UBS: Embodied robots are unlikely to see an "electric vehicle moment" within five years

Author | Zhou Zhiyu

Editor | Zhang Xiaoling

As the dawn of science fiction finally pierces the horizon of reality, humanoid robots, creations that once lingered in the writings of Isaac Asimov, are stepping out from the spotlight of laboratories and moving towards the bustling production lines of factories and future home lives in an unprecedented manner.

The endgame of this journey towards the "Robot Era" is undoubtedly enticing. UBS Securities recently released a report indicating that in the future, embodied robots will serve as both production tools in factories and optional consumer goods in daily life. In the baseline scenario, by 2050, the global stock of embodied robots will surpass 300 million units, with an annual demand reaching as high as 86 million units.

A market encompassing hardware, software, and service ecosystems worth $1.4-1.7 trillion is gradually unfolding, with an average annual compound growth rate exceeding 40% over the next twenty-five years. This is a figure that could drive any capital market into a frenzy, a massive wave comparable to the revolution of smartphones or electric vehicles.

However, as futurist Roy Amara stated, people tend to overestimate the short-term impact of technology while underestimating its long-term effects. Wang Feili, UBS Securities' industrial analyst in China, clearly pointed out at a media communication meeting on July 7 that although the long-term outlook is limitless, the market may need to lower expectations for the next five years.

A key inflection point is the "electric vehicle moment," where, after overcoming technological bottlenecks, sales surged from 1 million to 10 million units within five years. For humanoid robots, this moment is "unlikely to occur" within five years and may have to wait until after 2030.

This also means that the industry landscape for embodied robots will undergo intense reshuffling in the next five years. Wang Feili admitted, "We believe that looking at the next three to five years, the complete machine manufacturers will still face certain pressures, and it is currently very difficult to determine who will become the final winner."

The significant technological gap between ideals and reality is primarily constrained by the "brain" of the robots.

Wang Feili told Wall Street Insights that the industry currently faces two major bottlenecks: first, the lack of a mature, specialized AI general model for robot training; second, the extreme scarcity of high-quality training datasets, which are very limited in terms of width, breadth, and quantity.

To illustrate this gap more vividly, UBS creatively introduced a grading system for intelligent levels, benchmarked against automotive autonomous driving, ranging from L0 to L5. If L5 represents a "complete entity" capable of autonomously completing multiple tasks in any scenario, then the vast majority of current robot demonstrations still remain at the L0-L1 stage, which requires human intervention or pre-programming. The industry's goal for 2025 is merely to advance to L2, where robots can partially operate autonomously in standard scenarios.

What does this mean? Automotive autonomous driving achieved L2 in 2020, and it is still waiting at the threshold of L2.99 for regulatory improvements. This indicates that the journey for robots to transition from L2 to higher levels of general intelligence will be a much longer road The challenges at the software level primarily stem from the lack of a mature, robot-specific "general artificial intelligence model."

Current models are far from being able to enable robots to make autonomous decisions and generalize operations in new environments. Accompanying this issue is the extreme scarcity of high-quality training datasets. Unlike autonomous vehicles, which can easily obtain vast amounts of road data, the data collection for robots regarding their actions and interactions is difficult and limited in quantity, severely slowing down their pace of intelligence.

According to observations made by UBS analysts after visiting automotive manufacturing plants, they found that even simple tasks like moving boxes, which are considered the first practical scenarios, currently see robots performing at "only about 30% of human efficiency."

Meanwhile, the "bodies" of robots are also far from ideal, with hardware constraints being particularly prominent. This constraint is first reflected in the core "joints" that determine their action precision and stability. In components such as planetary roller screws, which require extremely high precision, and six-axis torque sensors that serve as the core of perception, domestic manufacturers still lag behind leading overseas companies in terms of technology. The high-performance SoC chips that drive the entire "brain" also currently rely on imports.

However, Wang Feili also expressed confidence that as Chinese companies continue to develop their technology, this bottleneck can gradually be resolved in the future.

Due to the numerous limitations of both software and hardware, UBS believes that the "electric vehicle moment" for embodied robots is unlikely to occur within the next five years, and may not arrive until after 2030.

Currently, embodied robots can only choose to enter from the most realistic and fundamental scenarios. Wang Feili pointed out that commercialization will follow the order of "industrial - service industry - household," and only with substantial applications in industry and services, including future entry into households, can it be said that embodied robots have truly achieved large-scale commercialization.

UBS predicts that a cost reduction of over 70% will be necessary to truly stimulate large-scale demand, especially in the consumer market, which is more sensitive to price than enterprise clients.

For investors, in this long marathon, grasping the rhythm and betting on the right segments at the right time is the key to success.

UBS provides a clear investment clock: in the short to medium term (the next 3-5 years), the upstream of the industrial chain will benefit first. While complete machine manufacturers are still under financial pressure from huge R&D investments and ecosystem building, the first wave of dividends will flow to those suppliers providing the core "bones" and "senses" for robots.

Among these, the automation components with the highest value proportion are undoubtedly the biggest winners, especially the planetary roller screws that constitute the robot's precise "joints" (accounting for about 14% of material costs) and various sensors that endow it with perception capabilities (totaling about 27%). In addition, the rare earth permanent magnet motors, batteries that provide its power source, and the semiconductor industry that gives it computing capabilities will all gain significant incremental markets in the wave of embodied robots.

In the long run, when technology matures and the market explodes, the industry's dominance will ultimately return to those midstream complete machine manufacturers that can define products and master ecosystems. Until then, the "water sellers" in the upstream of the industrial chain are undoubtedly the most stable winners in this gold rush A brand new trillion-level track is being opened, but the road to the finish line will not be achieved overnight. For all participants, what is needed now is not only imagination but also the patience and foresight to traverse cycles.

Risk Warning and Disclaimer

The market has risks, and investment requires caution. This article does not constitute personal investment advice and does not take into account the specific investment goals, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Investment based on this is at one's own risk