Goldman Sachs's latest humanoid robot report: Focus on 2025WRC product iterations far faster than a few months ago!

The Zhitong Finance App learned that Goldman Sachs participated in the 2025 World Robotics Conference (WRC, August 8-11) on August 9 and communicated with 9 major humanoid robot companies (Astribot, Booster Robotics, Galbot, Kepler, LimX, Paxini, RoboTera, Spirt.ai, and Premium Choice, all uncovered private companies, arranged in alphabetical order). Goldman Sachs spoke with the CEOs, CFO, COO, Head of Research, Vice President of Product, and Capital Markets Manager of these companies and watched live product demonstrations.

New developments and key highlights of the exhibition

Compared with the 2025 Shanghai World Artificial Intelligence Conference (WAIC), this exhibition had significantly higher household/consumer visitor traffic. The latter focuses more on a wide range of artificial intelligence academic and industrial development topics. Humanoid robots are only a part of them, so there are many professional visitors; while the World Robotics Conference focuses entirely on robotics technology and product demonstrations.

Goldman Sachs believes this indicates that the demand potential for education, companionship, and entertainment is likely to be strong in the short term (in addition, demand for scientific research, artificial intelligence model training, government, developers, and stage performances). In the long run, this is an early sign: as long as robot products are practical, affordable, and safe, in addition to manufacturing demand, the potential for consumer demand may be very impressive.

Recently, Yushu released a new product, R1, priced at 39,900 yuan (height 121 cm/weight 25 kg/26 degrees of freedom); Engine.ai also announced plans to launch SA02 (height 125 cm/weight 25 kg/26 + 2 degrees of freedom) with a price of 38,500 yuan. Both products address the above requirements, and the product specifications are more simplified.

Regarding technological inflection points, the most positive point Goldman Sachs has heard is that it will take at least 1-2 years to scale up data, optimize reinforcement learning (RL) models, or transition to an end-to-end mode of operation after 1 to 2 years in a rules-based mode of operation. According to some opinions, this is mainly because the industry is still struggling to obtain enough high-quality real-world training data.

Goldman Sachs also learned that human multi-modal remote operation is a transition solution, but needs to ensure low latency (two companies have already mentioned Tesla's new restaurant in Los Angeles, where the Optimus robot is already in the popcorn service). Wang Xingxing, CEO of Yu Shu, defined in his opening speech that the robot's “ChatGPT moment” is to be able to complete tasks in a new environment with common skills. He believes that this moment will take at least 2 to 3 years, but not more than 10 years (news source).

Despite this, Google's new artificial intelligence video generation model Veo3, launched in May 2025, sparked a buzz among robotics companies. The model can generate and integrate audio (including dialogue, sound effects, and background music), and excels at generating text and image prompts, simulating real-world physics, and accurate lip sync. Its capabilities raise questions: is the underlying AI architecture likely to be better than the combination of the vision-language-motion (VLA) model and reinforcement learning (RL)? The latter has been viewed by some as the main artificial intelligence solution for robots and embodied intelligence since the beginning of this year. Therefore, Goldman Sachs emphasized that given the rapid co-evolution of software and hardware structures (the two are intrinsically linked in the development of artificial intelligence), investors should be wary of potential product redesign risks. However, the industry is progressing rapidly, with new breakthroughs every few months, which is encouraging.

It is worth mentioning that Nvidia performed well at the World Robotics Conference, delivering keynote speechs/seminars showcasing its strategic initiatives in the field of physical artificial intelligence (Physical AI) and general robotics, including cloud-scale model training, physics-based simulation (such as Isaac SimLab) for synthetic data generation, and real-time edge artificial intelligence deployment (such as Jetson AGX Thor). The company also recently launched Cosmos Reason, an advanced visual language model (VLM) designed to give robots common sense and the ability to understand real-world scenarios. According to Goldman Sachs, almost all robotics products still use the Jetson AGX Thor or the broader Jetson platform.

Before and after this exhibition (July-August), around 50 human-robot exhibitors released more than 20 new products (in addition, about 120 industrial robot and parts companies participated). Goldman Sachs sees the diversification of wheeled and biped robots (equipped with full rotation actuators or rotational-linear combination actuators). Compared to Goldman Sachs's February/May survey, overall product performance (full body control, WBC) improved significantly in terms of speed and fluidity. I was pleasantly surprised that all of the company's live demonstrations that Goldman Sachs observed worked so well.

On the application side, Goldman Sachs found four main directions:

a) Education, exhibition hall guidance and stage performances;

b) A standard platform for developers and scientific research/artificial intelligence training;

c) Manufacturing and logistics;

d) Consumer/Pension/Companionship.

