With the help of deep accumulation in the CPU field, the company entered the research and development of DCU products in October 2018, forming a high-end processor product layout mainly based on the dual series of CPU and DCU.

2025/06/2409:20:51 hotcomm 1454

(report producer/author: Northeast Securities, Huang Jing, Li Jiu)

1. Haiguang Information: The leader of the independent advanced microprocessor industry

1.1. Haiguang Information: The leading domestic high-end processor company

Haiguang Information, the domestic high-end processor Light of hope. The company was founded in 2014 and is mainly engaged in the research and development of computing chip products and systems such as high-end processors. After years of hard work, the company has designed and mass-produced a number of high-end processor products that are compatible with the x86 ecosystem and have a performance of the same mainstream internationally, which can provide strong computing capabilities for cloud computing , big data analysis, distributed cloud storage and other scenarios. The company is committed to becoming China's most important computer chip design company and providing a core computing engine for the prosperity of China's information industry. On June 21, 2022, the company's listing application was approved by the China Securities Regulatory Commission and was subscribed on August 3, 2022. The issuance price is 36.00 yuan per share.

company combines the advantages of Chinese Academy of Sciences and local state-owned assets, and is committed to becoming a neutral chip company. According to the company's prospectus, the company has no controlling shareholder or actual controller. Since 2018, the company's top two shareholders are Zhongke Shuguang and Chengdu State-owned Assets (including Chengdu Industrial Investment Co., Ltd., Chengdu Gaotou Co., Ltd. and Chengdu Jicui Co., Ltd., which are joint actors). As of July 26, 2022, the shareholding ratios of the two were 32.10% and 19.53% respectively. The company has a total of 4 participating holding companies, namely Haiguang Integration, Haiguang Microelectronics, Haiguang Hangzhou and Zhixiangerwei, with direct shareholding ratios of 70%, 49%, 100% and 1.14% respectively.

's main products include Haiguang General Processor (CPU) and Haiguang Coprocessor (DCU). Through cooperation with AMD, the company has obtained the core technology of x86 processor design, and launched Haiguang No. 1 product design in 2016, officially entering the field of x86 processor design. Adhering to the product R&D strategy of "selling one generation, verifying one generation, and R&D one generation", the company has established a complete research and development link and process for high-end processors, and has continuously made breakthroughs in technology, and the performance of the CPU series products has been improved generation by generation. With the help of deep accumulation in the CPU field, the company entered the research and development of DCU products in October 2018, forming a high-end processor product layout mainly based on the dual series of CPU and DCU.

Haiguang CPU has a complex on-chip architecture and can be applied in multiple scenarios. According to the different product specifications of Haiguang CPU, 1-4 dies need to be packaged on a substrate. The internal structure of the die is very complex, including hardware circuits such as processor core, on-chip network, various interface controllers, etc., but also integrates complex program code (" microcode system"). Based on the company's cutting-edge technology, Haiguang CPU series products are compatible with the international mainstream x86 processor architecture and technical routes. They have advantages such as advanced process, excellent system architecture, and rich software and hardware ecology. They are widely used in application scenarios of servers and workstations, covering telecommunications, finance, the Internet, education, transportation and other fields.

Haiguang CPU has now formed a multi-series and multi-generation product matrix. With the continuous upgrading of technology, Haiguang CPU products are constantly iterating and product series are constantly expanding. From the perspective of application scenarios, Haiguang CPU products can be planned as 7000, 5000 and 3000 series, which are respectively used in high-end servers for complex fields such as data centers and cloud computing, mid- and low-end servers for government affairs, enterprises, and education, and workstations and edge computing servers for entry-level computing fields. From a generational perspective, Haiguang CPU products can be divided into Haiguang No. 1, 2, 3, 4, etc. As of the end of 2021, Haiguang 1 and Haiguang 2 have been commercially used, Haiguang 3 has completed laboratory verification, and was officially released for mass production in June 2022. Haiguang 4 is in the R&D stage.

Haiguang products are named in a mixture of numbers and letters.From the perspective of product categories, C86 refers to the CPU series products, and C87 refers to the DCU series products; from the perspective of product models, gear symbols 7, 5, 3, and 8 refer to high-end, mid-end, low-end processors and coprocessors, and intergenerational symbols 1, 2, and 3 refer to first-generation, second-generation and third-generation products, for example, Haiguang One is the first-generation CPU product, and the specific models include 31xx, 51xx and 71xx series; while the second-generation CPU product Haiguang Two product models include 32xx, 52xx and 72xx series.

