(Report Producer/Author: Shen Wanhongyuan Research, Pu Mengjie, Liu Yang, Dai Wenjie)
1. Three major businesses, involving five types of automotive electronic products
At the beginning of its establishment, Hengrun started by selling foreign automotive electronics research and development tools as an agent, entered the field of automotive electronics in 2005, and ushered in a turning point in the development of automotive electronics business in 2008 by undertaking the SAIC-GM project of Shanghai Xiaoshi Lamps. In 2003, Jingwei Hengrun was established, and in 2003-2004, Jingwei Hengrun mainly relied on agents to sell R&D tools in the field of foreign automotive electronics; in 2005, the company decided to focus on the field of automotive electronics. In 2008, the company ushered in a turning point in development, Shanghai Xiao ito Light 1 took SAIC-GM's AFS (Adaptive Turn headlight system) project to find Jingwei Hengrun, initially Jingwei Hengrun was only a chip for Shanghai Xiaoshi Car Lights and Japanese companies to negotiate prices, when Jingwei Hengrun gave a price of 75% compared to competitors in the case of product functions similar to that of international counterparts. In the end, the order that should have belonged to a Japanese company "changed hands" Jingwei Hengrun, and the products were installed in sachine-GM-LaCrosse and other models. Since the first automotive electronics order received by Jingwei Hengrun was a joint venture customer, SAIC-GM's more than 5,000 pages of design and testing standards for automotive electronics and supporting mass production experience have enabled Jingwei Hengrun's ability to rapidly improve. The company began to lay out intelligent network-related product research and development in 2012, and began to do port autonomous driving in 2017 (it has been put into operation in Qingdao Port, Tangshan Port and Rizhao Port).
Matrix Partners' current business involves automotive electronics, automotive IT outsourcing services, and port autonomous driving solutions, among which automotive electronics involves intelligent driving domain, networked products, body domain, chassis domain, and electric system. In 2020, the company's total revenue was 2.473 billion yuan, of which automotive electronics, automotive IT outsourcing, and port autonomous driving accounted for 73%, 26%, and 2%, respectively, and gross profit accounted for 59%, 39%, and 2%, respectively.
2. Intelligent driving: The key point is the in-depth cooperation with Mobileye
Matrix Partners' intelligent driving revenue comes from both commercial vehicle and passenger car markets, and in 2020, intelligent driving revenue from commercial vehicles and passenger cars will be about 50% each, and ADAS total revenue will be 403 million yuan. The main customers of Jingwei Hengrun Passenger Vehicle ADAS system are SAIC Passenger Vehicle, FAW Hongqi, FAW Besturn, Geely Automobile, Dongfeng Liuqi, Aiways Automobile, JIANGHUAI Automobile, Southeast Automobile, etc.; the main customers of commercial vehicles are FAW Jiefang, China National Heavy Duty Truck, Shaanxi Heavy Truck, Jiangling Heavy Duty Truck, etc.
Jingwei Hengrun focuses on the VO route to the ADAS market in front of passenger cars, and the 1V1R route in the commercial vehicle dual warning market. The technical route chosen by the Jingwei Hengrun ADAS system is mainly divided into three categories: VO (pure vision), 1V1R (one camera and 1 millimeter wave radar), and 1V5R (1 camera and 5 millimeter wave radar). Among them, in the passenger car forward ADAS market, Jingwei Hengrun is mainly BASED ON THE VO solution; in the commercial vehicle ADAS market, Jingwei Hengrun is mainly based on the 1V1R scheme.
Since there are great differences in the characteristics of the industry such as passenger cars and commercial vehicles, both in terms of industry driving force and competitive landscape, and the technical routes chosen by Matrix Partners Hengrun in the passenger car and commercial vehicle ADAS markets are also very different, we discuss the future fundamentals of Hengrun in the passenger car and commercial vehicle markets.
2.1 Passenger cars: price points
Matrix Partners' passenger car ADAS business has three main characteristics: (1) the main revenue comes from forward ADAS, while non-low-speed ADAS; (2) almost all the main control chips of the ADAS system are purchased Mobileye; (3) the main visual (VO) route is mainly focused.
The reason why Jingwei Hengrun has formed the above three obvious characteristics is essentially the most critical reason is the very close cooperation between Hengrun and Mobileye. The reason why Hengrun focuses on the pure visual route is because Mobiley is firmly taking the pure visual route (but mobileye released the FMCW lidar SoC in January 2021, the follow-up pure visual route may change); the main reason why Hengrun's main revenue comes from forward ADAS is because Mobileye's EyeQ3 and EyeQ4 forward visual perception algorithms are leading the world, and the share of the forward ADAS system master chip is very high, but Mobileye is in Parking is not available on The EyeQ3 and EyeQ4.
According to data from Zosz Auto Research Institute, the main suppliers of front-view systems for passenger cars in China are Denso, Bosch, Aptiv, Kosta, Panasonic, Weininger, Continental, Jingwei Hengrun, etc., of which the top 10 in 2020 occupies more than 90% of the market share, while only Jingwei Hengrun is shortlisted in China, with a market share of 3.6%. Hengrun's forward ADAS system market share among its own brands is 16.7%, ranking second only to Bosch. In addition to Matrix Partners Hengrun, Aptiv, ZF, Kosta, Wistron Capital, Ehang Yuanzhi, Zhixing Technology, etc. have also established long-term cooperative relations with Mobileye. Among Mobileye's customers, except for Nissan, which is an automaker, the rest are all Tier1, and its Tier1 customers are mainly ZF TRW, Aptiv, Magna, Valeo, Densan, Hengrun, and Keihin.
