兴证策略：四大指标看TMT当前位置与内部轮动

Core Viewpoint - The current AI sector shows a moderate level of congestion, with no significant excess returns accumulated in the TMT sector relative to the overall A-share market [1][5][6]. Group 1: Congestion Level - Most AI sub-sectors are currently at a moderate congestion level, with low congestion levels observed in upstream computing (IDC, high-speed copper connections, AI chips, GPUs), midstream software services (office software, cloud computing, SaaS, cybersecurity, AI Agent), and downstream applications (fintech, gaming, humanoid robots, smart healthcare) [2]. Group 2: Rolling Return Differential - The rolling 40-day return differential between TMT and the overall A-share market remains around 5%, indicating that it has not reached the overheating threshold of 10% [5]. Group 3: Trading Volume Proportion - The trading volume proportion for TMT is currently around 30%, which is still below the historical overheating range of 40%-45%, indicating no overheating signals [6]. Group 4: Internal Rotation of the Industry Chain - The recent AI market rally since June has been primarily driven by the upstream North American computing chain, with two notable divergences: upstream hardware outperforming midstream software services and downstream applications, and within upstream hardware, components like optical modules and PCBs significantly outperforming domestic chip chains [9][11]. Group 5: Performance of Sub-sectors - Since June, the North American computing chain, represented by PCBs and optical modules, has accumulated substantial excess returns compared to other directions in the industry chain. Since August, the focus of the upstream computing market has shifted towards lower-tier areas such as liquid cooling, copper connections, optical fibers, and domestic chips, with significant gains observed in sectors like robotics, AI smartphones, electronic governance, and gaming [11][13].