Haiguang DCU is compatible with "CUDA-like" environment. Haiguang DCU is a type of GPGPU. It is fully compatible with ROCm ("CUDA-like") GPU computing ecosystem. It can better adapt to and adapt to international mainstream commercial computing software and artificial intelligence software. It is widely used in computing-intensive application fields such as big data processing, AI, and commercial computing. It provides high-performance, high-energy-efficient computing power for applications, and supports high-complexity and high-throughput data processing tasks. From the perspective of product structure, the composition of Haiguang DCU is similar to that of CPU, with simpler structural logic than CPU, but more computing units. In terms of product types, Haiguang DCU product is planned to be Haiguang 8000 series. Currently, the Shensuan No. 1 product of this series has been commercially available, and Shensuan No. 2 is in the R&D stage.

1.2. Financial Analysis: Revenue grows rapidly, opening a performance harvest period

The company's revenue continues to grow at a high rate. Before 2018, the company focused on technology and product research and development, and its revenue scale was relatively small. Since 2019, with the large-scale production of Haiguang No. 1, No. 2 and Shensuan No. 1, the company's operating income has grown rapidly. In 2021, the company achieved operating income of 2.31 billion yuan, and increased by by 126.07% year-on-year, and the CAGR from 2017 to 2021 was as high as 258.40%. In the first half of 2022, the company has achieved operating income of 2.53 billion yuan, which has exceeded the annual revenue level in 2021. With the future large-scale shipments of Haiguang III, the company's revenue is expected to maintain steady growth.

Haiguang CPU 7000 series contributes more than 60% of its revenue, and intergenerational replacement helps accelerate revenue. From the perspective of product line, from 2019 to 2021, the 7000 series CPU products contributed the majority of revenue (65% in 2021) and maintained rapid growth, from 325 million yuan in 2019 to 1.502 billion yuan in 2021. The revenue of the 5000 series and 3000 series products has also increased year by year, with revenue of RMB 192 million and RMB 378 million respectively in 2021. In 2021, the company's DCU products (8100 series) achieved sales for the first time, with revenue of 239 million yuan, accounting for 10.3% of the company's operating income. From an intergenerational perspective, Haiguang 2 achieved mass production in 2020, and its revenue was equally comparable to Haiguang 1; in 2021, Haiguang 2 became the main tool for the company's revenue, achieving revenue of 1.78 billion yuan that year, accounting for as high as 77.06%.

sales are released quickly, making up for the decline in unit prices across generations year by year. In terms of product sales, the 7000 series and 3000 series contributed the main sales. After the intergenerational conversion, Haiguang No. 2 quickly increased its volume, replacing Haiguang No. 1's share, and the overall sales showed a rapid upward trend. From the perspective of product unit price, Haiguang No. 2 has increased significantly compared with Haiguang No. 1, and the unit price of each sub-series has decreased year by year with the launch time. Benefiting from the rapid increase in the company's products, overall revenue has grown rapidly.

net profit attributable to shareholders turns losses into profits, opening a period of performance harvest. From 2017 to 2020, the company's net profit attributable to shareholders was negative, mainly because the company was still in the early stage to the rapid development period, with high R&D investment and relatively small revenue scale. In addition, the company implemented multiple equity incentives for key employees and also generated large share payment expenses. In 2021, the company's net profit attributable to shareholders turned losses into profits, achieving 327 million yuan; in the first half of 2022, the company's net profit attributable to shareholders exceeded the previous year's level, reaching 476 million yuan. In terms of profitability, the company's net profit margin has been rising year by year, reaching 18.94% in 2021. In the first half of 2022, it continued to explore the upward space of net profit margin, achieving 27.04%.

(report producer/author: Northeast Securities, Huang Jing, Li Jiu)

1. Haiguang Information: The leader of the independent advanced microprocessor industry

1.1. Haiguang Information: The leading domestic high-end processor company

1.2. Financial Analysis: Revenue grows rapidly, opening a performance harvest period

scale effect raises the gross profit margin level, and the cost control ability continues to be enhanced. The company's main business cost of includes direct costs (direct material procurement fees, packaging and testing fees) and indirect costs (self-developed intangible assets amortization, intellectual property fees and other manufacturing costs).In 2021, the company's direct costs accounted for 64.19% of the main business costs ( direct materials 36.21%, packaging testing fee 27.97%), and the indirect costs accounted for 35.81% of the main business costs (self-developed intangible assets amortization 25.30%, intellectual property fees 10.05%, and other manufacturing costs 0.46%).

As the company's business scale grows, the proportion of self-developed intangible assets amortization amount in operating income continues to decline, which was 32.86%, 17.18% and 11.14% in 2019-2021, respectively. Under the scale effect, the gross profit margins of comprehensive and various businesses have increased accordingly. In 2021, the company's comprehensive gross profit margin reached 55.95%, an increase of 6.04 percentage points from the previous year. The company's expense control capabilities continue to increase, and the period expense rate has declined year by year, from 74.23% in 2019 to 37.88% in 2021, mainly due to the significant increase in the company's sales scale and the growth rate of various expenses is lower than the revenue growth rate. Under the scale effect, we expect the company's gross profit margin and period expense ratio to be further optimized, and we will continue to explore the upward space for profit.