The forward ADAS market currently presents two camps: the Mobileye camp + forward perception algorithm Self-developed or the cooperation camp with third-party algorithm companies - of which the Mobileye camp occupies 36% of the Chinese passenger car forward ADAS market in 2021. The former is represented by ZF, Aptiv and Matrix Partners, and uses Mobileye's EyeQ series chips as MPU in almost all forward ADAS camera all-in-one machines; the latter is represented by Bosch, Continental, Desay SV, and Intelligent Driving Technology. According to the statistics of Zosie Automobile Research Institute, In 2020, Mobileye has occupied more than 30% of the market share of China's passenger car forward vision chip market; according to the data of Gaogong Intelligent Vehicle, Mobileye will occupy 36.3% of China's forward ADAS visual perception market in 2021.
Mobileye "chip + visual perception algorithm" sold together, this way greatly reduces the threshold of the entire forward ADAS market, we speculate that the value of the visual perception algorithm in its single EyeQ chip is about 230 yuan; the non-Mobileye camp Tier1 needs to develop its own visual perception algorithm, and the front-view camera MPUs they choose mainly include Xilinx FPGA, Renesas V3H, NXP S32V and other chips. Bosch's second-generation single currently uses a Xilinx FPGA (model XA7Z020) as an MPU in the MPC2 of the ADAS camera, and the price of a single FPGA is about 120 yuan, while the average price of a single Mobileye EyeQ series chip is about 350 yuan, and it can be inferred that the value of the visual perception algorithm included in the EyeQ series chips is about 230 yuan. The proportion of vision in the entire intelligent driving technology will still be very large, basically undertaking about 70% of the working condition analysis work, becoming the mainstream perception scheme in mass production.
But as automotive electrical and electronic architectures move from distributed to domain-centralized, opportunities for the non-Mobileye camp are likely to increase. The entire ADAS system will evolve along the "FCM All-in-One (+FCR All-in-One) – Forward ADAS Controller + Low-Speed ADAS Controller – All-in-One Controller", which will have a profound impact on the rivalry between the Mobileye camp and the non-Mobileye camp. We will elaborate on this in the following pages.
Passenger car L0-L2 class ADAS controllers are distributed architecture: forward ADAS system, side and rear ADAS system, automatic parking system, panoramic surround view system corresponding to 1 controller equipped with corresponding algorithms to achieve; therefore, the above subsystems usually use the "all-in-one" scheme (that is, intelligent sensors, sensor modules and calculation unit ECUs, algorithms are integrated). For example, in addition to the camera module in the front-view camera IFC also contains the ECU motherboard and algorithm (this form of front-view camera is called FCM), and the ECU motherboard contains 2 chips, namely "safety core" and "performance core" - "safety core" generally uses MCU, responsible for ensuring functional safety, the calculation requirements are not high; "performance core" generally uses MPU, undertakes a lot of computing, and allows low functional safety level. Millimeter-wave radar also contains 1 MIIC chip and 1 MCU as well as millimeter-wave radar perception algorithms, which are all-in-one machine form (known as FCR). For example, the Hirain FAS-Cam front-view camera solution released by Matrix Partners in May 2020 is a typical FCM all-in-one, FAS-Cam integrated Mobileye EyeQ4, equipped with Infineon AURIX TriCore TC234L chip. The EyeQ4 corresponds to the above-mentioned MPU (Performance Core) and the Infineon TC234L corresponds to the above MCU (Safety Core).
The L2+ stage of the passenger car will gradually become a domain controller architecture in two phases – the first stage: low-speed ADAS controller + forward ADAS controller; and the second stage: the parking integrated domain controller (low and medium computing power). In the first phase, the self-parking APA and the surround view system AVM are fused into a single ECU, with 4 cameras and 12 ultrasonic radars collectively referred to as low-speed ADAS systems; in terms of driving functions, the forward ADAS system and the side-back ADAS system are fused into one ECU (forward-looking MPU + converged MPU) or still in the form of an intelligent all-in-one (front-view MPU). In the second phase, all driving and parking functions are fused into a single domain controller for calculations, i.e. "parking in one".
L2+ domain controller architecture can be divided into: 1V5R: 1 domain controller + 1 FCR + 2 SRR + 2 SFR + 1 FCM, only integrate the driving function, each sensor is still a smart sensor, camera and millimeter wave radar directly provide structured data to the domain controller, the domain controller inside the main MCU (such as TC397), responsible for the structured data fusion algorithm calculation can not be responsible for the calculation of the perception algorithm. 5V5R (row berth integration): 1 domain controller + 1 pure hard front-view camera + 4 surround view cameras + 5R, at this time the front-view camera is no longer an "all-in-one" scheme but just a pure hardware (without algorithms and ECU motherboard cameras), in addition to the domain controller in addition to the MCU will also add an MPU, to be responsible for forward vision and surround vision perception.