Currently, a) and b) have contributed most of the sales volume; for c), the surveyed companies said that the success rate of specific tasks such as picking, sorting, assembly and inspection has reached 80 to 99.5%, and further improvements are mainly aimed at long-tail and abnormal situations; with regard to whether the robot operation speed (reasoning and execution) can match the customer's production line speed, the current pragmatic goal of robot companies is non-production line tasks to increase initial sales and meet the needs of future flexible manufacturing systems (FMS).

Goldman Sachs also learned that manufacturers value the efficiency of the entire system (such as human-robot collaboration) rather than the efficiency of a single robot versus a single human at the same task. One company stated that the customer considered an 18-month payback period to be a reasonable criteria for placing an order.

In terms of cost reduction, Goldman Sachs notes that these robot companies already account for a fairly high share of domestic parts (probably around 80%). This year, they also further reduced costs through design optimization, and some hardware specifications were lowered. In the future, scale effects will still be the key to reducing the cost curve.

According to recent government subsidy news, media reports say that during Beijing's “Yizhuang Robot Consumption Festival” (August 2-17), consumers will receive a subsidy of 1,500 yuan for each robot that costs more than 10,000 yuan, and that companies that buy a single robot that costs more than 5 million yuan will receive a subsidy of 250,000 yuan for each robot. The JD platform also offers online discounts: consumers and businesses can get 600 yuan off on purchases over 6,000 yuan, and 50 yuan off on purchases over 500 yuan.

Stock views

Some low-cost robot products have recently been launched or received sales subsidies, and some robot hardware specifications have been lowered. This may bring short-term sales opportunities to uncovered humanoid robot parts companies. Goldman Sachs maintained a “neutral” rating for Green Harmonics, Mingzhi Electric, and Zhongda.

In view of potential product redesign risks and component competition risks, Goldman Sachs is more inclined to give Sanhua Intelligent Control A/H shares a “buy” rating because its long-term positioning in the field of actuator assembly has high visibility (including customer relationships and technology iteration). Goldman Sachs is optimistic about Sanhua Intelligent Control's growth potential in the field of humanoid robot actuators based on a solid lead in its core business. It expects both revenue and net profit to grow at a compound annual rate of 19% from 2025 to 2030, and the target price for 12 months is HK$33.2/RMB 30.4, respectively.

Investment arguments, target prices and risks

Sanhua Intelligent Control (02050/002050.SZ, buy/buy)

Sanhua Intelligent Control is a global leader in HVAC control and thermal management components. Goldman Sachs gave Sanhua Intelligent Control A/H shares a “buy” rating because it has strong growth potential in the field of human-like robot actuators on the basis of a solid lead in its core business. The compound annual growth rate of revenue/net profit is expected to be 19% from 2025 to 2030. In terms of HVAC, Goldman Sachs believes that with the increase in commercial HVAC market share and the expansion of sensor products, the revenue growth rate is expected to surpass the residential HVAC industry. In terms of electric vehicles, the company's revenue will continue to grow, benefiting from an increase in the global penetration rate of electric vehicles and a moderate increase in bicycle supporting value. Furthermore, Goldman Sachs believes that humanoid robots are an important long-term technology trend, and Sanhua Intelligent Control is expected to occupy a key position in the supply chain as a highly visible actuator assembler. The catalysts include advances in humanoid robot function technology, mass production by major customers, and a further increase in the penetration rate of electric vehicles in Europe. Compared to the average valuation of Chinese industrial technology stocks covered by Goldman Sachs, the stock's valuation is reasonable.

Goldman Sachs gave Sanhua Intelligent Control H/A shares a 12-month target price of HK$33.2/RMB 30.4, based on an expected price-earnings ratio of 21 times in 2030, at a 9.5% cost of capital (CoE) discount until 2025.

Main risks: (1) increased competition in the field of electric vehicle thermal management; (2) Tesla's electric vehicle sales fall short of expectations; (3) weak real estate completion data in China is dragging down demand for HVAC.

Best (300580.SZ, neutral)

Goldman Sachs believes that with steady growth in the auto parts business (including fuel vehicles and electric vehicles), the company is expected to become a competitive supplier of humanoid robotic planetary roller screws (PRS). Given that the company is expected to enter and benefit from the high-specification humanoid robot supply chain (Goldman Sachs predicts a compound annual growth rate of 80% for global high-specification humanoid robot shipments from 2024 to 2035), Goldman Sachs uses a long-term valuation method for it. With accumulated precision manufacturing experience in fixture products and production capacity support from equipment procurement, Goldman Sachs expects Best to occupy 10% of the global PRS market for high-specification human robots starting 2027. At the same time, the commercialization of machine tool parts is progressing smoothly, and ball screws and linear guides have begun to be used by some domestic machine tool manufacturers in 2024. Despite this, compared to Chinese industrial technology stocks and human-like robot supply chain stocks covered by Goldman Sachs, Goldman Sachs believes that its risk-reward ratio is reasonable, so it maintains a “neutral” rating.

