Galaxy Securities: Optimistic about the rebound at the bottom of AI applications and focusing on left-hand layout opportunities

The Zhitong Finance App learned that Galaxy Securities released a research report saying that the artificial intelligence sector maintained a volatile trend in December, falling slightly, and trading volume shrank. The overall market showed some differences, dominated by short-term subject matter trading opportunities. The bank believes that the market's current expectations for the Bank of Japan's interest rate hike and the Federal Reserve's interest rate cut have been fully interpreted. Currently, the capital is still relatively concentrated in fields such as AI hardware, optical modules, PCBs, etc., where performance is expected to continue to be realized. Of course, the technology sector shows a fragmented structure within the technology sector. The line also sees the section There was a marginal improvement in AI application companies' performance in the third quarter. Currently, the bank still suggests that the AI sector be placed on the left to support performance in segments and leading companies.

Galaxy Securities's main views are as follows:

The Nvidia H200 was approved for sale in China, and the domestic computing power industry chain faced a moment of change

On December 8, local time in the US, Trump said he would allow Nvidia to sell H200 chips to “approved customers” in China, and 25% of sales revenue would be handed over to the US government. The bank believes that in the short term, domestic computing power chips will be impacted to a certain extent. Of course, the exact amount of quota each major manufacturer can get is still unknown. Purchasing a certain amount of H200 will help domestic models improve the efficiency of the training process. In the long run, the impact on domestic computing power chips is limited and even forced domestic computing power chips to accelerate breakthroughs. The bank is still firmly optimistic about the domestic computing power industry chain. The long-term logic will be to suspend the replacement rhythm without changing the autonomous and controllable route on the training side, accelerate the implementation of large-scale domestic models and ecological applications through short-term H200 support, and promote the closed-loop commercial logic of the industrial chain. Domestic chips will continue to penetrate the reasoning side and industry scenarios. Bridging the gap in single-card performance through “supernodes” is beneficial to domestic computing power chips and ecological development in the long run.

Meta spent billions to acquire Manus, highlighting the commercial value of AI applications

On December 30, Beijing time, Manus released news that Manus will join Meta to begin a new round of exploration. Meta may spend several billion dollars to acquire Manus this time. This will be the third largest merger and acquisition since the establishment of the company after the company's establishment, after WhatsApp and Scale AI previously acquired. As a general agent, Manus can independently perform complex tasks such as research, coding, and data analysis. Currently, it has processed 147 trillion tokens and created more than 80 million virtual computers to provide intelligent services for millions of individuals and small and medium-sized enterprises around the world. The bank believes that Meta's acquisition of Manus is an important part of further strengthening its AI capabilities and accelerating the integration of AI technology into consumer-grade and enterprise-level product matrices. By integrating Manus into Meta's products, it can directly commercialize sales and provide Meta with a software layer with high profit margins. The bank believes that this acquisition will accelerate the commercialization of AI applications. 2026 will be a major year for AI applications, and it is recommended to continue to pay attention to investment opportunities in AI applications.

Investment advice: Focus on the following racetrack segments and companies

1. Domestic computing power industry chain: IFC, Zhongke Shuguang, Haiguang Information, etc.; 2. IDC service providers and computing power leasing: Runze Technology, Guanghuan New Network, Daowei Technology, Chengdi Xiangjiang, etc.; 3. Domestic credit innovation manufacturers: China Software, Softcom Power, Dameng Data, etc.; 4. AIAgent and applications: Jinshan Office, Huaxun, Daotong Technology, Tonghuashun, Jiahe Meikang, Guoneng Nissin, Shenxun, Caiheng Bioelectronics, Wanxing Technology, etc.; 5. Cloud computing vendors: Kingdee International, Jinshan Cloud; 6. All-in-one computers and end-side AI: ArcSoft Technology, Hikvision, Zhongke Chuangda, Huaqin Technology, Fluorite Network, etc.; 7. Supply, distribution, and application companies in the data element industry chain: Shenzhen Sanda A, Shanghai Steel Union, etc.; 8. EDA: Huada Jiutian, Guilun Electronics, Guangliwei, etc.

Risk warning: Technology iteration falls short of expectations; competition from tech giants increases risk; legal and regulatory risk; supply chain risk; downstream demand falls short of expected risk