The above 1V5R and 5V5R schemes have in common in that millimeter-wave radar is still an "all-in-one" form (intelligent sensor), but the most critical difference is that the computing tasks of the forward-looking visual perception algorithm have been transferred from the camera to the domain controller. The dominant configuration of L0-L2 level forward ADAS is now the 1V1R scheme (i.e., FCR+FCM scheme), pure radar (single FCR scheme) and pure vision (single FCM scheme), although quantitative but not mainstream; the convergence of forward ADAS and lateral rear ADAS is often collectively referred to as the "driving ADAS system", that is, the 1V3R scheme.
In the future, 5V5R berth will become the mainstream configuration of ADAS, and its overall cost will be reduced by 20-30% compared to the "1V1R+APA" split solution. Neusoft Ruichi pointed out: "Thanks to the sensors of the driving ADAS and the reuse of domain controllers, the new generation of mobile and berth integrated domain controllers save 20%-30% compared with the traditional 1V1R+APA technology solution." Nujie Technology also pointed out in the public account: "Taking the TDA4VM with 8TOPS hashrate as an example, if you use 6 cameras, 5 millimeter wave radars, 12 ultrasonic radars, integrated high-precision maps, and the solution of integrating L2+ HWP and AVP into the vehicle, it will save a lot of cost and be more competitive in the market than the use of two separate systems to achieve two functions." ”
The biggest advantage of the pure vision (VO) forward ADAS solution is the low price, the biggest disadvantage is the inability to accurately measure the distance, and the second is that objects cannot be seen at night. (1) Advantages: low price. The biggest advantage of the pure vision solution is the low price cost, according to the economic report of Jingwei Hengrun, its ADAS products (at least half of which are pure visual forward ADAS systems) will average about 1311 yuan / set in 2020, while the price of 1V1R will rise to 250 US dollars / set due to the addition of a millimeter wave radar (about 1724 yuan / set at the 2020 exchange rate). (2) Disadvantages: Difficulty in ranging. Pure visual forward ADAS generally uses monocular vision, but the deep learning algorithms used in traditional monocular vision are mostly based on 2D image recognition, and the depth information of the object cannot be obtained lack of stereoscopic space modeling capabilities, so the perception distance error is large, so the use of "camera + millimeter wave" radar fusion scheme can achieve better results, because millimeter wave radar is good at speed measurement and ranging can make up for the shortcomings of camera ranging difficulties.
However, Tesla and Mobileye chose a pure visual algorithm route threshold is very high, requiring "massive data + powerful deep learning algorithm capabilities + large-scale AI supercomputing center investment", we estimate that Tesla only spent $142 million on AI servers; Mobileye adopted REM crowdsourcing methods and stock models equipped with EyeQ chips to collect massive data. Tesla calls its deep learning neural network HydraNet, the entire HydraNet contains 48 different neural networks, through these 48 neural networks, you can output 1000 different prediction tensors, theoretically, Tesla's super network, can detect 1000 kinds of objects at the same time. Accomplishing these operations is not simple, Tesla has spent 70,000 GPU hours training deep learning models. Tesla's supercomputing center used to train deep neural networks ranked fifth in the world in computing power, with a total of 3 AI data centers, the first computing center for auto-tagging had 1752 GPUs, the second and third computing centers for training had 4032 GPUs, and the other had 5760 GPUs, assuming that the server price of 2 NVIDIA V100 and 2 A100 GPUs was $23,700 and $25,600, respectively (data source: IDC), This gives an estimate of up to $142 million for the purchase of AI servers alone.
The 5V5R parking integrated domain controller has a variety of chip solutions: single TDA4, dual TDA4, 1 TDA4+1 horizon J3, dual horizon J3, 1 TDA4+1 Mobileye EyeQ4, and so on. EyeQ3 and EyeQ4 can only be used as a driving system SoC, not as a parking system SoC, so if you want to use the EyeQ4 chip for the parking integrated ADAS domain controller, you can only choose the "1 EyeQ4 + 1 TDA4" chip solution, of which EyeQ4 is used as the driving SoC and TDA4 is used as the parking SoC. Starting with EyeQ4, each generation of EyeQ chips will have at least two versions: the lower version is used to support the L2 level single camera solution, and the higher version is used to support the L2+ level multi-camera solution. The high version of the EyeQ4 can support up to three cameras (for example: the three current sight cameras on the BMW Autopilot Assist Pro are supported by EyeQ4, the conventional camera + fisheye camera for driving recorder + narrow-range camera), the high version of The EyeQ5/EyeQ6 can support a greatly improved number of cameras, which can achieve 360° viewing angle coverage around the vehicle. (Source: Future Think Tank)
Therefore, for the forward ADAS vendors of the original Mobileye camp, the trend of integrated parking requires them to increase their capabilities in three aspects:
(1) Parking algorithm capability (for the non-Mobileye camp Tier1 requires it to have strong parking algorithm ability and driving algorithm capability): Since the computing power of the ADAS parking integrated domain controller is mostly between 8-16TOPS, how to efficiently run the upper-level driving algorithm and parking algorithm under limited computing power resources requires software and hardware manufacturers or software and hardware departments to do joint optimization on the upper-layer application algorithm, middleware and low-level software, and the software and hardware are high on the ADAS system Degree-coupled features. Therefore, system manufacturers also need to have strong algorithm capabilities - among them, the perception algorithm in the driving system Mobileye's EyeQ4 chip has been provided, and the system-level Tier1 only needs to make its own decision-making, planning, and control algorithms; but the parking and surround view algorithms in the parking system Mobileye's EyeQ4 are not provided, and need to be solved by the system manufacturers themselves.
