CMSC: AI technology progress drives terminal prosperity, AIDC demand potential is large

CMSC released a research report stating that, recently, due to the tight server resources, DeepSeek officially announced the suspension of API service recharge. The current tight server resources of DeepSeek are the result of a short-term imbalance between supply and demand. In the long term, the improvement in AI efficiency will inevitably drive a more widespread application at the terminal. The cycle of Jevons Paradox, where "efficiency improvement cost reduction application popularization demand surge resource tension," may be re-enacted in the AI industry. In addition, low-cost models represented by DeepSeek are expected to enhance the adaptation of the domestic computing power industry chain and promote the accelerated expansion of domestic computing power platforms. Failure to expand in a timely manner is the main reason for the tight server resources. This time, due to the upcoming end of preferential pricing (until February 8, 2025), the concentrated recharge by users further intensified the server load. DeepSeek significantly reduced training and inference costs through algorithm optimization (such as hybrid expert models, reinforcement learning, etc.), but the hardware expansion failed to match the sharp increase in the number of users, leading to tight server resources and the suspension of API service recharge. Overseas AI-related capital expenditures remain strong. Recently, overseas CSP companies such as Meta and Amazon have successively released their financial reports for 2024, with capital expenditure in Q4 exceeding market expectations. In the 2025 fiscal year, the capital expenditure plans of these companies all show a significant increase, with the main investment direction being AI-related infrastructure construction. In addition, Trump has also announced The Stargate Project, planning to invest $500 billion in AI infrastructure in the United States. Apart from the United States, France also plans to invest 109 billion in artificial intelligence projects in the coming years. Overall, there is a strong determination and clear direction in overseas government and mainstream enterprises' investment in AI-related projects. Risk warning: AI terminal demand falls short of expectations, enterprise capital expenditures fall short of expectations, escalation of trade barriers, etc.