2. Industry: CPU and GPGPU market long slope and thick snow

2.1.CPU: Downstream growth steadily, my country's CPU market space is vast

2.1.1.CPU: The core component of a computer, with wide downstream applications

CPU is the core component of a computer. CPU refers to central processing unit (Central Processing Unit), which is the computer's computing and control core, the final execution unit of information processing and program operation, and the core component of the computer. The essence of a CPU is a super-large-scale integrated circuit, which is used to interpret computer instructions and process data in computer software, and is responsible for controlling and configuring all software and hardware resources of the computer. According to the instruction set adopted, the CPU can be divided into two categories: CISC ( complex instruction set ) and RISC (lite instruction set). The characteristics of the two architectures are different. Among them, the x86 architecture is the representative of CISC, and ARM, MIPS, Alpha, etc. are representatives of RISC. The production of

CPU includes three links: design, manufacturing, packaging and testing. As one of the most typical integrated circuit products, the industrial chain structure of CPU is consistent with that of integrated circuits, and its production includes three links: design, manufacturing, packaging and testing. Among them, design refers to converting the specific requirements of the system, functions and performance into physical layout; the manufacturing link refers to the process of building a complete circuit according to the physical layout on the prepared wafer material; and packaging test is to cut, solder and package qualified wafers, connect the chip circuit to external devices, and conduct functional and performance testing of the packaged chip.

CPU can be applied to servers, workstations, PCs, mobile terminals, embedded devices and other devices. For different device types, there are certain differences in the technical indicators of the CPU architecture, functions, performance, reliability, energy efficiency ratio, etc. Haiguang CPU is mainly used in computing and storage devices such as servers and workstations. Among them, the CPU chip of the server has relatively strict requirements in terms of performance, reliability, scalability and maintainability. It is the processor with the strongest data processing capabilities, the most complex design process and the highest reliability. It is used in key scenarios in major industries. The work station is a high-end microcomputer, which mainly provides single users with more powerful performance than personal computers, and is mainly used in scientific and engineering computing, software development, computer-aided design and other scenarios.

2.1.2. Space: The server market is growing steadily, and my country's x86 CPU market has a broad space

The global server market is growing steadily. Over the past decade, the global server market has generally maintained steady growth. According to IDC data, global server shipments in 2020 were 12.129 million units, with sales of US$91.02 billion, an increase of 3.26% and 4.37% year-on-year respectively. It will continue to maintain an upward trend in the future. Whether in terms of amount or quantity, the x86 architecture occupies the main market share of the server. Since the x86 processor started early and the ecological environment has obvious advantages over other processors, servers that use x86 processors occupy an absolute majority in the market. According to IDCh data, from 2014 to 2020, the average sales of x86 architecture servers accounted for about 91%, and the shipment volume accounted for more than 97%.

my country's x86 server market size growth rate is higher than the global level, and it is expected that the CAGR will reach 8.8% from 2021 to 2025.According to IDC data, in 2020, the global x86 server market sales were US$82.65 billion, a year-on-year increase of 3.31%; sales volume was 11.802 million units, a year-on-year increase of 1.82%. In the same year, the size of China's x86 server market was US$21.87 billion, a year-on-year increase of 16.5%; shipments were 3.439 million units, a year-on-year increase of 8.1%. The growth rate of the x86 server market size and shipment volume in my country is significantly higher than that in the world. In the future, with the recovery of downstream demand and the country will accelerate the progress of new infrastructure such as 5G, industrial Internet, big data centers, artificial intelligence, etc., it is expected that my country's x86 server shipments will continue to grow at a compound growth rate of 8.8%, reaching 5.252 million units by 2025.

my country's x86 server CPU market space is vast, and it is expected that the shipment volume will reach tens of millions in 2025. According to IDC data, my country's x86 servers are mainly dual-channel servers. From 2016 to 2020, the proportion of dual-channel servers remained above 80%. Based on the calculations of the shipment volume and distribution of x86 servers in my country in 2020, it is estimated that the shipment volume of x86 CPU chips in my country will be 6.981 million. Assuming that the distribution of the number of channels in 2021-2025 is consistent with that in 2020, combined with IDC's scale forecast for my country's x86 servers, it is estimated that the CPU shipments of my country's x86 servers will reach 10.662 million in 2025, with a compound growth rate of 8.8% from 2021-2025.