(2) TDA4 and other parking system SoC development and landing capabilities: Chasing Technology pointed out: "For those chips that can be mass-produced in the next three years, such as TDA4, Qualcomm, and Horizon J3, there are only a dozen TOPS hash rates." Although such chips have been listed for some time, the number of companies that can really make hardware and software in China is only single digits. And even if the underlying driver of the hardware is opened, even less can completely port the automatic driving function on it. Traditional Tire1 only has manufacturing capabilities; companies that only do algorithms do not have the ability to develop and port embedded systems. "Desay SV officially launched R&D work based on TI TDA4 chip in July 2019, and in November 2021, it officially released the TDA4-based mass production product IPU02, which is installed on the Geely Xingyue L, which is the world's first TDA4 mass production project in an advanced driver assistance landing. From the chase and the development process of Xiwei, it can be explained that it is very difficult to run through a chip platform at the same time from the software and hardware, and it takes at least 2 years to start system-level development to mass production based on a specific chip platform.
(3) Millimeter wave radar capability: In the 5V5R lane parking integrated system, although the mmwave radar is still the form of intelligent all-in-one machine FCR, but the millimeter wave radar capability is also one of the key competitivenesses for related manufacturers to increase the share of the ADAS market: Analogy with the current distributed driving ADAS system competition pattern, the 1V1R market has long been monopolized by overseas Tier1 giants, the key reason is that the performance, stability and other indicators of domestic mmWave radar are not as good as overseas Tier1, The overall performance of China's local Tier1 "FCM+FCR (domestic)" system is not as good as that of overseas Tier1 giant "FCM+FCR (self-developed)" - the same reason is also applicable in the integrated parking market.
We estimate that in 2021, the market size of China's passenger car forward ADAS will be about 13.2 billion yuan, with a penetration rate of 40%, of which the 1V1R and VO market sizes will be 9.8 billion and 600 million yuan, respectively, and the low-speed ADAS market size will be 4.9 billion yuan, with a penetration rate of 12%. According to the data of Gaogong Intelligent Vehicle: in 2021, the front-loading insurance volume of passenger cars forward ADAS was 8.0789 million units, +29.51% year-on-year, and the corresponding penetration rate was 40%; of which the L2 level insurance volume was 3.9562 million units, +77.65% year-on-year. Among all forward ADAS solutions in 2021, 1V1R, VO, and pure radar solutions account for 71%, 6%, and 14% respectively – based on this, we estimate that the corresponding market sizes of 1V1R and VO solutions are 9.8 billion and 600 million, respectively, and the total market size of all forward ADAS is 13.2 billion yuan. According to our statistics in the "Zhilian Automobile Series Depth 19: The Differentiation of New and Old Forces Is Clear, and the Wave of Domestic Substitution has Arrived (2021 Automotive Technology Penetration and Market Share Tracking Report)": The penetration rate of China's passenger car automatic parking market in 2021 is 12.17%, assuming that the unit price of around the view + parking is 2000 yuan, it can be estimated that the low-speed ADAS market size in 2021 will be 4.9 billion yuan.
We expect that in 2025, China's forward ADAS market size will be about 58 billion yuan, with a penetration rate of 80%, of which the integrated parking, forward ADAS (non-berthing integrated) and low-speed ADAS (non-berthing integrated) will be 43.7 billion, 10.5 billion and 3.7 billion, respectively, and the corresponding penetration rates will be 45%, 35% and 10% respectively. We believe that because the "5V5R" scheme can be reduced by 20-30% compared to the current mainstream "1V1R+ APA" scheme, the subsequent 5V5R parking integrated solution will replace the current mainstream 1V1R and VO schemes, and the domestic local Tier1 parking integrated scheme will begin to be mass-produced in 2022, and we expect that the penetration rate of the parking integration will reach 5% in 2022 and 45% in 2025, corresponding to 1V1R and VO The share will decline - we estimate that the overall market size of China's passenger car ADAS will be about 58 billion in 2025, of which the integrated parking, forward ADAS (non-berthing integrated) and low-speed ADAS (non-berthing integrated) are 43.7 billion, 10.5 billion and 3.7 billion, respectively, and the corresponding penetration rates are 45%, 35% and 10%, respectively; the market size of 1V1R and VO solutions in forward ADAS (non-mobile parking integration) is 7.8 billion and 500 million, respectively.
