The core of the review of the major events of humanoid robots in the past week is the leadership of technology giants.
1.On February 24, 2024, humanoid robotics startup Figure AI raised about 6$7.5 billion, with a pre-investment valuation of about $2 billion. Currently, Amazon founder Jeff Bezos, Nvidia and other big tech companies are investing in Figure AI: Bezos has committed $100 million through his company Explore Investments LLC, Microsoft will invest $95 million, Nvidia and Amazon affiliates will provide $50 million each, Intel's venture capital arm will contribute $25 million, and LG Innotek will provide $8.5 million. At the same time, Samsung Investment Group pledged to contribute $5 million. OpenAI, which was once considering buying Figure, will invest $5 million.
Founded in January 2022, FigureAI focuses on the design of autonomous general-purpose humanoid robots. Its founder, Brett Adcock, has been known for 1The $100 million sale of a recruitment company allowed the $2.7 billion EVTOL company to ring the bell and be called "the next Elon Musk." He put $10 million in the first round and an additional $20 million in Series A. Prior to the latest round of large-scale investment from leading companies such as NVIDIA and Amazon, Figure has received $70 million in Series A financing led by Parkway Venture Capital in April 2023, with participation from FJ Labs, Moving Capital, Tamarak Global, Bold Ventures, and Aliya Capital Partners. Three months later, it received $9 million in Series A+ financing from Intel Capital.
Most of the company's employees come from well-known companies such as Tesla, IHMC, Apple, Cruise, etc., who want to build the world's first commercially viable humanoid robot and join the human workforce. According to its model planning, the robot is about 168 meters, weighs 60 kg, has a payload of 20 kg, is fully electric, and can run for 5 hours after charging. These robots will be autonomous, able to automate repetitive (or unskilled) tasks, and look and behave like humans, and can be used in everything from warehousing to retail and even caregiving. On January 7, 2024, Figure released ** showing that its product Figure 01 has demonstrated the ability to autonomously complete tasks such as making coffee. On January 19, Figure announced that it had signed a commercial agreement with BMW to deploy Figure 01 at BMW's U.S. manufacturing facility.
2.On February 24, 2024, NVIDIA established a new research division, "Gear" (Generalist Embodied Agent Research). Led by Professor Jim Fan and Prof. Yuke Zhu, Senior Research Scientist for AI and Head of the AI ** Program, the new department focuses on four key areas: multimodal basic models, general-purpose robotics research, basic agents in virtual worlds, and simulation and synthetic data technologies, and is committed to leading the way in future technology waves and driving the continuous progress and development of the entire AI and robotics industry.
"Embodied intelligence" is intelligent systems that understand, reason, and interact with the physical world. Currently, Gear has spearheaded the EUREKA project (human-level reward design by coding large language models), the Voyager project (an open embodied agent driven by a large language model), the VIMA project (a multimodal prompt for general robot operation capabilities), and the Minedojo project (an open embodied agent built with an Internet-scale knowledge base).
3.On February 20, Nvidia announced that it will host GTC2024 at the San Jose Convention Center from March 18 to 21. CEO Jensen Huang will unveil the latest breakthroughs in accelerated computing, generative AI and robotics. Among the 12 heavyweight guests listed by NVIDIA, three are bigwigs in the field of robotics, including Vincent Vanhoucke, distinguished scientist and senior director of robotics at Google Deepmind, Aaron Saunders, chief technology officer of Boston Dynamics, and Dr. Feifei Li, professor of computer science at Stanford University. Companies such as Agility Robotics, Boston Dynamics, Disney and Google Deepmind will also exhibit 25 robots, including humanoid robots, industrial manipulators and more.
4.OpenAI has provided support to 1x Technologies, a Norwegian humanoid robotics company. During the Chinese New Year, 1x Tech tweeted** that based on a scenario from a logistics center, their robots are already able to work autonomously. 1x indicates that each of the behaviors in the ** is controlled by a vision-based neural network that emits actions at a frequency of 10Hz. The neural network consumes images and emits actions to control the driver, arms, grippers, torso, and head. It does not include remote operation, computer graphics, editing, acceleration, footstep track playback, all controlled by neural networks, all autonomous, and all times the speed;
5.On February 24, Musk tweeted a ** showing the humanoid robot Optimus hanging out in the lab. It can be seen that the Optimus robot walks very smoothly. Optimus' team of engineers also says that in the latest **, Optimus has reached a walking speed of 06 meters per second or 134 mph. Since the demo in December, the speed has increased by more than 30%. This is mainly due to their improved vestibular system, foot trajectory, and ground contact logic, upgraded motion planners, and reduced cycle latency throughout the robot, while also adding slight torso and arm sway, making the Optimus more stable overall.