2.1.3. Schematic: Overseas giants form a monopoly, and domestic manufacturers continue to accumulate strength

Overseas x86 CPU market two heroes compete for the dominance. For the x86 architecture, the dual giants of Intel and AMD monopolize the global market. Among them, Intel relies on its powerful x86 ecosystem to occupy a leading position in the general CPU market, with market share remaining at around 80% for many years. Since AMD released the Zen series CPU in 2017, its market share has been increasing and its momentum is catching up. According to Mercury Research, in 2021Q4, Intel and AMD's global x86 CPU market share reached 74.4% and 25.6% respectively, and the two market share in the server field reached 89.3% and 10.7% respectively.

my country's domestic CPUs have formed a market structure of "four technical paths and six major manufacturers". As the most important core component of chips and even the information industry, CPU is the focus of my country's long-term support and development. At present, my country's domestic CPUs have gradually formed a market structure of "four technical paths and six major manufacturers", including Kunpeng and Feiteng using ARM architecture, Haiguang and Zhaoxin using x86 architecture, Shenwei using Alpha architecture, and LoongArch/MIPS architecture, Loongson using LoongArch/MIPS architecture. At present, most of the six major domestic CPU manufacturers adopt cooperation with foreign countries, including purchasing instruction set authorization, technical cooperation, etc.

policy is favorable, and domestic CPU manufacturers are expected to benefit. Against the backdrop of intensifying competition in the field of science and technology, the support of Chinese government for domestic CPUs has gradually increased. In August 2020, the State Council issued the "Several Policies to Promote the High-Quality Development of the Integrated Circuit Industry and Software Industry in the New Era", pointing out that China's chip self-sufficiency rate should reach 70% by 2025, creating new opportunities for domestic CPU manufacturers. With the continuous improvement of domestic CPU performance and the continuous improvement of the software ecosystem, information products based on domestic CPUs have been batch reused. With the acceleration of informatization in the future, the demand for domestic CPUs will continue to increase, and domestic CPU manufacturers are expected to benefit.

2.2.GPGPU: HPC. AI integration development creates a high-growth market

2.2.1.GPGPU: GPU for general computing

GPU can be differentiated into two branches: traditional GPU and GPGPU. GPU (Graphics Processing Unit, graphics processor ), also known as display core , vision processor , and display chip, is a microprocessor specializing in image and graphics-related computing on PCs, workstations, game consoles and some mobile devices. The original intention of the GPU is to improve the computer's processing performance of images, videos and other data.With the manifestation of GPU's advantages in parallel computing performance and the expansion of the scope of parallel computing applications, GPUs have gradually differentiated into two branches: 1) Traditional GPU: continues the purpose of graphics and image processing, built-in dedicated computing modules such as video encoding and decoding, 2D and 3D acceleration engines and image rendering; 2) GPGPU (General-Purpose computing on Graphics Processing Units, graphics processor for general computing), as a computing coprocessor, for the needs of different computing scenarios, it has added computing instructions such as vectors, tensors, and matrixes to improve the accuracy and performance of floating-point computing . Haiguang DCU products are designed and released based on general GPGPU mechanisms and are a type of coprocessor suitable for computing-intensive and computing acceleration fields.

GPGPU is widely used in the fields of commercial computing, big data processing, and AI. With the development of technology and the gradual improvement of the ecology, the use of GPGPU has been further explored. GPGPU has been widely used in commercial computing and big data processing in weather forecasting, industrial design, gene engineering, financial engineering and other fields. Currently, in application scenarios such as supercomputing and big data processing, computing power is mostly built using CPU+GPGPU or with dedicated acceleration chips. Taking supercomputing as an example, 70% of the top 500 global supercomputing systems in 2020 use GPGPUs, and 20 of the top 25 use GPGPUs. In the field of AI, using GPGPU to run model training algorithms in the cloud can significantly shorten the training time, reduce energy consumption, and reduce the application cost of AI. According to Frost & Sullivan, GPU accounted for the largest proportion of AI chips in 2019, reaching 36.54%, and is expected to increase to 51.4% in 2024. GPU is also the preferred acceleration solution for AI servers. According to IDC research, the proportion of GPU-accelerated AI servers in my country exceeded 90% in the first half of 2021.

2.2.2. Space: Downstream data centers and AI tracks are in high prosperity, and GPGPU development potential is great

Digital economy The "explosive" growth in data volume and computing power demand in the era boosts GPGPU demand. With the development of new technologies such as artificial intelligence and data mining, the integrated circuit industry has ushered in a stage of leading development in data centers, and the calculation and processing of massive data have also become a new driving force for the development of the industry. According to Cisco, global data center load tasks will reach nearly 5.7 million in 2021, more than twice that in 2016. With the continuous development of cloud computing, the construction of cloud data centers and super data centers around the world is accelerating. Cisco expects 628 super data centers with stronger computing power to reach 53% of the total data centers in 2021. With the continuous growth of computing power demand, the demand for computing accelerated hardware mainly based on GPGPU will also continue to rise.