The market segment corresponding to the ADAS business of Matrix Partners was previously mainly the above-mentioned VO solution, followed by the 1V1R solution. The ADAS controllers that Matrix Partners have launched are mainly divided into three categories: intelligent driving domain controller ADCU, on-board high-performance computing platform HPC, and automatic parking controller; Hengrun has not yet launched a medium and low computing power L2 row parking integrated domain controller, but it is currently developing L3 level intelligent driving domain controllers. Both ADCU and HPC are used in the ADAS system for driving. According to the research projects disclosed in the prospectus, Hengrun is developing a large-power intelligent driving domain controller for L3 level, with an estimated R&D cost of 245 million yuan, and the R&D period is planned for 2018-2023, and is currently under research and development (the molecular project has been completed). In February 2022, Matrix Partners and Black Sesame signed a strategic cooperation agreement, and it is expected that Hengrun will adopt huashan series chips in L3-level intelligent driving domain controllers; in addition, since most of the automakers' intelligent driving computing chips are using NVIDIA Orin, we expect Hengrun to also develop intelligent driving domain controllers based on Orin.
(1) The MPU of the intelligent driving domain controller ADCU is based on EyeQ4, and the MCU uses Infineon, which has been mass-produced on the Red Flag E-HS9. (2) The on-board high-performance computing platform HPC selects two high-performance chips OF TI TDA4 and Infineon TC397, Jingwei Hengrun has started self-developed HPC product research and development since 2019, and has launched two generations of products so far, and has been fixed-point to win through science and technology, Baoneng Automobile (but has not yet been mass-produced). (3) There are two chip solutions for the automatic parking system controller: TC297+TDA2 for APA, and TC397+TDA4 for AVP. Supporting customers include SAIC Maxus and Baoneng Automobile.
Summary: Mobileye's share of China's forward ADAS market will increase from 30% + to about 50% in 2-3 years, and we expect that Hengrun's low-cost VO solution equipped with Mobileye chips and the advantages of Mobileye visual perception algorithm in the passenger car ADAS market can still be in the low-cost and forward-facing ADAS vehicles with forward-to-high ADAS demand, and with the increased penetration rate of this type of forward ADAS in low-priced cars of its own brands, we can still obtain revenue growth. In 2021, the penetration rate of ADAS in the forward of self-brand 100,000-150,000 price models is about 40%, and the penetration rate of models below 100,000 is about 10%, and these two price models are the revenue increments of the follow-up Hengrun (the penetration rate of 60-70% of the 150,000-300,000 price points is already at a high level). However, the L2-level row-berth integration scheme and the L3-level large-power domain controller plan can be observed for hengrun's subsequent progress and product release.
2.2 Commercial vehicles: the second share of double warning, waiting for the dividend of the medium and light truck policy
Commercial vehicle ADAS is a strong policy-driven market. In 2017, the Ministry of Transport of the Mainland stipulated in the "Technical Requirements for the Safety of Operating Passenger Cars" and the "Technical Conditions for the Safety of Operating Trucks" that: (1) Buses over 9 meters should be compulsorily installed with the "Lane Departure Alarm Function" and the "Vehicle Forward Collision Warning Function" (dual early warning system) before april 2018, and the automatic emergency braking system (AEB) should be installed from April 2019. (2) Trucks: Heavy trucks and tractors with "more than 18 tons" and "maximum speed greater than 90km/h" should be pre-installed with mandatory dual warning systems from September 2020; heavy trucks with "more than 12 tons" and "maximum speed greater than 90km/h" and tractors with "maximum speed greater than 90km/h" should be mandatorily installed with AEB from May 2021.
The commercial vehicle front-loading dual early warning system market size in 2021 is 1 billion yuan, and the AEB market size is 151 million yuan. In 2021, the sales volume of new buses over 9 meters was 32,700 units, the sales of heavy trucks and tractors with a maximum speed of more than 12 tons and a maximum speed of more than 90 km/h were 340,000 units (AEB must be installed), and the sales volume of new vehicles with a maximum speed of more than 18 tons and a maximum speed of more than 90 km/h and tractors with double warning (LDW+FCW) was 659,500 units, accounting for 47.39% of the new vehicles of medium and heavy-duty operating trucks. The unit price of the dual early warning system is about 1500 yuan / set, and the unit price of AEBS is about 2000 yuan / set, which can be calculated that the market size of the passenger car double warning system in 2021 is 0.49 billion yuan, and the market size of the AEB market is 0.65 billion yuan; the market size of the heavy truck and tractor double warning system is 989 million yuan, and the market size of AEB is 0.86 billion yuan.
However, due to certain strategies adopted by commercial vehicle manufacturers to circumvent regulatory requirements, the actual implementation of regulations has been less than expected. Because of the higher price of vehicles and higher maintenance costs due to the double warning and AEB, which requires an additional fee of a few thousand yuan - most heavy truck and tractor manufacturers choose speed limit strategies to circumvent regulatory requirements, and the number of insured vehicles with standard AEB before operating trucks in May-December 2021 is 42,900 units, accounting for 14.16% of all sales of operating trucks that meet the regulatory requirements to be equipped with AEB, and the remaining 86% (about 260,000 units) of new cars should be installed but finally not installed according to regulatory requirements In the truck double early warning system market, the actual installation volume in 2021 is 659,500 units, there are 40,000 new cars that must be installed but not installed in line with the requirements of the regulations, and some car companies avoid the regulatory requirements by restricting the total mass of the body, speed limit devices, etc.; while the passenger car factory takes the way of controlling the length of the body to avoid the regulatory requirements, and in 2021, more than 5,000 new production buses are near the critical point required by the regulations, and the length is controlled within 9 meters (between 8.9 meters and 9 meters).
