On January 16th, it was learned from CCID Consulting Co., Ltd. that according to the company's recently released "2024 China Domestic Server Market Research Report" (hereinafter referred to as the "Report"), the domestic server market is expected to enter a critical period of development in the next three years. The growth rate of the domestic server market is much higher than the overall level of China's server market. The report shows that in 2023, the domestic server market will maintain rapid growth, with sales increasing by 28.3% year-on-year and market share rising from 11.8% in 2022 to 16.7% in 2023; The sales revenue increased by 80.9% year-on-year, with a growth rate more than 70 percentage points higher than that of China's server market. The market share increased from 17.2% in 2022 to 27.1% in 2023. State owned enterprises have become pioneers in deepening the application of domestically produced servers in the industry. The investment and layout of central state-owned enterprises in infrastructure have a demonstrative effect on the promotion of domestic servers. From the perspective of industry distribution, the total sales of domestically produced servers in the telecommunications, finance, and government sectors will account for over 85.0% in 2023; Although domestic servers in industries such as energy and transportation started relatively late, their market share has increased rapidly. The East China, Central South, and North China regions are the main battlefields for domestic server demand. The demand for computing power in Beijing, Shanghai, Guangzhou, Shenzhen and other cities is large. Large enterprises, industry headquarters and Internet companies often choose to deploy servers in these cities. In terms of manufacturer competitiveness, the report shows that in the domestic server market in 2023, brands such as Huakun Zhenyu, Henan Kunlun, and Inspur Information have performed well in the market, with a total sales share of over 50%. Among them, the market share of domestic servers with ARM architecture exceeds 50%, surpassing domestic servers with X86 architecture, making them more competitive in the market. From the specific calculation types, the sales and revenue of domestic general-purpose computing servers account for 95.0% and 67.4% of the total market share, respectively. Although the sales proportion of domestically produced accelerated computing servers is not high, the high selling price of training servers has driven a significant increase in sales, accounting for over 30% of the market sales. In the 2023 domestic accelerated computing server market, Henan Kunlun, Inspur Information, Huakun Zhenyu, Baode Computing, and Shenzhou Kuntai all rank among the top five in terms of sales and revenue. Among them, in the Ascend series server market, brands such as Henan Kunlun, Huakun Zhenyu, Baode Computing, Shenzhou Kuntai, and Changjiang Computing have strong market competitiveness, with their total sales accounting for over 70% and sales accounting for over 90%. It is worth noting that driven by factors such as continuous technological innovation, accelerated demand release, and policy support, the domestic server market is expected to enter a critical period of development in the next three years. The report predicts that by 2028, the sales share of the domestic server market is expected to exceed 50.0%, and the sales share is expected to approach 60%. From an industry perspective, the proportion of new additions in the government, telecommunications, and finance sectors remains stable, with the three major telecom operators experiencing rapid growth in their intelligent computing scale. It is expected that 2025 will continue to be a major year for telecom operator procurement; The application of large-scale models in the financial industry will extend from the inside out, which is expected to drive an increase in demand for domestic accelerated computing servers in the market. From the perspective of demand, with the iteration and improvement of domestic large-scale models, the application traffic of domestic models will gradually shift to mobile devices, and the demand for inference will gradually be released. It is expected that by 2026, the demand for inference may exceed the demand for training. (New Society)