Goldman Sachs gave a 12-month target price of RMB 23.7, based on an expected price-earnings ratio of 32 times in 2030, and discounted 9.5% of the cost of capital until 2025.

Upside risks: (1) accelerated development of transmission components; (2) accelerated expansion of production capacity for electric vehicle parts. Downside risks: (1) turbocharger penetration rate growth is slowing; (2) profit margins are lower than expected due to competition or price pressure.

Green harmonics (688017.SH, neutral)

Green Harmonic is a local leader in the harmonic reducer market in China, and Goldman Sachs gave it a “neutral” rating. Goldman Sachs is optimistic about the expansion of the Total Addressable Market (TAM) for the company's harmonic speed reducers, as its application scenarios continue to expand at home and abroad, covering industrial robots, collaborative robots, human/service robots, CNC machine tools and other high-end equipment (solar/semiconductor/medical/aerospace). In 2024, the company began bulk supply to the Chinese factories of at least two of the top four international robot brands, and is expected to contribute more significant revenue in 2025. Although Goldman Sachs recognizes the company's long-term growth potential and gradual contribution to the humanoid robot business, its valuation is reasonable compared to the covered Chinese industrial technology stocks.

Goldman Sachs gave Green Harmonics a 12-month target price of RMB 112.9, based on an expected price-earnings ratio of 45 times in 2030, discounted to 2025 at 11.5% of the cost of capital.

Upside risk: (1) mass production of human-like robots and/or technological progress faster than expected; (2) penetration of core robot customers (especially overseas brands) is faster than expected. Downside risks: (1) Domestic industrial/collaborative robot demand falls short of expectations; (2) increasing competitive pressure from overseas and local brands.

Narushi Electric (603728.SH, neutral)

Mingzhi Electric is a domestic leader in motor and drive solutions, and Goldman Sachs gave it a “neutral” rating. Although there is a certain gap between its motors in terms of speed and acceleration parameters, Goldman Sachs expects to occupy a market share in the frameless motor supply chain for high/medium/low specification robots by 2027, with significant cost advantages and capacity expansion potential at the new Taicang plant. Goldman Sachs expects Mingzhi Electric's revenue to grow in the field of frameless motor applications for humanoid robots, and is expected to increase supporting value by providing modules (rather than single components). However, Goldman Sachs's main business outlook is conservative. The reasons include: 1) the servo motor market share growth trajectory is weaker than that of its peers; 2) the organic growth rate from 2014 to 2023 was 10%, and the M&A growth rate during the same period was 23%; 3) factory relocation/ expansion of overseas production capacity may disrupt operations. Considering the company's growth potential and main business (motors and drives) performance in the field of humanoid robots, Goldman Sachs believes that the risk-reward ratio is reasonable compared to the covered Chinese industrial technology stocks.

Goldman Sachs gave Mingzhi Electric a 12-month target price of 49.8 yuan, based on an expected price-earnings ratio of 37 times in 2030, at a 9.5% capital cost discount until 2025.

Upside risks: (1) humanoid robots are developing faster than expected; (2) the market share among original equipment manufacturers (OEMs) for humanoid robots is higher than expected. Downside risks: (1) demand in the terminal market falls short of expectations; (2) the motor and drive business market share is growing slower than expected.

Ling Yunguang (688400.SH, neutral)

In the rapidly growing machine vision sub-field, Goldman Sachs expects Ling Yunguang's market share in traditional machine vision systems to grow moderately, relying on solution capabilities to serve major customers and some self-developed software, cameras, and lenses. However, starting in 2025, Ling Yunguang's FzMotion motion capture system began increasing its exposure in the humanoid robot data collection business and became the second growth engine. The compound annual growth rate from 2025 to 2030 is expected to be 73%. As production of humanoid robots grows, Goldman Sachs expects the humanoid robot motion capture business to contribute 18% of revenue and 17% of net profit by 2030. Combined with other underperforming machine vision terminal markets (consumer electronics, batteries, solar energy), Goldman Sachs expects Ling Yunguang's overall revenue/net profit compound annual growth rate of 16%/24% from 2025 to 2030. Overall, compared to Goldman Sachs covering the stock average - 4% potential decline, the stock's target price has some room to decline, and the valuation is reasonable, so it maintains a “neutral” rating.

Goldman Sachs gave Ling Yunguang a 12-month target price of RMB 22.1, based on an expected price-earnings ratio of 30 times in 2026, discounted at 11.5% of the cost of capital until 2025.

Upward/downside risks: (1) the penetration rate of the new energy sector is faster/slower than expected; (2) the flat panel display (FPD) industry is growing better/worse than expected; (3) visual software development is progressing faster/slower than expected.