Matrix Partners ranks second in the 20% market share of truck dual warning systems, second only to Neusoft Ruichi, and fourth in the 9% market share of truck AEB. (1) Trucks: The market share of the dual early warning system is the first and second respectively Neusoft Ruichi and Jingwei Hengrun. AEB market share of the first and second are Fortek and WABCO (acquired by ZF for $7 billion in 2020), and Knorr-Bremse's main AEB solution, ranked 6th, is based on camera and radar sensor data fusion processing technology, with high accuracy; because AEB and dual early warning systems, AEB also involves chassis control (including interface opening authority), which is still mainly in the hands of foreign suppliers such as ZF (WABCO) and Knorr-Bremse (these two). All are the head manufacturers of chassis control systems for commercial vehicles), so foreign suppliers have a significant advantage in AEB's market share over dual early warning systems. (2) Bus: Jingyida is the Tier1 of Yutong Bus, and Yutong ranks first in the market share of more than 30% of the bus market, so that Jingyida ranks first in the bus dual early warning system and AEB market; Zhejiang Rongzhong's perception solution supplier is Jimu Intelligence.
In the truck visual perception program, ZF, Knorr-Bremse, Jingwei Hengrun use Mobileye, Neusoft Ruichi, Fortek self-developed or purchased third-party algorithms; in the bus visual perception program, intelligent driving technology as Tier2 occupies more than 40% of the share. (1) Trucks: In terms of visual perception schemes, Frettec, Qingzhi Technology, and Neusoft Ruichi all use independent research and development perception schemes or domestic sensing schemes (for example, smart driving technology MAXIEYE), while ZF, Knorr-Bremse, and Jingwei Hengrun all use Mobileye solutions. (2) Bus: In 2020, Bus ADAS Visual Solution Providers Smart Drive Technology ranks first with a market share of 42.62%, Tier2 model. In terms of bus (LDWS, AEBS) forward vision perception algorithm suppliers, intelligent driving technology, polar intelligence, magic vision intelligence, starship, and Forretac ranked in the top five, these manufacturers are domestic solution suppliers.
In addition to the above-mentioned commercial vehicle regulations ADAS market, Hengrun has also laid out the "port unmanned" track since 2015, and has now landed in Qingdao Port, Tangshan Port and Rizhao Port, and has launched more than 20 intelligent port vehicles. In terms of the overall solution of high-level intelligent driving, the company has cooperated with FAW Jiefang and Suzhou Zhitu since 2017 to carry out port smart card collection projects in Qingdao Port, Tangshan Port and Rizhao Port. Truck drivers account for more than 50% of the entire port transportation costs, the current average cost of truck drivers is about 150,000 to 200,000 yuan / year, and the annual labor cost of mainland ports for truck drivers is about 50 to 100 billion yuan. At present, the cost of unmanned collection of cards (vehicle + transformation + deployment) is about 1.2-1.8 million yuan / unit, compared with the automatic guided vehicle (AGV) of magnetic nail navigation technology, each price of 500-800 million has obvious advantages. The domestic companies that do port unmanned driving mainly include: Xijing Technology, Mainline Technology, Sinian Intelligent Driving, Changjia Fengxing, Feibu Technology, Yuanrong Qixing, Tucson Future, Jingwei Hengrun, etc., and traditional engineering companies such as Sany Offshore and Zhenhua Heavy Industry are also in the layout.
3. Two major highlights: gateway upgrade + integration of gateway and body domain controller
There are two main products of Jingwei Hengrun Network: gateway and T-BOX. The ECU in the same domain is connected by cable and can communicate freely, but if you want to communicate between the ECUs of different domains, you need to let the "gateway" assist, because the communication protocol between different domains is different (CAN/LIN/FlexRey/Ethernet, etc.), the network is equivalent to a "translator" role, and the main function of the gateway is to forward data securely and accurately between different domains. In-vehicle gateways are mainly used for communication between different domains in the car, while T-Box is mainly used for communication between the car and the outside world.
In 2021, the penetration rate of automotive independent gateways has reached 99.5%, and almost every car is standard; in 2021, the penetration rate of automotive TBOX has reached 65.19%, and the industry has entered a mature stage. Traditional vehicle gateways are divided into two types: independent gateways and integrated gateways, of which independent gateways are the mainstream solutions that account for the vast majority of existing models. According to the statistics of Gaogong Intelligent Automobile, in 2020, the number of independent gateways equipped with new passenger car insurance in China is 17.6776 million units, with a penetration rate of 92%; in 2021, the number of independent gateways is 20.0486 million, the penetration rate is 99.5%, and the amount of insurance is +13% year-on-year. According to the statistics on the penetration rate of the "Internet of Vehicles" in our recently released "Zhilian Automobile Series Depth 19 - The Differentiation of New and Old Forces is Obvious, and the Wave of Domestic Substitution has Arrived (2021 Automotive Technology Penetration Rate and Market Share Tracking Report)" (which reflects the change in T-BOX penetration rate), 2018-2020 is the T-BOX high-speed penetration stage, but from 2021 onwards, the T-BOX penetration rate has slowed down, and the penetration rate has reached a relatively high level of 65%. Although it is expected that the follow-up can still rise to 100%, it is expected that the follow-up industry growth rate will maintain a steady growth level, not as good as 2018-2020.
