Text | BFT robot
01
core point of view
Core point of view: The development of the humanoid robot industry is still in the 0-1 stage. The current investment logic of the industry is biased towards event-driven theme investment, but the growth space of humanoid robots that can be implemented in service scenarios is very broad and deserves long-term attention. This article will discuss the following hot issues: D What are the changes in the humanoid robot industry at the current node? 2 How to judge its market space? 3 Which links in the Tesla humanoid robot industry chain specifically benefit my country? 4 What is the supply pattern and barriers in each link?
Marginal changes in the market: Al's development exceeds expectations and multiple players have entered the game. Pay attention to events such as Tesla AlDAY. We believe that the biggest marginal change in the market in 2023 compared to 2022 lies in: On the one hand, humanoid robots are the largest landing scenario of A+ machinery. ChatGPT access to robots is expected to make human-computer interaction more intelligent and diversified, and the industrial landing is expected to accelerate. On the other hand, there is an upsurge of humanoid robots around the world. Tesla, Boston Dynamics, Engineered Arts, 1X Technologies, UBTECH, Dyne, Xiaomi, Fourier Intelligence and other domestic and foreign companies have entered the market one after another. Son Masayoshi, founder of the venture capital firm Softbank Group, also expressed his strong interest in the AI revolution at the shareholder meeting in June this year.
Judgment of market space: The growth is clear, and the long-term value is expected to exceed that of electric vehicles. According to Goldman Sachs forecast, under the ideal situation of revolutionary breakthroughs in technology, the CAGR of humanoid robot sales in 2025-2035 can reach 94%, and the market size in 2035 will reach 154 billion US dollars. If we refer to the development history of electric vehicles, under optimistic conditions, the sales CAGR of humanoid robots in 2025-2035 can reach 59%. From the perspective of labor replacement, humanoid robots will gradually achieve mass production promotion from the B-end to the C-end within a two-year return on investment period. Tesla CEO Musk pointed out at the shareholder meeting in May this year that the value of humanoid robots in the long-term dimension will exceed that of electric vehicles.
Benefit link: Grasp the opportunities for Chinese manufacturing companies to benefit from the hardware of humanoid robots. Referring to the development of smart phones and electric vehicles, the Apple IOS system and Tesla FSD system are self-developed and controlled by the company, which is the core part of the software algorithm. In terms of hardware, outsourcing is usually used to meet cost reduction requirements. Therefore, Chinese manufacturing companies are expected to gain opportunities to benefit from the humanoid robot hardware industry chain by virtue of their obvious cost advantages.
Cost splitting: Tesla Optimus joint cost accounts for more than 51%, focusing on motors, leadscrews, reducers, and sensors. According to Tesla data, we assume that Optimus will use 14 rotary actuators, 14 linear actuators and 12 hand actuators respectively, and expect 3D rotary actuators: mainly composed of frameless torque motor + harmonic reducer / planetary reducer + torque sensor + encoder + bearing; 2 linear actuators: mainly composed of frameless torque motor / stepper motor + planetary roller screw / ball screw + torque sensor + encoder + bearing. 3 Hand actuator: mainly composed of hollow cup motor + planetary gearbox + ball screw + sensor + encoder. Combined with our expected price, the manufacturing cost of Optimus is about 22,500 US dollars, of which the total cost of joint actuators accounts for 51.42%, and the proportions of rotary joints, linear joints and hand joints are 20.87%, 22.41% and 8.13% respectively. Judging from the ranking of value and technical barriers, we can get:
Component cost ranking: frameless torque motor (10.34%), planetary roller screw (9.85%), torque sensor (7.51%), harmonic reducer (5.91%) encoder (5.54%), coreless motor (4.43%), bearing (3.02%).
Ranking of technical barriers: planetary roller screw, six-dimensional torque sensor, coreless motor, harmonic reducer, frameless torque motor, planetary reducer.
Marketization Outlook: The key to mass production is "usable" for application scenarios and "affordable" for large-scale cost reduction. Humanoid robots are still in the early stages of commercialization, and the pain points of marketization lie in the mining of rigidly needed application scenarios, large-scale cost reduction, and the establishment of industry standards.
