Goldman Sachs research "douses cold water": The turning point for humanoid robot technology is still unclear, and practical applications will take at least five more years!

Humanoid robots, the "future warriors" that the tech industry and capital markets have high hopes for, will they really be able to quickly occupy factory floors and enter thousands of households as expected? A report from Goldman Sachs on the 27th may pour cold water on the overheated expectations.

Goldman Sachs analyst Jacqueline Du's team conducted an on-site investigation of Yushu Technology after the Spring Festival. The results show that humanoid robots still have a long way to go before they can truly "take up their posts."

At least in the next 2-3 years, humanoid robots will find it difficult to achieve the same work efficiency as human workers. Meaningful applications are expected to emerge only after 5-10 years. To achieve large-scale applications, multiple iterations in both software and hardware are required.

Goldman Sachs predicts that by 2027 and 2032, global shipments of humanoid robots will reach 76,000 units and 502,000 units, respectively. This growth rate is lower than market expectations, indicating that the commercialization process of robotic technology may be slower than many anticipate.

A recent report from Morgan Stanley also pointed out that although there has been rapid progress in motion control, achieving practical applications for humanoid robots still requires continuous improvements in algorithms and hardware. However, China's humanoid robot industry is showing signs of accelerated development.

"Can run and jump" ≠ "Can work," the commercialization path of humanoid robots is long

Yushu Technology's H1 humanoid robot was one of the focuses of this investigation. The H1 showcases powerful hardware performance, capable of walking, dancing, and possessing excellent self-balancing abilities. This capability is mainly attributed to Yushu Technology's rich experience in motion control accumulated during the development of robotic dogs.

However, behind the impressive movements lies a key issue: the H1 only has 19 degrees of freedom (DoF). This means it cannot handle complex and delicate tasks. For example, it cannot perform fine operational tasks like assembling parts or conducting complex household chores as humans do.

Goldman Sachs' report cites company comments stating that humanoid robots will find it difficult to achieve the same work efficiency as human workers in the next 2-3 years, but meaningful applications may only emerge over a 5-10 year span.

Additionally, Yushu Technology's humanoid robot currently only uses rotary actuators (planetary gear reducers) in its hardware design. This means that to achieve large-scale applications, multiple iterations in both software and hardware are necessary.

So, how far are we from household humanoid robots? Wang Xingxing recently mentioned in an interview with reporters from China Central Television:

"I personally feel that by the end of this year, the entire AI humanoid robot will reach a new level. If all goes well, by next year or the year after, some basic applications, such as in the service industry or some industrial sectors, can basically be pushed forward, but household applications may be a bit slower because they may have higher safety requirements and will still need the technology to be relatively more mature." In addition to humanoid robots, Yushu Technology's robotic dog products also showcased their technological strength during the research. Yushu Technology's robotic dogs are equipped with advanced gait control algorithms, enabling them to easily navigate complex terrains and hazardous environments. The company currently delivers thousands of robotic dogs each year, widely used for patrolling, inspections, and as substitutes for dangerous tasks such as fire rescue and inspections of substations and nuclear power plants. Yushu Technology holds a 60% to 70% share of the global quadruped robot market, with 50% of its revenue coming from overseas markets.

However, Goldman Sachs analysts pointed out that although robotic dogs perform excellently in specific scenarios, their application scenarios remain limited. While Yushu Technology's robotic dogs are equipped with LLM (Large Language Model) APIs and have voice control capabilities, they still cannot achieve fully autonomous operation. This means that the autonomous decision-making ability of robotic dogs in complex environments is still insufficient, requiring human intervention and guidance.

"Technological Iteration" ≠ "Singularity Arrival": Goldman Sachs and Morgan Stanley Cautiously Predict

Goldman Sachs' report noted that despite significant progress in robotics technology over the past few years, it is still insufficient to achieve multitasking and autonomous operation in complex environments. This indicates that the turning point for humanoid robot technology has not yet arrived.

Goldman Sachs expects that by 2027 and 2032, global shipments of humanoid robots will reach 76,000 and 502,000 units, respectively. This growth rate is below market expectations, indicating that the commercialization process of robotics technology may be slower than many anticipate.

In this context, Goldman Sachs believes that current investment opportunities are mainly concentrated in the components of the robotics supply chain. In particular, actuator assemblers and harmonic gear suppliers, due to their products having higher determinism and stronger technological barriers, will become the focus of future investments.

Previously, Morgan Stanley's research report also pointed out that despite rapid progress in motion control, achieving practical applications for humanoid robots still requires continuous improvements in algorithms and hardware, but China's humanoid robot industry is showing signs of accelerated development.

A key finding is that most Chinese manufacturers plan to achieve mass production of hundreds to thousands of humanoid robots by 2025.

Leading domestic players like Agibot are expanding their order scale to meet their aggressive production targets for 2025. Leju expects sales to reach 500-1,000 units by 2025, while MagicBot aims to sell 400 units in 2025, expanding to 15,000 units by 2027.