Independent gateway competition pattern: Continental group first, Hengrun eighth. According to data from Gaogong Intelligent Vehicles, continental, Bosch, and OFILM (ranked among the top three for the first time) account for 50% of the market share in terms of gateway (including traditional CAN, CAN FD, Ethernet, etc.) suppliers. Among them, OFILM (Huadong Automobile and Electric, Nanjing Tianqing), Jingwei Hengrun and BYD ranked among the top three domestic suppliers. In addition, in terms of the overall market concentration of independent gateways, the share of TOP10 suppliers is nearly 90%.
Traditional body controllers (BCMs) are being upgraded to body domain controllers, which currently have a penetration rate of 4.31%. According to the statistics of Gaogong Intelligent Vehicles, in 2021, the number of new passenger cars equipped with BCM integrated domain controllers (products in the transition stage) in the Chinese market (excluding imports and exports) was 842,300 units (according to the statistical caliber of bicycles, some models are equipped with multiple domain controllers), and the penetration rate was 4.13%. For example, one of the highlights of BYD's e-platform 3.0 is the left/right body domain controller, which integrates BCM, security gateway, key center, air conditioning control, tire pressure monitoring, instrument control, parking assistance, smart key and other modules, and the extended version supports up to 32 traditional distributed ECU functions.
Body controller BCM competitive landscape: Continental group occupies the first place with a 31% share, and the top 7 are all foreign Tier1 or Sino-foreign joint ventures with international Tier1 joint venture background. In terms of total BCM controller market share, Continental, United Automotive Electronics and Itecker ranked in the top three, with a total share of 88.54% of the top ten suppliers. At the same time, Chinese suppliers such as Matrix Partners Hengrun, Neusoft, Noble Technology, and Desay SV are almost at the same starting line as Continental Group and UMC, and have the opportunity to break the monopoly pattern of foreign suppliers in the traditional BCM and gateway markets.
There will be three subsequent changes in the automotive gateway industry: (1) Ethernet will replace CAN as the backbone network in the car, and the central gateway will also be upgraded from CAN to Ethernet. (2) The central gateway is integrated with the body domain controller and merges into one to become the central server. (3) The automobile changes from the ECU distributed architecture to the multi-domain controller architecture, and the central gateway becomes a "service gateway"; in the longer term, when multiple domain controllers are fused into a central computing platform, the "service gateway" will further evolve into a "regional gateway".
Trend 1: Traditional CAN gateways are upgraded to 100 Gigabit Ethernet Gateway, and in 2021, 100 Gigabit Ethernet Gateway will account for 13% of the overall independent gateways, and it is expected that this proportion will continue to rise until it is completely replaced. According to the statistics of Gaogong Intelligent Vehicles, in 2021, the number of new passenger car vehicles in the Chinese market (excluding imports and exports) with gateways above 100 megabytes is 2.5149 million, with a penetration rate of 12.33%, and Ethernet gateways above 100 megabytes account for 13% of the overall independent gateways. The unit price of the traditional CAN gateway is about 250 yuan, the unit price of the 100 Megabit Ethernet gateway is 500 yuan, and the price of the service gateway is about 1000 yuan, according to the assumption of the penetration rate of each type of gateway in 2022-2024 on the second page of this report, it can be calculated that the overall market size of China's passenger car independent gateways in 2022-2024 is expected to be 65, 73 and 8 billion yuan, respectively.
Trend 2: As automotive E&E architectures change, central gateways (distributed ECU architectures) will become service gateways (multiple domain controller architectures) and eventually regional gateways (central computing architectures), from 1 to 4. There is only 1 central gateway on the car in the central gateway and service gateway stage, but when the automotive electrical and electronic architecture becomes only 1 central computing unit, there will be 4 regional gateways distributed in the four corners of the body, and the number of gateways will increase. There are three implementation forms of service gateway: the gateway as the gateway for service scheduling, the gateway computing platform (only the visual part of the intelligent driving on the whole vehicle is the largest demand for computing power, but the other vehicle logic is I/O processing, the computing power demand is not large, so the vehicle logic other than vision can be put on the service gateway for processing, this case is called the gateway computing platform), and the whole vehicle logic and vision are all put on a strong heterogeneous chip for processing.
Trend 3: The integration of central gateways and body domain controllers has gradually become a trend, representing products such as Continental HPC and OFILM fifth-generation body domain controllers, which have been designated by the car factory and some have been mass-produced. Continental's HPC integrates a central gateway into the body domain controller, and in 2020, Continental's first-generation body HPC will be mass-produced on Volkswagen's ID.3 based on the MEB platform as an in-vehicle application server (ICAS1, based on renesas R-Car M3), mass-produced on ID.4 in 2021, and 500,000 Volkswagen vehicles equipped with HPC by March 2022; Continental's second-generation HPC was designated by the GAC Group in December 2021 The first order in China, based on the NXP S32G399, is ready to be put into production at the first time on the manufacturer's electric vehicle platform in 2023 (the "Xingling" electronic and electrical architecture released by GAC), and is planned to be installed in the next five models including electric vehicles, hybrid vehicles and fuel vehicles. The fifth-generation body domain controller product released by OFILM in December 2020 combines the body domain controller and the Ethernet gateway into one, and has also been designated by domestic car manufacturers. The Central Electronic Control Module (CEM), developed by Great Wall's Noble Technology, uses Renesas' new generation of MCU products and switches provided by NXP, and is the first highly integrated electronic control unit in the industry to integrate functions such as body domain controllers, BCM, PEPS, and Gigabit Ethernet gateways, and will enter mass production in the second half of this year.