Chief Mechanic Li Zhe:
The logic of the industry megatrend has not changed, today NVIDIA said that it will release the latest breakthroughs in the field of accelerated computing, generative AI and robotics on March 18, 2024 GTC, and Google Deepmind employee Twitter said that the application of AI in robots will also have big results disclosed in the near future (Google also participated in the NVIDIA conference on March 18).
Robots are still the core scenario sector of the AI application layer: Previously, everyone paid attention to the Meta model in the AI field, OpenAI's 1X robot**, and Tesla and Sora generation capabilities.
Ideas for stock selection and sector rotation:
Recently, the screw segment has performed the best, mainly due to.
1) When the short-term catalyst is the domestic first-class chain audit factory, the probability of industrial progress in the screw plate is the largest, and the core target with the deepest decline in the screw plate is found.
2) When the short-term catalysis is expected to be the progress of robot algorithms, the sensor sector of technology mapping logic performs best.
3) After the screw company takes the lead in leading the rise, the good performance is often the processing equipment lathe, the grinding machine company (turning + grinding process), and the final weight-reducing special plastics (not only PEEK, PEEK is only a possible path).
4) The core target or diffusion logic of the low-level domestic ** chain.
Related industry chain companies:
T1: Sanhua Intelligent Control, Tuopu Group, Mingzhi Electric, Green Harmonic, etc.
Sensors: Donghua Testing, Huapei Power, Lingyun Co., Ltd., Keli Sensing, etc.
Screws: Best, Wuzhou Xinchun, Beite Technology, Hengli Hydraulics, etc.
Equipment Special plastics: Zhejiang Heideman, Huachen Equipment, Tanaka Seiki, Zhaomin Technology, etc.
Possible diffusion logic low valuation targets: Dongfang Precision, Neway CNC, Guosheng Zhike, Huaxiang shares, Shuanglin shares, etc. (equipment renewal and replacement policy + new quality productivity); Domestic companies with low stock prices: Buke shares, precision forging technology, Zhongda De, etc.
Risk warning: the risk that the R&D and expansion of new robot products are less than expected; The risk of uncertain technical route of robot parts.
Strategy: First, the market has bottomed out, but what we really need to pay attention to is the switch of the medium and long-term main line.
After a sharp ** from late January to early February, the market has seen a significant ** since February 6, with the Shanghai Composite Index recovering 3,000 points for 8 consecutive days, and the main broad-based indices have risen by more than 10% since February 6. From a structural point of view, TMT leads the whole market, and it cannot be ignored that the resource industry, which has a strong performance in the early stage of the market, has not fallen in this round, and coal and petroleum and petrochemical are still more than 10%. To a certain extent, this has broken the cognitive expectation of a large number of investors that "the market will enter the market and the style will also be switched". Long-term fundamental changes between different sectors have begun to be reflected in market pricing, which cannot be explained by a simple historical experience of "style switching". Under the background of "de-financialization", the repair of the economy comes more from the repair of physical activities than from the expansion of profits under financialization, and the pattern of net profit growth of listed companies will exist for a long time. Therefore, the pattern of inter-plate dominance will not change. Therefore, the order of recommendation for the long-term main line of the market has not changed from the previous review "Embrace Resources, Beyond Dividends". In addition, the thematic investment opportunities of robots are also one of the recommended directions in our previous annual strategy "Noah's Ark". It is worth noting that the robot theme index also significantly outperformed the market in this round**, with the robot index **28% since February 6, outperforming the market by about 14%. Next, let's interpret the investment opportunities of VisionNav Robotics from a strategic perspective.
Second, the strategic perspective of the robot investment opportunities.
First of all, on the whole, robots are actually the best combination of AI + manufacturing, and China's advantage over other countries in the world lies precisely in the manufacturing industry: whether it is the proportion of high-tech product exports, the development index of manufacturing power or the resilience index of the industrial chain, it all indicates our position in the world's top two in manufacturing. This makes it very likely that the application field of AI in China will be developed from the field of to B or to G instead of to C, and the regulatory and policy risks faced by AI+ manufacturing are also lower in comparison, and enterprises in the field of robotics may take the lead in completing the realization from capital expenditure to revenue to corporate profits.
