计算机行业周报：板块继续下跌，Google发布TurboQuant-20260330

Investment Rating - The report maintains a "Recommended" rating for the industry [5] Core Insights - The computer (Shenwan) index fell by 3.44% during the week of March 23-27, 2026, underperforming major indices due to market conditions. The Shanghai Composite Index decreased by 1.09%, the Shenzhen Component Index by 0.76%, and the ChiNext Index by 1.68% [1][10][19] - Google Research introduced TurboQuant, a vector quantization compression algorithm that reduces the key memory usage of large models to 3 bits without losing accuracy. This innovation addresses the significant memory consumption of KV cache during model inference, which expands linearly with context window growth [3][20] - The report suggests that the continuous rapid iteration of large model-related software and hardware technologies is expected to lower application costs, promoting the widespread adoption of AI applications and driving sustained growth in AI hardware demand [3][20] Summary by Sections Market Review - The Shenwan computer index experienced a decline of 3.44%, with sub-sectors such as computer equipment, IT services II, and software development showing respective declines of -3.12%, -3.14%, and -3.90% [1][10][12] Major Events - Significant announcements include: 1. Guanglian Da reported a revenue of 6.068 billion yuan for 2025, a decrease of 2.17% year-on-year, while net profit increased by 61.77% to 405 million yuan [2][17] 2. Zhiming Da's revenue grew by 61.87% to 709 million yuan, with net profit soaring by 425.27% to 102 million yuan [2][17] 3. Kingsoft Office achieved a revenue of 5.929 billion yuan, up 15.78%, and a net profit of 1.836 billion yuan, an increase of 11.63% [2][17] 4. Hengsheng Electronics reported a revenue of 5.783 billion yuan, down 12.13%, but net profit rose by 18.01% to 1.231 billion yuan [2][17] Investment Perspective - The report highlights that the computer index's significant drop is influenced by international factors, with the introduction of TurboQuant potentially impacting the storage sector's stock prices due to its focus on reducing the key cost pain point of large models [3][19][20]