past, China has long relied on overseas systems in the field of industrial simulation. In the era of physical AI, if we still only use simulation tools, data formats, and evaluation systems defined by others, even if China's embodied intelligent robot industry has strong hardware, rich manufacturing scenarios, and application requirements, it will be difficult to take the initiative in the next generation of physical AI infrastructure. ”Yang Haibo said, 'We need to use our full stack self-developed simulation capabilities to participate in the joint construction of global physical AI infrastructure rules, and provide more independent and collaborative training and evaluation support for Chinese embodied intelligence enterprises to go overseas.'. ”

Yang Haibo further explained that whoever can complete task adaptation, stress testing, and risk identification through highly credible simulation before the embodied intelligent robot enters the real scene, and verify its stability, safety, and deployability in advance, is more likely to reduce overseas landing costs, shorten delivery cycles, and seize opportunities. Chinese companies need to prove that embodied intelligent robots can work stably before going global, "said Yang Haibo.

The departure of embodied intelligent robots to sea is a long-term obstacle race that tests intelligence, endurance, and perseverance, requiring crossing numerous levels.

Compliance is an unavoidable threshold. A company brought a embodied intelligent robot to a professional organization for overseas market compliance certification. After testing according to the standards one by one, there were too many non compliant areas. However, at this point, the robot had already been finalized and could not be rectified at all. We have seen too many situations like this, "said Dan Ming, General Manager of Industrial Machinery at T ü V Rheinland Greater China's Solar and Commercial Product Services Department. He used EU standards as an example to introduce that a humanoid robot not only needs to be certified in terms of mechanical instructions/regulations, wireless instructions, network security, electromagnetic compatibility, etc., but also needs to be certified separately for components such as power boards, joint modules, lithium batteries, and safety controllers.

The more you recite the 'Tightening Spell', the tighter it becomes. Huang Yu, Senior Manager of the Electronic and Electrical Products Service Department at T ü V Rheinland Greater China, reminds that the obligation to disclose vulnerabilities under the EU Network Resilience Act will soon come into effect. The bill requires manufacturers of products with digital elements to proactively report exploited security vulnerabilities and incidents to the EU authorities. Violating regulations can result in fines of up to 2.5% of global revenue or 15 million euros, significantly increasing the compliance costs for global digital products entering the EU market.

At present, the scale of embodied intelligent robots going global is generally small, and compliance issues have not yet been particularly highlighted. However, once the quantity increases in the future, it will definitely be targeted by overseas regulatory agencies. "Shan Ming suggests that overseas enterprises should prioritize compliance requirements, thoroughly study the regulations and technical requirements of export target countries, and integrate relevant rules from the design source to avoid detours in the later stage.

Liu Zhufu, Vice President and co-founder of Yuejiang Technology, has a deep understanding of this. Yuejiang Technology has ranked first in China's collaborative robot exports for 8 consecutive years, with products sold to over 100 countries and regions worldwide. None of our orders are obtained through low prices, "said Liu Zhufu. The certification system in the European and American markets is extremely strict, and Yuejiang Technology's experience is to pave the way with certification in advance, opening the door to overseas markets with dozens of international certifications.

At the same time, several interviewed experts called for promoting international standard mutual recognition at the national level to reduce the cost of repeated testing. Pei Xuan analyzed that when new energy vehicles go global, the global vehicle safety and energy consumption standard system has basically taken shape, and China can only passively benchmark and adapt. At present, international standards for the safety, performance, and human-computer interaction of embodied intelligent robots are still in the drafting stage. Relevant departments should actively promote the deep participation of domestic enterprises and research institutions in the formulation of relevant standards by international organizations such as the International Organization for Standardization, the International Electrotechnical Commission, and the International Telecommunication Union, integrate China's engineering practice and industrial chain technical parameters into global rules, and enhance the discourse power of standards.