Secondly, from the perspective of the industrial chain, compared with the original industrial robots, humanoid robots have a relatively low penetration rate due to complex structure and high cost, but this also indicates a huge market space in the future. At the same time, because the number of hardware and accuracy requirements required by humanoid robots are much higher than those of ordinary industrial robots, this also means that humanoid robots have a longer and wider industrial chain, and the value of the industrial chain that can be created is greater. At present, the localization rate of the industrial chain represented by the reducer and servo system in China has also been improving, although the absolute value of the proportion is still low, but it also means that there is a huge space for substitution in the future.
Finally, from a longer-term perspective, robots are likely to be the key to helping humans get out of this round of "stagnation". This round of resource supply contraction is difficult to quickly reverse in the long-term context of energy transition, and de-globalization has just begun, which will further hinder the efficiency of resource allocation. At the same time, aging has become a problem that most manufacturing countries and resource countries have to face, which leads to the rise of labor bargaining power and wage rigidity, which may eventually form a cycle of wages - inflation expectations forcing goods to rise - continue to raise wages. In order to break this cycle, the transformation to capital-intensive automation and intelligent industries is a more likely path. Empirical studies have also shown that if we take into account the 1993-2007 period between different regions of the United States, the addition of one robot per 1,000 workers would reduce the number of employed people by 034%, the level of wages fell by 05%。In fact, the world's manufacturing powers have taken action, especially for China, and the further upgrading of high-end manufacturing is also urgent and necessary.
Automotive: Nvidia's layout of embodied intelligence + ** business visit is imminent, optimistic about the robot sector.
Hot Events:1From February to 24, NVIDIA announced the establishment of a new research department - "gear", the full name of which is universal embodied agent research, which will focus on 1) multimodal basic models; 2) Research on general-purpose robots; 3) basic agents in the virtual world; 4) Simulation and synthetic data techniques.
2.From March 18 to March 21, NVIDIA will hold the GTC conference in 2024 to announce the latest breakthroughs in accelerated computing, generative AI, and robotics. At the conference, companies such as Agility Robotics, Boston Dynamics, Disney and Google Deep Mind will exhibit 25 robots, including humanoid robots, industrial manipulators and more.
Industry changes: Robotics is a combination of AI, sensors, and actuators, and NVIDIA has established a research department "Gear" to realize intelligent applications across multiple modalities and scenarios, which will accelerate the development of intelligent agents that can adapt to various environments, have a wide range of skills, and can operate effectively in the virtual and real worlds, and promote the continuous progress and development of the entire artificial intelligence and robotics industry.
We judge that 2024 will be the year of rapid breakthroughs in robot software technology, and it will also be the first year of mass production of humanoid robots.
Valuation: Affected by factors such as the impact of ** and Tesla's financial report in late January (AI Day postponed, mass production may be postponed to 2025) and other factors, the robot sector has been significantly adjusted, and there has been a ** recently, but the valuation is still at a relatively low point. At present, the valuations of the core targets in 2024 are: Tuopu 27x, Best 29x, Beite Technology 40-50x, Shuanghuan 21x, and precision forging 18x.
Higher decline than before: Compared with the high point in the early stage of 2023, the overall decline of the robot target is -20 30%, such as Beite Technology -16%, Sanhua Intelligent Control -21%, Tuopu Group -29%, Best -30%, Shuanghuan Transmission -34%, and Precision Forging Technology -37%;
Lower increase than before: Compared with Tesla's financial report on January 25, 2024, the overall increase in robot targets is +10-20%, such as Best 23%, Sanhua Intelligent Control 17%, Tuopu Group 13%, and Beite Technology 12%.
Considering the certainty, valuation, and location, it is recommended to lay out the underlying capabilities with strong technical capabilities, close cooperation with T or Tier1, and strong fixed-point certainty
1) Tier1: Tuopu Group, Sanhua Intelligent Control;
2) Lead screw & equipment: Beite Technology, Best, Wuzhou New Year;
3) Reducer: double-ring transmission, precision forging technology.
Risk warning: the mass production progress of robots is not up to expectations, the progress of large models is less than expected, and the localization level of the industrial chain is not up to expectations.
Electric new: key recommendation: Sanhua Intelligent Control, Mingzhi Electric and Flextron Electric.
1. Sanhua Intelligent Control: The profitability of the main business is strong and the robot rhythm is steadily advancing.
