Kuaishou (01024) introduces "multi-image reference" mode with AI technology, Zhongjin optimistic about its AI competitiveness.

Recently, Kuaishou's Keling AI officially launched the "multi-image reference" mode globally, further consolidating Kuaishou's leading position in the field of AI video generation. At the same time, CICC pointed out in its latest research report that as AI enters the application exploration stage, domestic internet companies exploring applications are more worth looking forward to. It is recommended to focus on companies such as Tencent and Kuaishou. Currently, the direction of AI applications includes dialogue Siasun Robot&Automation, programming assistants, office efficiency tools, education, entertainment, and empowering traditional businesses. Looking ahead to 2025, CICC believes that AI Agent may become a focus of competition. Compared to the model side, which is more in a catch-up state, Chinese internet companies have strong competitiveness in the application side. Taking Kuaishou as an example, since the launch of Keling AI, it has undergone more than 20 updates and upgrades in just over half a year, surpassing Sora in terms of iteration speed and usability. The new "multi-image reference" mode can understand and integrate different subjects in multiple images, creatively generate blended videos based on user text descriptions, further overcoming the consistency challenges in AI video generation. Previously, Keling AI had launched the "face model" feature globally, supporting users to train and customize face models by uploading multiple video clips, and continue to generate video content based on that face model. As the industry's first video model customization feature, the "face model" innovatively meets the user's need to create multiple shots containing the same person. With the introduction of the "multi-image reference" feature, Keling AI further solves the consistency challenges in video generation, helping users unleash their creativity, freely combine various image elements, and greatly improve the accuracy and usability of video generation. CICC analysis believes that whether the supply dividend can bring about demand resonance depends on whether the product effect can meet user needs, whether costs can be controlled, whether the business model is sustainable, etc. Compared to the model side, which is more in a catch-up state, Chinese internet companies have shown strong competitiveness in the direction of AI applications compared to Chinese and American internet companies. Currently, the valuation of Chinese internet companies is basically not inclusive of AI expectations. In terms of dynamic P/E ratio system, domestic internet companies such as Tencent and Kuaishou have PEG (dynamic P/E ratio) ratios of 0.96 times and 0.35 times based on this year's P/E ratio/this year's profit growth rate, respectively, while Meta and Alphabet have ratios of 2.38 times and 1.93 times, respectively. CICC believes that as training methodologies become more mature, the gap in computing power becomes smaller, and talent mobility increases, Chinese AI-related companies are expected to gradually narrow the gap with leaders and lead the way. It is recommended to focus on companies such as Tencent and Kuaishou.