According to Securities Daily, Apple's machine learning research department recently unveiled a paper and video revealing the development of an AI lamp-shaped robot. Guo Mingqi, an analyst at TF International Securities, noted that Apple is exploring both humanoid and non-humanoid robot projects for its future smart home ecosystems.

However, Guo Mingqi predicts that Apple's related robot products are still in the nascent proof-of-concept (POC) stage and may not enter mass production until 2028 or beyond.

Nevertheless, it is undeniable that the robotic race is intensifying.

Huawei has long been involved in the realm of humanoid robots, while Unitree Robotics gained popularity during the Spring Festival Gala. Additionally, tech giants such as Tesla and BYD have also set their sights on this burgeoning field.

From Boston Dynamics to Honda, OpenAI, and Microsoft, the advent of the generative AI era has sparked a new wave of competition in robotics.

Robots are comprised of three key technology modules: motion modules, sensor modules, and artificial intelligence modules. Traditional robots often require expertise in just one of these areas. For instance, industrial robots primarily focus on motion control technology, whereas cleaning robots emphasize navigation and sensing technology.

Humanoid robots, however, differ significantly. They must be versatile across various application scenarios rather than being single-tasked in specific contexts. This makes them significantly more complex, requiring higher technology integration, fusion, modeling of larger datasets, and a stronger understanding of language and instructions. Prior to this, AI data and models were largely developed in isolation, leading to slow iteration speeds and high costs that were slow to decrease.

The prevalence of large models has drastically altered this landscape.

In just a few years, the number of model parameters has surged from billions to trillions. Consequently, large models are evolving from single-modal large models for text, speech, vision, etc., to general AI that integrates multiple modalities. This allows voice, vision, decision-making, control, and other technologies to be seamlessly integrated with humanoid robots, comprehensively enhancing their capabilities.

In April 2023, AI company Levatas collaborated with Boston Dynamics to integrate ChatGPT and Google's text-to-speech technology into the Spot robot dog, successfully enabling human interaction.

The rapid evolution of underlying technologies has demonstrated the potential for large-scale commercialization of humanoid robots, with major global technology companies actively making attempts and preparations. However, based on the current situation, humanoid robots still have a considerable journey before truly becoming a staple in households.

Firstly, the capabilities of humanoid robots currently available on the market are relatively limited and lack substitution benefits.

For example, household service humanoid robots do not yet possess the ability to fully replace existing human daily services, while humanoid robots in the field of commercial guidance and reception can only answer simple questions and cannot address all customer inquiries.

Given this, humanoid robots that lack rigid substitution are still not very appealing to C-end consumers.

It is also evident from the product positioning of participants that the primary focus of current efforts is still exploring B-end scenarios. For instance, the first batch of mass-produced Tesla bots are likely to be deployed in superfactories, while UBTech's focus is on exploring the application of humanoid robots in related industrial scenarios such as new energy vehicles and 3C electronics in collaboration with enterprises.

Secondly, numerous shortcomings in basic technologies have yet to be addressed. For instance, robot batteries need to support continuous operation for up to 20 hours, but most current humanoid robots can only work continuously for less than 2 hours. Another example is cost, which often hovers around tens of thousands of dollars, a price that obviously hinders widespread adoption. Manufacturing costs need to be reduced by 15%-20% annually in the future.

Humanoid robots may not be a lucrative business in the next three years, but within 30 years, it is an industry that cannot be ignored. From the industry's fundamental perspective, the value of humanoid robots lies in replacing high-cost human labor, an event that is highly inevitable.

Regarding current technology and cost issues, they will not pose long-term problems.

Taking costs as an example, as long as continuous commercialization can be achieved, any new technology and product will transition from high prices to low prices, as has been the case with computers, smartphones, electric vehicles, and so forth. In the past, the cost of single humanoid robots such as Honda's ASIMO and Boston Dynamics' Atlas was as high as $3 million and $1.9 million, respectively. Now Tesla can produce them for $20,000, and the price is bound to drop further in the future.

Technology follows a similar trajectory, continuously iterating and upgrading. Due to the immaturity of lithium battery technology, when Boston Dynamics launched the first-generation humanoid robot Atlas in 2013, it still required power cables. However, when the second-generation Atlas was launched in 2016, it utilized an independent lithium battery.

Musk predicts that the long-term demand for humanoid robots will reach 10 billion units. Even if only one-tenth of this expectation is ultimately met, the industrial space will still be incredibly vast.

With the largest population and manufacturing industry, China is already the world's largest robot consumer market, with robust demand from both B-end and C-end in the future. Moreover, China's software and hardware technologies in the field of robots are not significantly lagging behind. The combination of these two factors fundamentally determines that China holds great promise in producing a batch of robot companies that can compete globally.

The current leading domestic humanoid robot companies are UBTech and CloudMinds, founded in 2012 and 2015, respectively. Based on the current situation, compared to UBTech and CloudMinds, the domestic humanoid robot companies that will truly emerge in the future are likely to be major players such as Xiaomi and ByteDance. Major companies possess obvious advantages in terms of talent, funds, market, and brand. Furthermore, although companies like UBTech started earlier, they have not accumulated a strong leading edge. The revenue share of UBTech's humanoid robots is currently only in the single digits.

The visibility and profitability of upstream players are significantly stronger. Analyzing the material costs of robots, reducers, servos, and controllers account for 35%, 20%, and 15% of the cost of industrial robots, respectively, totaling 70%. Considering that humanoid robots have more joints and degrees of freedom, the proportion of these components may be even higher.

The reducer field boasts numerous players, including Han's Transmission, Leadshine, Tongchuan Technology, Zhongdali De, Guomao, etc., all of which possess certain production capabilities. However, there is only one leader: Green Harmonic Drive. The company has established a virtuous development cycle of "research and development - expansion - profitability - further research and development and expansion." As a manufacturing enterprise, Green Harmonic Drive has achieved a net profit rate of over 30%, which is no small feat.

The class differentiation in the servo motor field is relatively pronounced. High-end production capacity is primarily held by foreign companies such as Mitsubishi, Yaskawa, Fanuc, and Siemens. Huichuan Technology, Jiangte Electric, Jiangsu Leili, Leisai Intelligence, and Haozhi Electromechanical are concentrated in the low-to-mid-end field, with Huichuan Technology being the absolute leader. In 2022, Huichuan Technology's market share in the domestic servo field reached 21.5%, an increase of 5 percentage points compared to 2021. With national substitution efforts and increased robot production, Huichuan Technology's potential expectations are also relatively high.

In the controller field, domestic controller companies are highly fragmented. Although there are a number of professional controller companies such as Canopen, Wanxun Autocontrol, GookTech, Invt, and Haide Control, they have yet to form effective market competitiveness. The current localization rate is less than 20%, and whether they can emerge in the future remains to be seen.

In the long run, humanoid robots are a blossoming flower of tomorrow that will eventually reach its full potential. Short-term speculation is futile. What is required is long-term tracking and attention to identify those key enterprises.

Disclaimer

This article involves content related to listed companies and is based on the author's personal analysis and judgment, drawing from information publicly disclosed by listed companies in accordance with their legal obligations (including but not limited to interim announcements, periodic reports, and official interaction platforms). The information or opinions presented in this article do not constitute any investment or other business advice, and Market Value Observation does not assume any responsibility for any actions taken as a result of adopting this article.

——END——