China Securities Co., Ltd.: AI computing power is a sector-wide market opportunity, with individual stocks in the industry chain generally having the opportunity for growth.

China Securities Co., Ltd. released a research report stating that currently, the penetration rate of AI is still low, the development of large models is still in the early stages, the industrial application cycle is just beginning, and the investment in computing power brought by large models is on the rise. Capital expenditure will increase with the growth of large model income, and there is a high ceiling for investment. The AI computing power sector involves advanced processes, GPU/ASIC, optical modules, PCBs, servers, switches, optical devices, copper connections, IDCs and their accessories (liquid cooling, power supply, power, comprehensive wiring, etc.), and is a sector-wide bull market, with opportunities for companies in the industry chain, and market value still has room for growth. The main points of China Securities Co., Ltd. are as follows: - The penetration rate of AI is still low, and the development of large models is still in the early stages - Compared with the user situation during the mobile internet cycle, the current penetration rate of large models is still low. For example, in July 2025, the number of active users of ChatGPT was 700 million, accounting for only 12.7% of the global internet population. In November 2024, the cumulative users of Douyin exceeded 160 million, with a penetration rate of less than 15% among mobile internet users in China. Currently, large models may still be in the early stages of "user cultivation + application exploration", and there is still a lot of room for growth in the number of users and penetration rate in the future. - Large models may ignite the industrial internet, upgrading the mobile internet, and the future outlook is optimistic - The company believes that AI large models may integrate and upgrade mobile internet applications, but are more likely to be widely used in the B-end, igniting the industrial internet. For example, in July 2025, Microsoft revealed that AI tools in their customer service center saved over 500 million US dollars in costs in 2024, and 35% of the new products at Microsoft were generated by AI. According to Gartner, in 2024, global IT spending on IT services and software was $ 2.71 trillion, and in China, it was $156.5 billion. If large models can penetrate this market in the future, the development potential is enormous. - The long-term space for capital expenditure driven by AI is expansive and sustainable - In the mobile internet cycle, capex shifted from network investment to computing power (cloud/edge computing) investment. The significant growth of CSP cloud business drove capex soaring, with comparable growth multiples, such as Amazon's cloud computing revenue increased by 24.8 times from 2013 to 2022, and capex increased by 17.5 times; Alibaba's cloud computing revenue increased by 18.5 times from 2014 to 2018, and capex increased by 9.2 times. From the perspective of operating cash flow, capex has not reached the company's ceiling. From 2013 to 2022, Amazon, Google, and Meta's annual capex highest proportion of cash flow from operating activities was 136%, 52%, and 63%, respectively, while in 2024, it was 72%, 42%, and 43%, respectively. In the future, the capital expenditure of major companies focusing on intelligent computing is expected to continue to grow with the growth of AI-related income.