通信：微信接入DeepSeek，推理算力供需矛盾加剧

Investment Rating - Industry rating: Overweight [1] Core Insights - The integration of WeChat with DeepSeek is expected to significantly enhance the demand for inference computing power, creating a supply-demand imbalance in the short term [3][5] - The AI investment narrative has evolved from focusing on large model development to addressing the bottlenecks in training computing power, leading to increased interest in AI hardware and applications [7] - The report anticipates a shift back to the computing power sector, particularly in inference computing, as the demand gap becomes more pronounced [8] Summary by Sections Industry Overview - WeChat is currently testing the integration of the DeepSeek R1 model, which allows users to access AI search capabilities without compromising personal data [3] - The integration is seen as a strategic move by Tencent to enhance its ecosystem and leverage its vast user base of 1.3 billion monthly active users [3] Technological Developments - Innovations in algorithms have reduced the unit demand for computing power, but the overall demand for computing power is expected to increase due to lower entry barriers for companies entering the AI space [4] - Predictions indicate that inference computing power will account for 95% of total computing power in the future, with a compound annual growth rate of 190% over the next five years in China [4] Market Dynamics - The current demand for inference computing power is surging, leading to challenges for internet companies that have integrated DeepSeek, as they face server overload issues [5] - The report highlights that cloud service providers need to invest in high-performance GPUs and upgrade their data center architectures to meet the growing demand [5] Investment Recommendations - The report identifies several companies that are likely to benefit from the developments in AI computing power, including those in AI chip manufacturing, server technology, and optical communication [8]