Nothing can stop the progress of technology. Even if I don't do it, someone else will.

Author|He Tanming

Editor|Yang Ming

The development and competition of humanoid robots have entered a new stage.

On November 28, Tesla, a technology giant leading in the field of humanoid robots, released its latest research and development progress: In the video, the robot Optimus catches a tennis ball in mid-air with this "new hand," moving very naturally and smoothly.

Screenshot of Optimus catching a tennis ball, Source: X Screenshot

Moreover, NVIDIA has recently been rumored to be seeking to establish a humanoid robot supply chain in Taiwan Province, China. 1X technologies, which OpenAI has bet on, is also gradually maturing, with an increasing number of application scenarios. It can be seen that global giants are intensively deploying in the humanoid robot market.

There are also many positive news domestically. On November 15, Huawei's Embodied Intelligence Center signed memoranda of cooperation with 16 enterprises, establishing cooperation with multiple enterprises in the field of humanoid robots.

Source: Artificial Intelligence Application Innovation Achievement Conference

The development of humanoid robots is in full swing, with global internet giants continuously increasing their investment. Against the backdrop of weak domestic manufacturers, will Huawei's entry bring new vitality to domestic humanoid robots?

01 A millennium-old vision finally becomes a reality thanks to AI

'Humanoid robot' is a vision that spans a thousand years. The first to propose the concept of a 'robot' was Aristotle, traceable in historical records. The first to create a 'robot' in real life was Leonardo da Vinci, who designed a 'robot warrior' powered by wind and water. After that, Switzerland developed a humanoid doll that could write, and Germany developed a soldier robot that could play the trumpet... It can be seen that human exploration and development of robots have entered a preliminary stage.

Photo of the 'robot warrior' designed by Leonardo da Vinci

However, this stage lasted much longer than everyone expected. Even a few years ago, Atlas, developed by Boston Dynamics and capable of complex movements, was still in the preliminary stage of humanoid robots by definition. For robots, the criterion for entering a new stage is not the complexity of movements or functions, but artificial intelligence.

The most critical aspect of humanoid robots is intelligence. Before that, no matter how advanced the robots developed by Boston Dynamics were, as long as they were still limited to the scope of human manipulation, they were still in the preliminary stage of humanoid robots.

However, a significant event occurred in 2022: The release of ChatGPT, developed by OpenAI. Its level of intelligence not only revolutionized the internet AI industry but also became a key point for humanoid robots to leap to the next stage.

Since then, global internet giants have successively entered the humanoid robot market, making embodied intelligence a reality. Consequently, global giants have successively entered the humanoid robot market.

After 2022, companies such as UK-based Engineered Arts, Tesla, and 1X technologies in the United States rapidly rose to prominence due to their advantages in artificial intelligence. In contrast, established manufacturers such as Honda and Boston Dynamics were gradually overshadowed by the times.

Meanwhile, many domestic manufacturers have also begun laying out humanoid robots, such as Xiaomi's CyberOne and XPeng's AI robot Iron. However, for now, domestic humanoid robots are still far behind European and American products.

Why is there a significant gap between domestic manufacturers and top European and American manufacturers? An important reason is the fragmentation of technology.

Display of core components of Leju robots

It is no exaggeration to say that humanoid robots are a culmination of cutting-edge technologies, including drive, transmission, sensing, AI intelligence, energy, etc.

However, these technologies are very fragmented in China. For example, in the field of dexterous hands, Zhiyuan Robotics and Xingdong Jiyuan are relatively well-developed domestically, but relatively lacking in AI. Companies equipped with AI include CloudMinds, Fourier Intelligence, and Leju Robotics, among others. However, CloudMinds' robots suffer from excessive latency and high costs, making mass production unfeasible. Fourier Intelligence's robots are more suitable for the medical field, while Leju Robotics has imperfections in battery life, obstacle avoidance, and other functions.

The fragmented situation among domestic humanoid robot manufacturers has resulted in significant flaws in each manufacturer's humanoid robots, with no product being considered world-class.

