The so-called AI chip concept generally refers to modules specifically used to handle computing tasks in artificial intelligence applications. It is the hardware used to process computing tasks that was born in the era of rapid development of artificial intelligence applications. In a broad sense, all chips oriented towards artificial intelligence applications are referred to as AI chips.
There are three main technical routes: general purpose (GPU), semi custom (FPGA), and custom (ASIC).
In the short term, GPU remains the main force in the market, and FPGA’s market share is rapidly growing. However, in the long run, ASIC is the more suitable choice for the field of artificial intelligence. In the future, the complexity of algorithms will only increase, and a dedicated chip architecture is even more needed to cope with this situation.
Encouraged by national policies, China’s emerging industries such as new energy, big data, and cloud computing are in a period of vigorous development, driving the rapid development of China’s AI chip market. According to China Industry Information Network, the global AI computer chip market showed rapid growth from 2019 to 2025, with a global AI computer chip market size of 6.2 billion US dollars in 2020, a year-on-year increase of 37.78%; It is expected that the global AI computer chip market size will reach 8.5 billion US dollars in 2021, a year-on-year increase of 37.10%; It is expected that the global AI computer chip market size will reach 30 billion US dollars by 2025.
In the situation where NVIDIA A100GPU cannot be exported to China, domestic AI chip products will benefit from domestic substitution logic.
Based on research and development investment and market performance over the past three years, Securities Star has selected the top ten concept stocks. As shown in the following figure:
April 19th Hot Topic | Top 10 Conceptual Stocks of AI Chips
Among them, the Cambrian, Opit, and Zhongke Shuguang markets have received high attention. Below is a brief review of these three stocks.
Cambrian: A scarce AI computing chip company in China. In the past 22 years, the company’s cloud product lines, such as Siyuan 290 and Siyuan 370, have successfully introduced multiple top customers such as Alibaba Cloud. In the rapid iteration of the new cloud training product Siyuan 590 chip, the floating point computing ability of the chip has significantly improved compared to the previous generation 290 product. For the Cambrian period, its biggest problem is that it has not yet achieved profitability.
Obit: The newly developed next-generation aerospace grade embedded artificial intelligence AI chip Yulong 810 has completed ECO work, achieving phased mass production and chip return. In terms of computing power customers, Obit stated in its annual report and interaction that the company provides computing power support for Hengqin Supercomputing Center. Recently, Opit released a quarterly report, with a net profit attributable to shareholders of the listed company of 27.2518 million yuan, a year-on-year increase of 240.82%.
Zhongke Shuguang: Zhongke Shuguang is the largest shareholder of Haiguang Information. Backed by the Chinese Academy of Sciences, Haiguang has deep research and development capabilities. According to Haiguang’s prospectus and Ping An Securities Research Institute, currently, the only one in China that can support FP64 double precision floating-point operations is Haiguang’s DCU Deep Computing No.1.