Due to the high customization and low budget pressure of ASIC servers, combined with the application needs of the future gradual transformation to AI recommendation era, the external estimate of ASIC's AI server shipment will continue to grow, and reiterate the "coded" ratings for Dunghai, Industrial Fulian, Xinhua, Xinhua, Qilin, Huscarta, Lianhua, Qilin, Fushida, Lianhua, Qilin, Qincheng, Taiguang, Jinxiang, Jinxiang, Taiyang, Jinhua, Taiwan Power, Fuxin Technology, Wangsi, Wangwei, and Xinhua.
Foreign data stated that in terms of AI chip market share, the ratio of GPU to ASIC is 64:36 in 2024, while in 2025 it is 62:38, and in 2026 it is estimated to be 60:40 in 2026, because ASIC servers are equipped with high customization and low budget pressure, and the application demand that will gradually shift to the AI recommendation era is expected to continue to grow.
Foreign information believes that with the advent of the inference era, AI servers using substrates may be more attractive to cloud service providers (CSPs), and estimates that the global shipment of AI servers will increase from the demand for 9 million AI chips in 2025 to 12 million chips in 2026, of which the ASIC shipment payment rate will rise to 38% in 2025 and to 40% in 2026.
Foreign point name: Since ASIC's AI server shipment volume will continue to grow, it is reiterated to give "code" ratings to Donghai, Industrial Fulian, Xinhua, Xinhua, Qilin, Huscarta, Lianhua, Brand New, Dual, Qincheng, Tai Optoelectronics, Jinxiang Electric, Taiyang Electric, Taiwan Electric, Fuzhou Industrial Technology, Wangsi, Wangwei, and Xinhua
Extended reading: Continue Mor's Law Assisting AI Development! Tokyo University has developed "new transistor" to replace silicon materials It is necessary to use NVIDIA generative AI technology! Assisting Taichung Science Museum to create AI smart navigation RobCraft AI Highlights! Daming Robot Germany Automatica exhibits three major AI collaboration applications