In addition to data centers, the acceleration of computing in the AI field has gradually formed a situation where GPGPU solutions are the main ones. According to data from the Forward Industry Research Institute, my country's AI chip market size will reach 78.5 billion yuan in 2024, with a compound growth rate of 45.11% from 2019 to 2024. With the advancement of artificial intelligence-related technologies, application scenarios will be more diversified, and the good versatility of GPGPUs and the perfect software ecosystem will be further demonstrated. GPGPUs are expected to occupy a large proportion of the market share in my country's artificial intelligence chip field.

As the downstream application scenarios gradually deepen, the development potential of GPGPUs is promising in the future. With the continuous improvement of GPGPU chip performance, its application scenarios in big data processing, artificial intelligence and other application scenarios will gradually deepen: 1) Commercial computing and big data processing: In the digital age, people have increasingly high requirements for information processing. Big data processing is not only applied to traditional fields such as oil exploration and weather forecasting, but also the demand for big data processing and commercial computing in the Internet, finance, data mining, and education fields also show a rapid growth trend; 2) Artificial Intelligence Processor: In the future, new algorithms will appear for different artificial intelligence application types and scenarios, and new requirements will be made for the performance and flexibility of artificial intelligence chips.

GPGPU has a complete software ecological environment and programming flexibility, which can better adapt to the development needs of multiple fields of AI and is expected to become the main force of artificial intelligence processors; 3) Intelligent Computing Center: The essence of the Intelligent Computing Center is the supply and production platform of computing power. It sells computing power to artificial intelligence companies and helps them turn technical models into reality. Due to the evolution of algorithms, artificial intelligence has put forward a huge demand for computing power, and the industry predicts that GPGPUs will drive the artificial intelligence inference performance to more than double every year. Under the response of national policies, local governments have set off a boom in the construction of large intelligent computing centers, and have layouts in Wuhan, Shanghai, Dalian, Zhengzhou, Xi'an, Hefei and other places.

2.2.3. Schematic: Nvidia is the dominant, and domestic manufacturers have great potential for

The global GPU market is "three parts of the world", and Nvidia is the dominant GPGPU market. From the perspective of the global landscape, the global GPU market has formed a situation where Nvidia, AMD and Intel have "three points of the world". In terms of GPGPU, in terms of global structure, Nvidia is the dominant, AMD and Intel are the followers, and more than 90% of the market in my country's GPGPU field is also in the hands of international giants such as Nvidia. The GPGPU market controlled by foreign giants has brought many problems, such as expensive prices, weak bargaining power for downstream industry users, single varieties, and industrial chain security risks. Realizing independent research and development of GPGPUs has become a major urgent need in the domestic chip industry.

The domestic production process is accelerating, and domestic GPGPU manufacturers have great potential. In recent years, under the influence of factors such as the macro environment (national policy orientation, strong calls for domestic substitution), talent reserves (NVIDIA and others have cultivated a considerable number of outstanding GPU talents in China), capital support (the semiconductor financing environment is improving, and Science and Technology Innovation Board provides an exit mechanism), a wave of entrepreneurship and financing has appeared in the GPGPU field in China. At present, major players such as Tianshu Intelligent Core, Biren Technology, Denglin Technology, Moore Thread have been gathered. In the future, we believe that in the face of broad market prospects, domestic GPGPU manufacturers will have great potential.

3. Advantages: establish a business in technology, establish a product, and build a high-quality industrial chain

3.1. Technology: Deep technical background, continuous research and development, innovation and accumulation of strength for long-term development

0 The company's management has a deep technical background and rich industry experience. The research and development and production of high-end processors require the industry's most cutting-edge technologies and the most advanced process processes, and have extremely high R&D and production barriers. Many members of the company's management have a technical background and are the company's core technical personnel. They have made great contributions to the research and development and mass production of Haiguang CPU and DCU products. In addition, the company's management has rich industry management experience. Chairman Mr. Meng Xiantang was once the deputy general manager of Guoke Holdings Co., Ltd.; Mr. Sha Chaoqun, director and general manager, has served as vice president and senior vice president of Inspur Technology.

A professional R&D team is the cornerstone of the company's development and growth. After more than six years of training, the company has a professional high-end processor research and development team. Most of the company's backbone R&D personnel have the employment background of well-known domestic and foreign chip companies, and have the experience of successfully developing x86 processors or ARM processors . Most core R&D personnel have more than 20 years of experience in R&D of high-end processors. The R&D team has a reasonable structure, solid theoretical foundation, rich practical experience and comprehensive skills, which can strongly support technological innovation and product research and development. The company has established a R&D team of thousands of people in Beijing, Chengdu, , Suzhou, , Shanghai and other places. As of the end of 2021, the company had 1,031 R&D technical personnel, accounting for 90.2%, of which 749 were postgraduate students, accounting for 72.65%. The number of R&D personnel and the proportion of R&D personnel in the company are at a high level in the industry, laying the foundation for the company's continuous development and growth.