Application scenarios : "Useful": Humanoid robots are expected to be applied first on the B-side. For example, Tesla's first batch of robots may land in its super factory, and the cost reduction & intelligence level will be improved and transferred to the C-side service scene;
Scale cost reduction: "affordable": Musk mentioned in AI DAY that the target price of Optimus is less than $20,000. We believe that Optimus is expected to become the first project in the humanoid robot industry to achieve mass production and commercialization through its powerful software algorithm capabilities, clear cost reduction goals and application scenarios.
02
AI empowers the humanoid robot industry to speed up, grasping the opportunity of my country's hardware revenue
2.1 How do you view changes in the 0-1 stage of humanoid robots?
In the 0-1 stage of the industry, the current changes lie in AI empowerment & multi-party entry & performance improvement.
The development of the humanoid robot industry is still in the 0-1 stage. On September 30, 2022, Tesla Al DAY first demonstrated the detailed parameters, development process and action capabilities of its humanoid robot, which aroused widespread attention from society, industry and the market at that time.
Change 1: Empowered by AI large models, humanoid robots become the best carrier of embodied intelligence. A large model empowers the humanoid robot industry in terms of language, vision, motion control, and reducing R&D costs.
(1) Language ability: The rapidly iterative ChatGPT has a language generation ability no less than that of a human, providing a solution to the robot's voice interaction problem. In April 2023, the artificial intelligence company Levatas cooperated with Boston Dynamics to connect ChatGPT and Google's speech synthesis technology to the Spot robot dog to achieve interaction with humans.
(2) Visual ability: With the blessing of the SAM model, the accuracy, recognition speed and zero-sample migration capabilities of machine vision have been revolutionaryly improved, which is expected to empower humanoid robots to adapt to more industrial and life scenarios.
(3) Motion control: The motion control system based on the AI model is still being iterated, and it is expected to be widely used in the balance and coordination control of humanoid biped robots in the future.
(4) Reduce R&D costs: AI virtual simulation testing greatly reduces testing costs, improves data collection efficiency, and accelerates product iteration cycles.
In addition, AI large models are also widely used in the fields of automatic decision-making, path design and planning of humanoid robots. As the best carrier of artificial intelligence, humanoid robots will fully enjoy the dividends of the rapid development of AI large models.
Change 2: The attention of industrial capital has increased, and multiple parties have entered the market to accelerate the layout.
Since 2021, overseas companies such as Tesla, Boston Dynamics, Honda, Engineered Arts, and 1X Technologies have released humanoid robot products or their iterations. Domestic companies such as UBTECH, Daxian Robotics, Xiaomi, Fourier Intelligence, and Puremi Technology are also gradually launching and updating their products. Correspondingly, many related component companies are also preparing to enter the robot field. Masayoshi Son, founder of venture capital firm Softbank Group, also expressed his strong interest in the AI revolution at the shareholder meeting in June this year.
Change 3: The latest version of Tesla Optimus was released in May, and the product performance upgrade function is complete.
In August 2021, Tesla publicly shared the humanoid robot concept Tesla Bot for the first time on the first AI Day, and then launched the first-generation prototype of TeslaBot with exposed wires in February 2022. Later, on the AI Day in September of the same year, it showed Optimus’s series parameters and related videos, including walking, watering flowers, and moving items in the factory. A humanoid robot driven by a highly actuated electric actuator. On May 16, 2023 (North American time), Tesla released the latest video of its humanoid robot at the 2023 shareholders meeting. We can see that its performance is more stable and its shape is more natural. The new changes mainly lie in:
Control ability: The robot demonstrates the ability to control the mechanical arm and the mechanical torque. It walks and grabs objects naturally, and can beat eggs without breaking them.
Environmental exploration and memory ability: The robot has inherited Vision A, installed the same unmanned driving system FSD, and connected to the same neural network as Tesla, which can more comprehensively and perfectly recognize the surrounding environment, explore the environment and generate memory.
Learning ability: The robot can learn actions based on human action demonstrations, and has upgraded end-to-end control capabilities to interact with various objects with manipulators.