Excessive fragmentation leads to a large number of humanoid robots with different functions and application scenarios on the market, which has certain benefits in terms of product pricing and consumer experience. However, it is not beneficial for the domestic industry ecosystem in the long run. Over time, domestically produced humanoid robots will gradually lose competitiveness, and only a few large manufacturers' humanoid robots will remain in the market.

02 Domestic manufacturers are fighting alone; what should Huawei do?

How can domestic forces be consolidated to create a top-tier domestic humanoid robot? Without Huawei, this difficult problem might take a long time to solve.

In the view of 'Jidian Shangye', Huawei's move was 'long planned'. As early as 2017, Huawei signed agreements with the University of Edinburgh and SoftBank to conduct research and development around AI+robots and registered patents in multiple robot-related fields. Subsequently, it also cooperated with companies such as CloudMinds and Leju Robotics.

If previous actions were merely providing technical support, signing memoranda of cooperation implies that Huawei may truly enter the humanoid robot market.

On November 15, 16 companies signed memoranda of cooperation with Huawei. Among the 16 companies collaborating with Huawei, Efort and Huawei have deep cooperation in embodied intelligence. Leju Robotics' products also use Huawei's HarmonyOS. There are also industrial robots specializing in different scenarios, such as Han's Robot, Topstar, and Zhongjian Technology, as well as upstream suppliers like Joyway and Hualong Xunda.

It is noteworthy that, among Tesla's Optimus robot partners, all but Joyway are component suppliers. Tesla has almost no cooperation with other robot manufacturers. This means that Huawei's humanoid robots can gain more manufacturing experience, and their final landing and service scenarios may even surpass Optimus.

However, it is worth noting that some relatively mature domestic humanoid robot manufacturers are not among those who have signed memoranda of cooperation with Huawei. Huawei seems to need ecological partners in various scenarios but not humanoid robots that have not been verified by the market. The reason is simple: Huawei may aim to become China's largest humanoid robot enterprise.

This means that, at least for now, Huawei does not intend to integrate the entire industry. Instead, Huawei's humanoid robots will compete with other domestic humanoid robots.

At this stage, there are not many humanoid robot products that have truly been put into production. Among them, Boston Dynamics' Spot is mainly used by GlobalFoundries for data collection for condition monitoring and predictive maintenance, radiation inspection in the Chernobyl region by UKAEA, and production and monitoring in factories. Agility Robotics' Digit is mainly used by companies such as Amazon, Ford, and GXO for handling and delivery work.

There are even fewer actual landing scenarios for domestic humanoid robots. Only Fourier Intelligence's GR-1 has entered mass production, previously used for greeting and reception at the Shanghai Expo. Zhiyuan Robotics' Expedition A1 and Xiaomi's CyberOne have not yet entered mass production.

If Huawei really enters the 'humanoid' market, how far can it go? Compared horizontally, surpassing Xiaomi, also a mobile phone manufacturer, does not seem difficult.

Whether in terms of technology or application scenarios, Xiaomi's CyberOne is actually a bit 'behind'. The first reason is that the AI is not intelligent enough. CyberOne is equipped with Xiaomi's self-developed AI, and the actual experience results of MiOS 2 show that Xiaomi is slightly lacking in cross-segment smart connectivity and AI experience. The second aspect is the lack of refinement. For example, in terms of hands, CyberOne's hands are simply a palm with a movable thumb. Although it can pick up things, it cannot perform more delicate hand movements, such as catching a tennis ball as demonstrated by Optimus, which may be challenging for CyberOne.

Product image of Xiaomi's CyberOne robot

Can Huawei achieve a technological breakthrough? The answer is obvious.

Although Huawei does not have its own product, it has provided technical support to Leju Robotics. The Kuafu MY robot, equipped with the Pangu large model, has a higher degree of anthropomorphism. It can not only perform various actions but also better complete tasks. In addition, due to the integration of the Pangu large model and the HarmonyOS ecosystem, the Kuafu MY robot's ability to autonomously perceive, understand, plan, and execute tasks has increased exponentially. After Huawei participates in the manufacture of humanoid robots, the experience of Kuafu MY can also be fed back to Huawei to achieve further breakthroughs.