The company has always attached importance to and encouraged R&D innovation, and has continued to maintain high R&D investment in recent years. From 2019 to 2021, the company's R&D expenditure increased year by year, from 865 million yuan in 2019 to 1.585 billion yuan in 2021, with a CAGR of 35.4%. In 2019 and 2020, the company's R&D expenditure exceeded its current revenue. Due to the rapid growth of the company's revenue, the proportion of R&D expenditure in current revenue has decreased.According to the horizontal comparison, the company's R&D investment rate from 2019 to 2021 was 228.04%, 106.60% and 68.60%, respectively, which was at a high level in the industry. In 2020, the company's R&D investment rate dropped by more than 100%, but it was still far higher than the comparable company average in the same industry.

AMD Cross-licensing provides strong technical support for the company. In March 2016 and October 2017, Haiguang Microelectronics and Haiguang Integration signed a "Technical License Agreement" with AMD, respectively, stipulating that AMD will license high-end processor-related technologies and software to the two joint ventures. Through the technical license and related technical support of AMD's high-end processor, the company has obtained the core technology of x86 processor design and entered the field of x86 processor design. In 2018, it developed and launched a domestic x86 architecture processor product that meets the needs of Chinese users and has the three major characteristics of "ecology, performance, and security", namely Haiguang No. 1. Since then, through independent innovation and product iteration, the company has mastered its own core CPU and DCU production technologies, and has consolidated its leading position in finance, communications, government and other fields.

company has made significant progress in all core links of processors and has mastered more than ten core technologies. The company adheres to the concept of providing users with high-performance, safe and controllable products, and continues to invest in the fields of CPU and DCU chip technology. Through years of research and development, the company has made significant progress in all core links of processors, breaking through and mastering more than ten core technologies including high-end general-purpose processor microstructure design, high-end coprocessor microstructure design, high-end processor SoC architecture design, and processor security. These core technologies are independently developed, and have reached a mature and stable state in terms of maturity. Most of them have been used in Haiguang processor products and are at the advanced level in the industry. The company has formed a large number of independent intellectual property rights during the R&D process. As of July 26, 2022, the company and its subsidiaries have 179 authorized patents (including 136 invention patents), 154 software copyrights, and 81 integrated circuit layout design exclusive rights, etc.

deep technology reserves, continuous research and development and accumulation of strength for long-term development. As a technology-intensive enterprise, the company adheres to the development strategy of independent research and development of core technologies. The company adheres to the market-oriented arrangement of R&D plans, aiming at customer needs, and combines the advantages and characteristics of its own technology and products to continuously optimize the organizational structure and R&D management process, effectively ensuring the sustainability of technological innovation. Through the smooth implementation of independent research and development projects under development, the company's technical position and product leadership will be consolidated.

3.2. Product: High performance + strong ecology, strong product strength builds solid barriers

3.2.1. Performance: Haiguang CPU and DCU performance leading domestic

company's product performance is strong, benchmarking the international leading level. The company was the first in China to successfully complete high-end general-purpose processor and coprocessor products, and realize commercial applications. The company's product performance has reached the level of the same type of mainstream high-end processors internationally and is in a leading position in the country. Haiguang CPU: Haiguang 7258 has similar performance to Intel's mainstream products during the same period. Select 6 products released by the global leader Intel in 2020 (similar to Haiguang 7285) that can reflect the performance level of its mainstream products that year. The SPEC CPU 2017 test was tested in the industry internationally, and compared the performance with Haiguang 7285. Through comparison, it can be seen that the measured performance of Haiguang 7285 is generally comparable to that of mainstream processor products released by Intel at the same time.

Haiguang CPU has strong security advantages. With the increase in system complexity, the growth of cloud computing and the emergence of new technologies, the difficulty and importance of computing environment security have also increased. Haiguang CPU supports the national secret algorithm and has built-in security processors that can perform memory encryption, virtualization encryption, and container encryption. It has strong security advantages and has well met the needs of various critical information infrastructure for data security. At the same time, Haiguang CPU supports general trusted computing standards, which greatly improves the security of high-end processors and provides users with more efficient security guarantees during data processing.

Haiguang DCU: The performance of Shenshou No. 1 is comparable to the international high-end product level.Haiguang DCU is the only accelerated computing chip in China that supports scientific computing and artificial intelligence to accelerate large-scale applications. Select Shensuan 1 to compare it with the high-end GPU products (model A100) of NVIDIA, a leading international GPU manufacturer, and AMD’s high-end GPU products (model MI100). It can be seen that in typical application scenarios, the company's Shenzhen Computing No. 1 index has reached the international high-end product level of the same type.

