IDC: In 2023, China's accelerated server market grew 104% year-on-year, and GPU servers are still dominant

The Zhitong Finance App learned that the International Data Corporation (IDC) released the latest “China Semi-Annual Accelerated Computing Market (Full Year 2023) Tracking” report. According to IDC data, China's accelerated server market reached 9.4 billion US dollars in 2023, an increase of 104% over 2022. Among them, GPU servers still dominate, accounting for 92% of the market share, reaching 8.7 billion US dollars. At the same time, non-GPU-accelerated servers such as NPU, ASIC, and FPGA accounted for nearly 8% of the market share at a 49% year-on-year growth rate, exceeding 700 million US dollars.

In 2023, in terms of manufacturer sales, Inspur, Xinhua, and Ningchang ranked in the top three, accounting for nearly 70% of the market share; in terms of the number of servers shipped, Inspur, Kunqian, and Xinhua ranked in the top three, accounting for more than 50% of the market share; from an industry perspective, the Internet is still the largest procurement industry, accounting for nearly 60% of the overall server market. In addition, most industries such as finance, telecommunications, transportation, and healthcare have more than doubled.

From a macro perspective, in addition to the two main tasks of developing higher computing power servers and chips and developing generative artificial intelligence, the AI industry will also become another focus of attention in the fields of comprehensive ecological construction and green sustainable technology. From the perspective of the market and industry chain, in the context of the continuous development of the AI industry and computing power market, the market demand for AI servers continues to increase, which will continue to drive sales growth of AI servers. IDC predicts that China's accelerated server market will reach US$12.4 billion by 2028. Among them, the non-GPU server market will exceed 33%.

The implementation of the big model will clearly drive the market for software adaptation solutions and computing power infrastructure for the AI industry. The development of artificial intelligence has begun to take shape. Applications such as voice recognition, face recognition, text generation, and graphic creation have already been implemented from research models, and the role of large model applications in daily production and life is becoming more and more obvious. Based on training on massive data and continuous optimization of models, large artificial intelligence models have more accurate execution capabilities and stronger scene migrability. Among them, AI software such as ChatGPT, Sora, Starfire, and Wenxin Yiman received even more enthusiastic response when it was publicly released. In 2024, the Chinese government is committed to the digital transformation of industries. Artificial intelligence will provide better solutions for a wide range of applications in complex scenarios such as the metaverse, urban governance, industrial manufacturing, and agricultural production. While these large models have received widespread attention, they have placed greater demands on the computing power of training AI servers.

In 2023, China's accelerator chip market will reach nearly 1.4 million units. From a technical perspective, GPU cards account for 85% of the market share; from a brand perspective, Chinese AI chip brands have shipped more than 200,000 pieces, accounting for about 14% of the total market share. Artificial intelligence chips for inference account for 67% of the market. The market continues to be enthusiastic about investing in the development of large models. The application scenarios of artificial intelligence chips are more diverse, and more targeted artificial intelligence chips are being invested in more accurate market segments. The performance of domestic chips has improved markedly, and related industrial chains are gradually being formed.

Du Yunlong, an AI infrastructure analyst at IDC China, believes that AI is a high-tech industry, and the construction of technology and supporting facilities is not complete. Coupled with interference from external factors, blind and frantic procurement, AI servers are currently in a stock market state, and there is uncertainty about technology upgrades, which may lead to a phased decline in AI server sales. Therefore, it is necessary to view the development of the AI industry rationally, recognize the complexity and long-term nature of chip technology, avoid excessive hype, and pay attention to healthy competition. Furthermore, with the continuous update of smart hardware, computing power is increasing while energy consumption requirements are increasing. Against the backdrop of increasingly acute energy and environmental protection issues, the industry has begun to pay attention to energy saving and emission reduction and exploring sustainable routes. Meanwhile, green and sustainable smart industry solutions are also a research topic in the AI industry. Ensuring increased computing power while requiring low energy consumption is a major pursuit in hardware updates.