Xiangcai Securities: Demand for computing power is high, and end-side AI continues to iterate

The Zhitong Finance App learned that Xiangcai Securities released a research report saying that with the development of generative AI technology, AI technology is being deployed to the end side by consumer electronics manufacturers, driving the advent of a wave of AI+ consumer electronics hardware innovation. In terms of cost, since ASICs have the advantages of low cost and customization compared to GPUs, adoption can reduce dependence on Nvidia. Therefore, major domestic and foreign manufacturers are developing their own ASICs to drive rapid growth in demand for computing power ASICs. The iteration of computing power chips and switches has greatly increased the value of stand-alone PCBs, and the sharp rise in volume and price has jointly promoted the rapid growth of the AI PCB market. It is recommended to focus on companies related to the end-side AI, ASIC, and PCB industry chains.

The main views of Xiangcai Securities are as follows:

Large model technology continues to be iterated, and the wave of AI terminal innovation is coming

After years of development, traditional consumer electronics terminals have entered an era of stock or slow development. Sales of smartphones and PCs are basically stable, and TWS has entered a period of slow growth. After the release of ChatGPT, domestic and foreign companies entered the big model field one after another, promoting the continuous iteration of big model technology. The development of large model compression technology has laid the foundation for deploying large models on the end side. End-side AI has advantages such as low cost, high performance, privacy and security. With the development of generative AI technology, AI technology is being deployed to the end side by consumer electronics manufacturers, driving the wave of AI+ consumer electronics hardware innovation.

Google and Amazon sell ASICs to the outside world, and demand for ASICs with computing power is strong

Since ASICs have the advantages of low cost and customization compared to GPUs, their adoption can reduce dependence on Nvidia. Therefore, major domestic and foreign manufacturers are developing their own ASICs to drive rapid growth in demand for computing power ASICs. According to Marvell's forecast, the market size of the computing power ASIC is expected to grow from 6.6 billion US dollars in 2023 to 55 billion US dollars in 2028, and the compound growth rate of the global computing power ASIC market size will reach 53% in 2023-2028. Currently, Google supplies TPU in batches to Meta and Claude, and Amazon supplies Trainium to Claude in batches, further proving the performance and commercial value of ASIC and enhancing the certainty of ASIC industry trends.

Demand for AI computing power is driving up PCB volume and price

Major AI companies and major global Internet companies have invested huge capital expenses to expand data centers. According to Dell'Oro's forecast, the compound growth rate of global data center capital expenditure will reach 21% by 2029, and the huge capital expenditure plans of AI companies and major Internet companies will drive the continuous expansion of computing power clusters. Computing power expansion will simultaneously amplify “network bottlenecks”, turn switches from “accessories” into “key production capacity” to improve data center performance, and drive rapid growth in demand for high-speed switches. At the same time, the number of PCB layers in AI servers and high-speed switches continues to increase, and the materials used in PCBs are constantly being upgraded, driving a significant increase in the value of stand-alone PCBs, and the sharp rise in volume and price has jointly promoted the rapid growth of the AIPCB market.

End-side AI implementation is accelerating, and the penetration rate of AI terminals is expected to increase rapidly

The development of large model compression technology has laid the foundation for deploying large models on the end side. Edge AI has the advantages of low cost, high performance, privacy and security. Huawei and Byte have successively demonstrated the great application value of end-side AI on mobile phones. Through deep cooperation with mobile phone manufacturers, Doubao Mobile Assistant has deeply embedded into the operating system and realized complex operations across applications, achieving a leap in the effectiveness of AI mobile applications, showing the potential of end-side AI applications. Other smartphone manufacturers have also released AI phones with system-level AI capabilities.

Microsoft released Copilot+PC, setting high standards for AIPC and enriching AIPC applications. Both Intel and AMD will release new AIPC processors in 2025. Driven by industry chain manufacturers and the support of AI capabilities, the AI terminal penetration rate is expected to increase rapidly. According to Counterpoint's forecast, in 2023-2027, the compound growth rate of advanced AIPC sales will be about 115%, and the compound growth rate of AI mobile phone sales will be about 32%.

Aspect of the target

The AI section on the end side suggests focusing on Rockchip (603893.SH), Hengxuan Technology (688608.SH), Lexin Technology (688018.SH), Zhongke Lanxun (), and Horizon Robot-W (09660); the ASIC section suggests focusing on VeriSilicon (Dubai), Aogee-U (), and Cambrian-U (). 688332.SH 688521.SH 688220.SH 688256.SH

Risk Alerts

Consumer electronics recovery falls short of expectations; demand for AI terminals falls short of expectations; ASIC demand falls short of expectations; AI technology iteration falls short of expectations; AI application development falls short of expectations; and capital expenditure of major AI companies and Internet companies falls short of expectations.