The profitability of the main business is strong. 1) Refrigeration business: In 22 years, the company's global market share of electronic expansion valves and four-way reversing valves has exceeded 50%, and with the acceleration of the industrialization of air source heat pumps and energy storage thermal management, related components are expected to provide new growth momentum for the company; 2) Auto parts business: In 22 years, the company's global automotive integrated components accounted for 60%, and the current orders in hand are sufficient, with the accelerated penetration of new energy vehicles at home and abroad and the development of new fixed points, the revenue of the auto parts business can maintain a compound growth rate of 50% in the past three years.
Forward-looking layout in the field of robotics and open up a new growth curve. In January 24, the company announced that it planned to invest no less than 3.8 billion yuan in the construction of robot electromechanical actuators and domain controller projects. The company has started the investment and construction of the robot production base, and the process of robot industrialization is advancing smoothly.
2. Mingzhi Electric: Pay attention to the company's three core changes (capacity, strategy, and downstream).
Capacity: The release of the new base is expected to break through the production bottleneck. In the early stage, due to capacity constraints, it faced the problem of scale growth, and the production-to-sales ratio of its control motor and drive system was close to 1:1, and the production plan was in a saturated state. At present, the layout of the new production base has been basically completed, and at the same time, automation technology and new capacity planning strategies have been introduced, so that the company is expected to break through the production bottleneck and further increase its market share.
Strategy: Expand product categories & improve overall solutions. Coreless motor: 23H1 increased by 67% year-on-year, mainly due to the steady release of production capacity and the business growth of related products in the application fields of mobile robots, medical devices and biochemical analysis instruments; Precision transmission business: 23H1 achieved operating income of 44.1 million yuan, a year-on-year increase of 29%; Motion control solutions: 23H1 total operating income of 34.8 billion yuan, a year-on-year increase of 12%.
Downstream: The business of emerging high value-added application areas maintained growth. The company's motion control business has achieved rapid growth in emerging high value-added applications such as photovoltaic equipment, industrial and factory automation, intelligent driving vehicles, mobile robots, and intelligent pump and valve control.
3. Flextron Electric: The main business is growing rapidly, and humanoid robots have opened up new space.
The company's main business performance is outstanding, and new products and new markets are growing rapidly. 1) New products: 3Q23 servo system and control system revenue exceeds 300 million yuan, YOY+70%; Servo and small PLC products have been sold in batches, medium-sized PLCs are also under development, and the full-level product line has been opened. 2) New markets: It has laid out markets such as India, the Middle East, and Southeast Asia, and achieved rapid breakthroughs with the help of general products such as the photovoltaic field, and the company's overseas revenue in 23H1 was YOY+174%, continuing to grow rapidly.
Humanoid robots have been laid out for a long time and enjoy the dividends of track growth. In 22 years, the robot industry department has been established, and at present, DC brushless coreless motor products have been launched, mainly for robots and medical fields, and have been tested in the medical field; Special series of low-voltage servo robots, frameless motors, servo integrated wheels, shaft joint module products, etc. have been launched, and a number of products are currently delivered in small batches.
Risk warning: the risk that downstream demand is less than expected; The risk that the market development is not as expected.
Computer: Computer: The embodied intelligent large model completes the last piece of the puzzle for the robot operating system.
The embodied intelligent large model realizes the closed-loop of autonomous decision-making and self-learning of robots, and opens a wave of end-to-end robot development. The embodied intelligence model endows the robot with stronger task generalization ability and higher task completion ability, promotes the robot to leap from the PLC level to the SLC level, and builds a closed loop of self-learning.
The improvement of interaction ability and cost optimization are the catalysts for the implementation of robots. The embodied intelligence model makes it possible for users to complete task-level programming in natural language, and the unprecedented improvement of interaction ability opens up the application scenario space. The high performance of the embodied intelligent software system may become an important solution to replace the high-performance needs of robot hardware, and the lower cost is expected to make the human robot enter the "hitting area" of large-scale shipments.
In the era of embodied intelligence, the robot cerebellum - AI module and supporting development platform ushered in a major upgrade. Continuously improving the efficiency of receiving information processing, improving the design inclusiveness of embedding **, and developing more diverse kits are the main directions for many manufacturers to improve their products. International leading manufacturers have made new improvements and innovations on the original robot AI chip platform: Tesla released the humanoid robot "Optimus", expanding the application of FSD chips from the field of intelligent driving to the field of intelligent robots; On the basis of the Jetson Orin series platform, NVIDIA has launched a series of intelligent robot platforms such as ISSAC AMR, continuing to consolidate its competitive advantage; Qualcomm's RB6 and RB5 AMR platforms, based on the original RB3 platform, expand the linkage application functions of AI+5G and actively embrace the new technological revolution. Intel combines the second-generation neuromorphic chip Loihi2 with its HERO platform to accelerate the application of brain-like perception; AMD's new pre-purchase hardware acceleration solution on the KRIA K26 SOM platform greatly reduces the platform flexibility and barrier to use. The new robot development platform represented by NVIDIA ISAAC and iFLYTEK Robot Superbrain Platform has achieved one-stop, end-to-end development, or become a new paradigm for robot development in the new era.