3.2.2. Ecology: A good ecology paves the way for future development

Haiguang CPU: compatible with x86 instruction set, with a mature and rich application ecological environment. It supports mainstream operating systems, databases, virtualization platforms or cloud computing platforms at home and abroad, and can effectively compatible with millions of system software and application software based on the x86 instruction set. The major domestic first-line system manufacturers and some emerging manufacturers have adopted Haiguang products for designing solutions. Various manufacturers have used Haiguang CPUs to form a complete product family around different types of markets, including general rack servers, AI servers, storage and network security products, which can cover all applications in the entire IT information system data center. Haiguang DCU: compatible with "CUDA-like" environments and can better adapt to international mainstream commercial computing software and artificial intelligence software. The company actively integrates into the domestic and foreign open source community and actively provides adaptation and optimization solutions for Haiguang CPU and DCU to the open source community, ensuring the compatibility of the company's processor products in the open source ecosystem.

Create a photosynthesis organization to build a prosperous ecosystem. Haiguang Industrial Ecological Cooperation Organization (hereinafter referred to as Photosynthetic Organization) is an ecological alliance initiated by the company. It is mainly committed to industrial cooperation and innovation, promote the construction of the digital economy, and promote the long-term development of the domestic information industry ecosystem. With the growth of the scale of Photosynthetic Organization, more and more domestic manufacturers have carried out ecological construction and adaptation of Haiguang processors, and have carried out research and development, mutual certification and continuous optimization in operating systems, databases, middleware, cloud computing platform software, AI technology framework and programming environment, core industry applications, etc., and developed a number of influential domestic complete machine systems, basic software and application software, and initially formed a domestic software and hardware ecosystem based on Haiguang CPU and DCU.

's company's products have entered many well-known domestic server manufacturers such as Inspur, Lenovo, H3C, and Tongfang, and the customer richness continues to increase. Since 2019, the company has expanded its sales team and its products have gained increasingly recognition from more and more customers. At present, the products of many well-known domestic server manufacturers including Inspur, Lenovo, H3C and Tongfang have been equipped with Haiguang CPU chips, and have been successfully applied to financial, energy and chemical customers such as Industrial and Commercial Bank of China, Bank of China, China Petroleum and Sinopec, and have also been widely used in the data center business of telecom operators. As the products gradually gain customer recognition, new customers continue to be introduced, and the richness of the company's customers continues to increase. Although the sales revenue of the company's products is still mainly concentrated in the top five customers, the proportion of the top five customers is gradually decreasing.

3.3. The implementation of IPO fundraising and investment projects will further enhance the company's competitiveness

fundraising and investment projects will further enhance the company's competitiveness. According to the company's prospectus, the company plans to raise 9.148 billion yuan to invest in the research and development of the new generation of Haiguang general processor and coprocessor, and the construction of advanced processor technology R&D centers. The implementation of this fundraising and investment project can promote the company's technological development, enhance core competitiveness, expand the application fields of sea light processors, promote the company's product sales, and achieve the company's healthy development.

The new generation of Haiguang General Processor R&D project: It plans to invest 2.878 billion yuan to raise funds. The project will integrate the latest achievements of the company's processor design technology and adopt advanced process processes. Compared with previous generations of products, the functions, performance and energy efficiency ratio will be significantly improved. The new generation of Haiguang universal processors are based on advanced process processes, expanding the on-chip shared cache capacity, supporting advanced DDR memory, supporting advanced PCIe standard interfaces, supporting cache consistency between multiple CPU chips, supporting hardware virtualization, and providing a complete hardware and software development platform and tool chain.

The new generation of Haiguang Coprocessor R&D project: It plans to invest 1.856 billion yuan to raise funds.This project will make full use of the company's R&D achievements in the DCU field, break through the key technologies of the new generation of Haiguang Coprocessor R&D, improve the DCU ecosystem, realize the large-scale sales of the new generation of Haiguang Coprocessor series products, and achieve the performance of products of the same period of the international leading companies.

Advanced Processor Technology R&D Center Construction: It plans to invest 2.414 billion yuan to raise funds. Through the implementation of this project, the completion of forward-looking technology research for the next generation of Haiguang CPU and DCU will significantly enhance the company's CPU and DCU chip technology research and development capabilities, and support the continuous research and development of the next generation of products; break through advanced chip packaging technology, and further improve the integration of the next generation of Haiguang CPU and DCU chips; overcome the key engineering technology for the research and development of high-end chip products, improve the quality rate of the next generation of Haiguang CPU and DCU chips, increase production capacity, and reduce production costs; break through the Haiguang processor software and hardware enable technology, improve the Haiguang chip ecosystem, and promote the large-scale sales of Haiguang CPU and DCU chips. Science and Technology and Development Reserve Fund: The reserve funds are approximately RMB 2.00 billion, and the specific purpose is to conduct forward-looking technical research and supplement working capital.