2.2 Aging & rising labor costs, huge potential space for humanoid robots?
The population is aging rapidly and there is a gap in the labor force.
According to statistics from the National Health Commission of the People's Republic of China, the proportion of China's population over 65 years old reached 10.5% in 2015, higher than the world average of 8.4%. In 2020, the proportion of China's population over the age of 65 will rise rapidly to 13.5%, and the growth rate will also be higher than the world average during the same period. According to the forecast of the United Nations Department of Economic and Social Affairs, in 2035, the proportion of China's population over 65 years old will reach as high as 22.5%, entering a super-aging society. Over the same period, the global average will also rise to 13.2%, which is close to severe aging. Accordingly, the world is facing an increasingly serious problem of population aging, and the supply of labor will face a serious shortage.
Per capita wages are rising year by year, and labor costs are rising.
According to the data from the National Bureau of Statistics, China's per capita wages will rise from 90,500 in 2019 to 114,000 in 2022, with a compound growth rate of 8%, and the pressure on labor costs will rise rapidly. Assuming that the annual growth rate of China's per capita wages will still be 8% in the future, it is expected that China's per capita wages will reach 144,000 yuan in 2025. In the same period, if Optimus is put into mass production, according to Musk, the price will be about 20,000 US dollars, or about 140,000 yuan, which is basically the same as the per capita wage in China. Accordingly, humanoid robots are expected to occupy a cost advantage and achieve an increase in penetration rate in industry and service industries.
2.3 Under the cost advantage, grasp the opportunity to benefit from my country's hardware field
The humanoid robot industry chain has a certain degree of overlap with industrial robots, but there are big differences in upstream parts and downstream application scenarios.
Industrial robots mainly rely on industrial control systems to operate, and have a high degree of integration and intelligence; humanoid robots involve the processing of various complex scenes, so they require both strong motion control capabilities and powerful perception and computing capabilities. Like industrial robots, humanoid robots involve joint actuators, but the structure is more complex, and it also includes visual sensors, mechanical sensors, power batteries, interactive devices, etc. Downstream is mainly used in military, service, inspection, consulting, distribution and other fields and some dangerous scenes.
The cost of a humanoid robot mainly includes hardware and software.
Among them, the hardware mainly includes processors, drivers, actuators, sensors, etc., and the software mainly includes related operating systems, underlying algorithms, and development platforms. According to the degree of importance, the ranking from high to low is: FSD system > reducer > various motors > visual sensor + mechanical sensor.
In terms of software: the difficulty of humanoid robot software mainly lies in the hands and legs, which need to achieve refined operation and stable gait. Although the environment recognition algorithm has benefited from the emergence of SAM and other models, it still has difficulties. Tesla has the visual recognition learning algorithm of the FSD system and the powerful computing power supported by the D1 chip, and regards independent research and development as a strong driving force for its own development.
In terms of hardware: drive, transmission and perception are key technical areas. The drive device and the transmission device are the core components used to control the joint motion of the robot, so that smooth and high-precision joint motion can be achieved, involving precise motor control, torque transmission and position feedback technologies, including motors, reducers and lead screws. In terms of perception technology, humanoid robots need to use sensors to perceive the surrounding environment in order to perform tasks such as environment perception, human-computer interaction and obstacle avoidance, including visual sensors, force sensors, position sensors, tactile sensors, etc.
Due to cost advantages, Chinese companies are expected to fully benefit in the field of hardware.
At present, there is a certain gap between domestic and foreign manufacturers in the hardware technology of humanoid robots. However, with the technological progress of domestic manufacturers and the obvious cost advantages brought about by the comparative advantages of the manufacturing industry, referring to the division of labor in the global supply chain of Chinese companies in the smartphone and electric vehicle industries, we believe that Chinese companies are expected to gain opportunities to benefit from the humanoid robot hardware industry chain by virtue of their strong cost reduction capabilities and scale advantages.
Report source: Essence Securities
Report Editor: Intelligent Robot System
For more exciting content, please pay attention to the official account: BFT Robot
The copyright belongs to BFT Robot. If you need to reprint, please contact us. If you have any questions about the content of this article, please contact us and we will respond promptly.