Investment suggestions: It is recommended to pay attention to the scarcity of companies that combine large models with robot business, and it is recommended to pay attention to iFLYTEK, Thunderda, EZVIZ Network, Hanwang Technology, etc.;
Risk warning: market competition is intensifying, and the development of humanoid robot technology is less than expected.
Metals: Metal industry: the cornerstone of metal materials to help the development of humanoid machines.
The development of humanoid robots has accelerated, driving the demand for metal materials. Humanoid robots have a human-like appearance, which can replace a variety of traditional robots to a certain extent, but also require more complex drive, sensing, control, and interaction systems. In the early days, humanoid robots were expensive, did not have intelligence, and had limited application scenarios, making it difficult to mass-produce them on a large scale. The rapid development of AI may become the brain of humanoid robots, and the cost of newly released humanoid robots has generally decreased in recent years. Once the humanoid robot overcomes the two shortcomings of high cost and unintelligence, the downstream application space may be opened, and the humanoid robot market space will usher in rapid growth. The rapid expansion of humanoid robot terminals will bring investment opportunities for the entire industry chain, including the upstream material end.
Rare earth magnets: key raw materials for motors, humanoid robots have become a new growth pole. Servo motors account for 22% of the cost of industrial robots and are an important part of robots, while humanoid robots require precise control and high torque, so they put forward high requirements for magnetic performance, which can only be met by rare earth permanent magnet materials. Under neutral expectations, Tesla's humanoid robot will bring 20,000 tons of high-performance NdFeB magnet blanks and 6,747 tons of NdFeB oxide demand in 2030, accounting for 7% of the total demand for NdFeB and NdNd oxide in 20221%/7.2%。In addition to the pull of demand, if the magnetic material company is considered to be in the form of magnetic components, the value of the single machine may far exceed that of new energy vehicles.
Precision cutting wire: harmonic reducer processing consumables, which is expected to benefit from the volume of humanoid robots. The reducer accounts for 32% of the cost of industrial robots and is an important part of the robot, which is mainly divided into harmonic reducer and RV reducer. Among them, the internal teeth of the steel wheel of the harmonic reducer need to be processed by the slow wire walking process, and the Berkenhoff company of Powerway Alloy has performed well in the electrode wire material industry for slow wire walking. It is estimated that in 2030, humanoid robots will bring new demand for precision cutting wire350,000 tons.
Lightweight materials (aluminum, magnesium, carbon fiber): robot skeleton, hundreds of boats compete for the waves. For humanoid robots, the benefits of lightweight are much greater than those of automobiles. Among the three common automotive lightweight materials aluminum, magnesium and carbon fiber composite materials, carbon fiber composite materials are highly expensive, which is not conducive to the overall cost reduction of humanoid robots. Compared with aluminum, magnesium has a smaller density, better shock absorption and heat dissipation properties, and we believe that it is an important alternative to the lightweight of humanoid robots. We expect humanoid robots to bring new demand for lightweight materials by 20301250,000 tons, the elasticity of demand for aluminum, magnesium and carbon fiber is 02%/12.5%/92.6%。
Investment suggestion: based on the potential demand for upstream materials based on the large volume of humanoid robots. We focus on recommending: Jinli Permanent Magnet, Zhenghai Magnetic, Baowu Magnesium, Bowei Alloy. It is recommended to pay attention to: magnetic-related targets of Zhongke Sanhuan, Ningbo Yunsheng, Earth Bear, Yingluohua, Zhongke Magnetics;Nanshan Aluminum, which is the target of aluminum processing, Jinbo Co., Ltd., Chujiang New Material, which is the target of carbon-carbon composite material processing, Xingyuan Zhuo Magnesium, Wanfeng Aowei, etc., which is the target of magnesium processing.
Risk warning: Tesla's humanoid robot volume is less than expected, the industry's key technology breakthroughs are less than expected, and the risk of raw material fluctuations.