4. Profit forecast

The company's revenue mainly comes from the sales of CPU products 7000 series, 5000 series, 3000 series and DCU products 8000 series. By series:

1) 7000 Series: 7000 Series is mainly used in high-end servers, facing complex fields such as data centers and cloud computing, and is the main tool for company revenue. From a quantitative perspective, with the recovery of downstream market demand and the country's progress in the construction of new infrastructure in seven major fields such as 5G, industrial Internet, big data centers, and artificial intelligence, CPU chips applied to the high-end server market will continue to grow rapidly. With the large-scale shipment of Haiguang III in the second half of 2022, the 7000 series is expected to accelerate its volume. We expect the sales growth rate of the 7000 series from 2022 to 2024 to be 50%/20%/10% respectively; from the perspective of price, the unit price of the product shows an increasing trend with the generation, and the 71xx products with lower unit price in 2021 still account for 38% of the shipment volume of the 7000 series. With the gradual replacement of the 73xx products with high unit price, the scale shipment of 74xx products and the low unit price 71xx products, the average unit price of the 7000 series has a lot of room for improvement. We expect the unit price growth rate of the 7000 series from 2022 to 2024 to be 40%/20%/20% respectively.

2) 5000 Series: 5000 Series is mainly aimed at the needs of mid- and low-end servers in the fields of government affairs, enterprises and education. From a quantitative perspective, with the continuous deepening of industrial development, the speed of domestic substitution in government affairs and other fields is expected to increase. According to the requirements of the State Council’s "Several Policies to Promote the High-Quality Development of the Integrated Circuit Industry and Software Industry in the New Era", China’s chip self-sufficiency rate will reach 70% in 2025, and there is still a lot of room for improvement. We expect the sales growth rate of 5000 series products from 2022 to 2024 to be 60%/25%/10% respectively; from a price perspective, the unit price of 52xx products in 2021 is significantly higher than the unit price of 51xx products, and the shipment of 51xx products is small. The increase in the future average unit price mainly comes from the pull of new products such as 53xx with higher unit price. We expect the unit price growth rate of 5000 series from 2022 to 2024 to be 25%/15%/15% respectively.

3) 3000 Series: 3000 Series is mainly used in fields such as workstations and edge computing servers. From a quantitative perspective, the market size of my country's edge computing servers shows a rapid growth trend. According to IDC data, my country's edge server sales will reach 22.2% from 2021 to 2025, showing great development potential. Downstream market demand boosted, and the sales of Haiguang 3000 series maintained rapid growth from 2020 to 2021, and are expected to maintain a high growth trend in the future. We expect the sales growth rate of the 3000 series from 2022 to 2024 to be 80%/30%/20% respectively; from the perspective of price, the 3000 series in 2021 is mainly dominated by 32xx products. With the release of new products of Haiguang No. 3 and No. 4, it will slightly drive the average unit price increase. We expect the unit price growth rate of the 3000 series from 2022 to 2024 to be 15%/10%/10% respectively.

4) 8000 Series: 8000 Series is a company's DCU product, mainly used in computing-intensive application fields such as big data processing, AI, and commercial computing. The company's Shensuan No. 1 has been commercially available, and Shensuan No. 2 is expected to be mass-produced in small batches by the end of the year.From a quantity perspective, the downstream big data processing, commercial computing and artificial intelligence technology markets are constantly expanding, and the demand for DCU products will continue to grow. As the better performance Shensuan No. 2 is commercially available, DCU product sales also maintain rapid growth. We expect the sales growth rate of the 8000 series from 2022 to 2024 to be 160%/50%/40% respectively; from the price perspective, the unit price of the company's products is showing a downward trend year by year. We expect this trend to be also applicable to the 8000 series products. The average unit price of the DCU will decline in 2022. With the large-scale shipment of the DCU Shensuan No. 2, new products are expected to boost the average unit price. We expect the growth rate of the 8000 series from 2022 to 2024 to be -5%/20%/10% respectively.

(This article is for reference only and does not represent any of our investment advice. If you need to use relevant information, please refer to the original text of the report.)

selected report source: [Future Think Tank] Future Think Tank - Official website According to IDC data, in 2020, the global x86 server market sales were US$82.65 billion, a year-on-year increase of 3.31%; sales volume was 11.802 million units, a year-on-year increase of 1.82%. In the same year, the size of China's x86 server market was US$21.87 billion, a year-on-year increase of 16.5%; shipments were 3.439 million units, a year-on-year increase of 8.1%. The growth rate of the x86 server market size and shipment volume in my country is significantly higher than that in the world. In the future, with the recovery of downstream demand and the country will accelerate the progress of new infrastructure such as 5G, industrial Internet, big data centers, artificial intelligence, etc., it is expected that my country's x86 server shipments will continue to grow at a compound growth rate of 8.8%, reaching 5.252 million units by 2025.