The body domain controller with integrated gateway will have about 500,000 vehicles insured in 2021, and we expect the domestic passenger car body domain control market size to be about 8.9 billion yuan in 2024. According to the monitoring data of The Gaogong Intelligent Automobile Research Institute, in 2021, the number of passenger cars in the Chinese market will exceed 500,000 vehicles equipped with body domain control (body control + gateway integration) (considering that some models are equipped with two controllers on the left and right of the body, the actual number of domain controllers is close to 1 million units). The unit price of the body domain controller with the integrated central gateway function is about 1000 yuan, while the traditional single body controller (BCM) is about 100-200 yuan; the 100 Megabit Ethernet off is about 500 yuan, and we expect the body domain controller with the integrated gateway to cost about 1500 yuan per unit. Based on the assumptions on the penetration rate of body domain controllers (without gateways) and body domain control (integrated gateways) on page 2 of this report, we expect the body domain controllers market size to be approximately $22,43,89 billion in 2022-2024. Among them, the original price of the high-performance service-oriented network gateway commonly used chip NXP S32G series chip is about 30-40 US dollars (2 to 3 times higher than that of traditional gateway chips). (Source: Future Think Tank)
4. Valuation: PS valuation is a reasonable market value of 16.5 billion this year
Based on the above section 2-3 analysis, our core assumptions for the revenue of Hengrun's three key automotive electronic product lines are as follows: Intelligent driving:
(1) Passenger cars: from 2022, the 5V5R berth integration plan will gradually replace the pure visual VO scheme and the current mainstream 1V1R scheme, so we predict that the VO and 1V1R market size will remain stable (although the share will be replaced by the integrated mooring, but due to the overall penetration rate of the forward ADAS, the market size remains stable), because Hengrun is the main VO scheme in the passenger car market, the 5V5R berth integrated scheme has not been launched, according to the fixed point of the company's prospectus, In 2020, Hengrun obtained geely automobile GEE 1.0 &AMA+, GEEA2.0, CMA1.0 a total of 3 vehicle platform project VO solution fixed points, so we expect Hengrun's market share of the VO market in the next 2 years will continue to increase. (2) Commercial vehicles: Since the dual early warning system for heavy-duty trucks and the AEB system were fully installed in September 2020 and May 2021 respectively, we expect Hengrun commercial vehicle revenue to remain stable from 2022 onwards.
Intelligent network connection: (1) gateway: Since the subsequent upgrade will be from the "traditional CAN gateway - 100 Megabit Ethernet Gateway - service gateway", so the gateway will appear unit price increase logic, we expect the average price of Hengrun independent gateway will increase from 289 yuan in 2021 to 450 yuan in 2024; on the other hand, because the body domain controller will integrate the gateway, so the revenue of the integrated gateway will be reflected in the "body control" revenue, and the penetration rate of the independent gateway will be 99.5% in 2021, which we expect will be This will drop to 80% in 2024. (2) T-BOX: Since the penetration rate of the T-BOX industry has reached a high of 65.19% at the end of 2021, the competitive landscape is also relatively stable, so we expect the revenue from T-BOX from subsequent companies to remain stable.
Body Control: We expect that the follow-up traditional body ECU will be replaced by body domain controllers - of which the body domain control penetration rate of the integrated gateway will be 2.48% in 2021, and we expect it to increase to 20% in 2024 at a unit price of 1500 yuan; the penetration rate of the body domain controller with the non-integrated gateway will be 1.65% in 2021, and we expect to increase to 10% in 2024, with a unit price of 1000 yuan. Hengrun body domain controller has been designated FAW, Huaren Express, we expect the company's body domain controller market share will reach 3% in 2022, in 2024 will reach 7%.
We expect that with the growth of intelligent driving business shipments to bring scale effects and body domain controllers gradually mass production, the company's automotive electronics business gross profit margin will be stable and rising, of which: (1) body control: Because the body domain controller is more complex than the ordinary body ECU development complexity and higher integration, so we speculate that the gross profit margin is also higher than the ordinary body ECU, so with the gradual increase in the proportion of body domain controllers, Hengrun body control business line gross margin will show an upward trend. (2) Smart Networking: The gross profit margin of 100 Megabit Ethernet Gateway and service gateway is slightly higher than that of can gateways that have matured and competed fiercely, so with the penetration of Ethernet Gateway and service gateway, we expect that the gross profit margin of intelligent network connection business will increase. (3) Intelligent driving: The company's VO products are released in Geely Automobile, and the gross profit margin increases with the scale.
(This article is for informational purposes only and does not represent any of our investment advice.) For usage information, see the original report. ）
Featured report source: [Future Think Tank].
![This "core" makes smart cars more "smart"! Black sesame wisdom and latitude constant run[fig]](data:image/gif;base64,R0lGODlhAQABAIAAAP///wAAACwAAAAAAQABAAACAkQBADs=)