一封致歉信，AI大模型龙头没了700亿港元

Core Viewpoint - The Hong Kong stock market showed a strong rebound in the morning session, but the AI large model sector experienced a notable decline, with significant differentiation among stocks [1] Group 1: Market Performance - As of the midday break, Zhiyu (02513.HK) fell by 23%, Haizhi Technology Group dropped by 18%, and MINIMAX-WP (00100.HK) decreased by 10%, contrasting sharply with the overall rise in the tech sector [1] - Zhiyu's market capitalization evaporated by over 700 million HKD during the session [1] Group 2: Company Issues - The recent adjustment was directly linked to Zhiyu's apology letter regarding the GLM Coding Plan revision, where the company acknowledged three main issues: insufficient rule transparency, slow rollout of GLM-5, and poorly designed upgrade mechanisms for existing users [3] - To restore user experience and market confidence, Zhiyu quickly introduced full refunds, one-click rollbacks, and service extensions as compensation measures [3] Group 3: Stock Performance and Ratings - Prior to the recent decline, Zhiyu's stock price surged significantly, closing at 725 HKD on February 20, with a total market value of 323.2 billion HKD, reflecting a cumulative increase of over 500% since its IPO [3] - Institutional ratings for the sector remained positive, with CICC rating Zhiyu as "outperforming the industry" and setting a target price of 688 HKD, while JPMorgan initiated coverage with a "buy/overweight" rating and a target price of 400 HKD for Zhiyu and 700 HKD for MINIMAX [3] Group 4: Short-term Outlook - In the short term, the unexpected traffic for GLM-5 led to capacity issues, with Pro/Lite versions facing throttling and refund pressures, compounded by profit-taking from previous gains [4] - Institutions generally view the current turmoil as an operational issue that does not alter the core technology and commercialization logic [4] - The market remains focused on technological implementation, user retention, and profit realization, with investors advised to monitor computing power progress, commercialization pace, and the effectiveness of compensation measures [4]