03 Why is Huawei entering the market again, despite already being 'in the game'?

Here's a question: Leju Robotics is already doing well in humanoid robots, so why is Huawei entering the market?

Robot appearance at Huawei Developer Conference

For humans, the development of humanoid robots is significant. A mature humanoid robot can enhance the quality of our daily lives, such as in education, companionship, and entertainment in family life. It can also replace us in dangerous and high-precision work, such as exploration in dangerous areas like deep seas and radioactive environments, and help us explore new fields, such as space exploration. As humanoid robot technology matures and costs gradually decrease, their entry into human daily life is not a distant vision.

Currently, global internet giants have a consensus on the humanoid robot market, vigorously developing humanoid robots or investing in robot startups. For example, Figure AI, a company established in 2022, has received over $675 million in investment from companies such as Microsoft, Amazon, NVIDIA, and OpenAI. The newly established startup Skild has also received $300 million in investment. Domestic companies such as Baidu, Xiaomi, and Alibaba have also entered the market, participating in the research, development, and investment of humanoid robots.

GGII previously predicted that the global humanoid robot market size is expected to soar to $20 billion by 2030, and global humanoid robot sales will surge from 11,900 units to 605,700 units.

This explains why Huawei is entering the market despite Leju Robotics' relative maturity. For Huawei, whether for commercial purposes or based on the significance of humanoid robots for future life, its layout in humanoid robots, as well as that of other major manufacturers, is understandable.

In short, Huawei may not be a manufacturer, but humanoid robots will definitely be an important part of the future HarmonyOS ecosystem. Moreover, Huawei may bring new advantages to domestic humanoid robots.

The first is artificial intelligence. Currently, most AI stays at the stage of 'command issuance - robot execution'. Huawei's Pangu large model and HarmonyOS ecosystem may endow humanoid robots with true 'intelligence,' enabling them to act accordingly based on the current environment or sudden situations without receiving commands.

The second is Huawei Cloud. It not only allows users to better control humanoid robots but also enables system upgrades and storage of private content for humanoid robots through Huawei Cloud.

The third is industrial chain collaboration. With years of experience in upstream and downstream collaboration in the industrial chain, Huawei can quickly organize advantageous production capacity in the industrial chain, which is crucial for the mass production of humanoid robots. At the same time, integrating resources also has the benefit of reducing production costs, which can not only avoid the dilemma of GR-1 but also allow humanoid robots to better enter people's lives.

In general, the industrial chain has confidence in the competitiveness of Huawei's robots. However, objectively speaking, there are also several issues that Huawei needs to address when entering the humanoid robot manufacturing market.

The first is technical issues, including the accuracy and battery life of humanoid robots. For example, can a humanoid robot climb like a human on a mountain path? How long can a humanoid robot last on standby or maintaining high-speed movement? These are challenges that Huawei needs to address.

The second is social and legal issues. What is the scope of autonomous decision-making for humanoid robots? If a humanoid robot's program goes wrong, should the legal judgment be aimed at the enterprise or the consumer? This is a difficult problem for Huawei and the world as a whole.

These issues are currently difficult to resolve, but 'Jidian Shangye' will continue to observe this market and see how it changes after the mass production and landing of humanoid robots by Huawei or other companies.

Zhang Ping'an, Executive Director of Huawei and CEO of Huawei Cloud, said at the 2024 Huawei Developer Conference, 'As everyone expects, let AI robots help us with laundry, cooking, and cleaning, giving us more time to read books, write poems, and paint.'

With the advancement of technology, Zhang Ping'an's words will gradually become a reality. However, there is still a long way to go, and Huawei has a long road ahead in the field of humanoid robots.

END

Producer: Huang Qiangqiang