Analysis of the Current Status and Future Development Trends of China's AI Chip Industry in 2023
-
AI chips, also known as AI accelerators or computing cards, are modules specifically designed to handle large-scale computational tasks in artificial intelligence applications (other non-computational tasks are still managed by CPUs). Currently, AI chips are mainly categorized into GPU, FPGA, and ASIC.
Industry Development History
The development of AI chips can be traced back to 2016 when AI chips emerged as a new field, with related technologies and compilers gradually maturing and forming relatively stable architectural designs. The AI chip industry has gone through multiple stages, from the initial GPU-dominated era focused on algorithm training, to the FPGA and ASIC era for inference tasks, and now to the era of dedicated AI chips. Each stage reflects the computational demands and technological advancements of artificial intelligence at different developmental phases.
Industry Development Policies
Currently, China's AI chip industry heavily relies on imports. In recent years, the national level has placed great emphasis on the development of the AI chip industry, issuing a series of supportive policies since 2016. These policies have created a favorable environment for the AI chip industry, promoting its prosperous growth.
Industry Chain Structure
The AI chip industry chain mainly includes chip design, manufacturing, packaging, and testing. The upstream design and manufacturing of AI chips require significant technical accumulation and R&D investment. AI chips have a wide range of applications, including intelligent security, autonomous driving, smartphones, smart retail, and intelligent robotics. With the improvement of technical capabilities and increased R&D investment, the design and manufacturing levels of AI chips will continue to rise.
Downstream Application Analysis
With the rapid development of China's internet, 5G, cloud services, and satellite positioning technologies, significant technical support has been provided for the advancement of autonomous driving. In recent years, the market size of China's autonomous driving industry has grown rapidly. Data shows that in 2022, the market size of China's autonomous driving industry reached 289.4 billion yuan. Autonomous vehicles are an important direction for future automotive development, and the demand from industries like autonomous driving will drive the expansive growth of the AI chip industry.
Analysis of the Current Status of the AI Chip Industry
China's AI chip industry market size continues to expand, gradually becoming an important force in the global AI chip market. Data shows that the market size of China's AI chips was about 6.4 billion yuan in 2018 and grew to 85 billion yuan in 2021, with an average annual compound growth rate of 67.7%.
Industry Investment Situation
The investment in China's AI chip industry has shown a trend of year-on-year growth. With the continuous development and popularization of artificial intelligence technology, the demand for high-performance, low-power AI chips is increasing. The AI chip market is expected to grow at a high speed, creating huge business opportunities. According to data, there were 93 investment cases in China's AI chip industry in 2022, with an investment amount of 21.515 billion yuan. Although the number of financing events decreased in 2022, AI chips remain a hot spot for both domestic and foreign capital.
AI Chip Market Product Structure
From the perspective of AI chip types, GPU is one of the most common types of AI chips. Its parallel computing capability and floating-point operation capability make it very suitable for artificial intelligence applications such as deep learning. In terms of market share, according to data, GPU occupies an absolutely dominant position, followed by NPU and ASIC.
Industry Competition Landscape
The competition landscape of the AI chip industry is mainly concentrated globally, with major chip manufacturers actively deploying in the AI chip market. Among them, NVIDIA is the world's largest AI chip supplier, with an absolute leading market share. According to data, China's AI accelerator card shipments in 2022 were about 1.09 million units, with NVIDIA accounting for 85% of the market share in China's AI accelerator cards, Huawei accounting for 10%, and Baidu accounting for 2%.
Key Enterprise Revenue
Key enterprises in China's AI chip industry include HiSilicon Semiconductor, Bitmain, Loongson Technology, and Jingjia Micro. HiSilicon Semiconductor is a company specializing in the research, development, and manufacturing of semiconductor chips. Its product line includes various types of AI chips such as GPU and ASIC, covering fields such as security and mobile phones. Loongson Technology is a company focusing on the research, development, and manufacturing of independently controllable computer chips. Its product line includes CPUs and GPUs based on the independent instruction set LoongArch. Jingjia Micro is a company specializing in the research, development, and manufacturing of graphics processing chips and artificial intelligence chips. Its product line includes GPUs and AI accelerators.
Industry Future Development Trends
The chip industry as a whole is encouraged and supported by policies, with AI chip development benefiting from domestic demand and the localization process.
Chips are the cornerstone of the information and digital age. As the world's largest semiconductor consumer market, China has insufficient self-sufficiency in chips and relies heavily on imports. To develop domestic chips and achieve import substitution, the government has introduced a series of policies in recent years to support the domestic chip industry.
In August 2020, the State Council issued the "Several Policies to Promote the High-Quality Development of the Integrated Circuit and Software Industries in the New Era," proposing support for the integrated circuit and software industries in eight areas, including taxation, investment and financing, research and development, imports and exports, talent, intellectual property, market applications, and international cooperation, to accelerate the development of integrated circuits and software. The development of AI chips will benefit from policy support for chip localization and the huge domestic market demand.
AI chip R&D will shift from technology-driven to scenario-driven.
Currently, AI chip design is more focused on technical requirements, such as chip architecture selection and performance improvement. As competition in the AI chip field intensifies, chip companies need to not only make breakthroughs in technology but also expand their application scenarios to seize more development opportunities. To adapt to the fragmented application market, future chip design needs to be customer-end demand-oriented, analyzing the feasibility of implementation from various aspects such as demand volume, commercial landing models, and market barriers, and achieving large-scale development of AI chips through scenario implementation.
AI chip development is shifting from cloud-focused to cloud-edge integration.
Cloud chips focus on non-real-time, long-cycle data for big data analysis and can support large-scale parallel operations. Currently, cloud AI chip applications are relatively mature. With the proliferation of applications such as smart speakers, autonomous driving, drones, and security monitoring, some inference and even training computing power from the cloud will migrate to the edge and terminal sides to support real-time intelligent processing and execution of local services. The demand for AI chips at the edge and terminal sides is more diverse, emphasizing low power consumption and low cost, with relatively lower technical requirements.
Driven by artificial intelligence and other factors, edge computing will gradually be applied in many fields such as public security, smart homes, and intelligent transportation. With the rise of edge computing, cloud-edge integrated computing power deployment solutions are becoming mainstream, not only optimizing algorithm structures but also fundamentally empowering various edge and terminal applications, providing better and more complete solutions.