Investment Rating - The investment rating for the industry is "Outperform the Market" (maintained) [1] Core Insights - The recovery of sales for NVIDIA's H20 chip is expected to clear inventory and boost capital expenditure (Capex) among domestic cloud providers, leading to increased demand in the cloud sector [3][4] - The introduction of the H20 chip, which is a compliant version tailored for the Chinese market, is anticipated to significantly reduce the time and costs associated with hardware investment and actual usage, accelerating the transition of domestic enterprises to cloud services [4][5] Summary by Relevant Sections AI Chip Computing Layer - The resumption of H20 sales will lead to inventory clearance and increased capital expenditure, with the potential for continued positive Capex growth if the supply of special GPUs proceeds smoothly [3][5] - The H20 chip, while having reduced performance compared to the H100, remains a primary choice for domestic AI developers due to its compatibility with the CUDA ecosystem [5][6] - The expected inventory loss related to H20 is projected to reach around $4.5 billion, with total inventory recovery potentially exceeding $10 billion [6] Domestic Cloud Providers - Domestic cloud providers maintained high growth rates in Q2, with expectations for accelerated growth in the second half of the year as customer solutions are implemented [9][10] - The recovery of H20 supply is expected to significantly increase capital expenditure among domestic cloud providers, positively impacting downstream demand [9][10] - Despite a slight decrease in the number of AI customers in Q2, the penetration rate of scenarios continues to improve, indicating a robust growth trajectory for cloud services [9] AI Model Layer - The second quarter saw a concentration of model development among leading domestic tech firms, which are rapidly catching up to international standards and fostering an open-source ecosystem [13][17] - Major companies like Alibaba, ByteDance, Tencent, and Baidu are actively enhancing their AI models and tools, contributing to a vibrant AI ecosystem [15][16][17]

Group 1 - The AI Agent sector is expected to be one of the hottest areas in AI by 2025, with major companies like Microsoft, Google, OpenAI, Baidu, Alibaba, and Tencent launching significant AI Agent products [1] - The current wave of AI entrepreneurship is characterized by lightweight teams, often consisting of just a few individuals, leveraging large model capabilities to initiate AI application projects [1][2] - Cloud computing companies such as Baidu Intelligent Cloud, Tencent Cloud, and Volcano Engine are actively positioning themselves in the AI Agent space, recognizing the business opportunity [1] Group 2 - Volcano Engine has introduced its latest model, Doubao 1.6, along with a full-stack AI Agent development tool, significantly reducing model prices to one-third of DeepSeek's [1][2] - The trend of small team entrepreneurship in AI applications is leading to a shift in investment strategies, with a focus on early-stage projects requiring lower capital [2] - The rapid development of large model technology and significant cost reductions are driving the explosion of AI application entrepreneurship, reminiscent of the internet boom from 2004 to 2006 [2] Group 3 - The competition among major players in the large model sector is intensifying, with companies like Baidu, Alibaba, Tencent, and Volcano Engine launching new models in the first half of the year [3] - The pricing and capabilities of large models are critical factors influencing companies' choices of AI Agent development platforms [4][5] - Baidu's Wenxin model 4.5 Turbo and 1 Turbo have significantly reduced token usage costs, making them more attractive to users [4][5] Group 4 - Major cloud vendors are enhancing their AI Agent development platforms to lower technical barriers, enabling non-technical product managers and entrepreneurs to quickly develop applications [6] - The demand for AI Agents is expected to reshape the cloud computing industry, with a focus on distributed architecture and market concentration [7] - The growth of AI-related businesses is driving revenue increases for cloud computing companies, with Baidu's AI cloud business recording significant growth in Q1 2025 [7]

Investment Rating - The industry investment rating is "Strongly Outperform the Market" with an expected industry index increase of over 5% in the next six months [29] Core Insights - The Hong Kong stock market showed a positive trend with the Hang Seng Index closing at 21,980.74 points, up approximately 2.74% during the week of April 21 to April 25, 2025 [11] - Significant net buying from southbound funds was observed, with Alibaba, Meituan, and Tencent being the top three stocks attracting investment [20] - The report highlights the low valuation of internet companies, with projected P/E ratios for 2025 indicating potential upside for stocks like Tencent, Pinduoduo, and Alibaba [4][19] Summary by Sections Internet Sector - Tencent's 2025 projected P/E is 16X, with strong fundamentals and new product launches expected [4] - Pinduoduo is recommended for attention after a recent pullback, currently valued at 8X P/E [4] - Alibaba's valuation is at 12X P/E for FY26, with expectations of continued growth in cloud services and e-commerce market stabilization [4] AI Sector - Baidu's AI Developer Conference introduced the Wenxin model 4.5Turbo, with input costs reduced by 80% compared to its predecessor [2][8] - The report anticipates continued growth in the intelligent driving sector, driven by major manufacturers and advancements in technology [2] New Consumption - Pop Mart reported a total revenue increase of 165%-170% year-on-year for Q1 2025, exceeding market expectations [5] - The introduction of new products is expected to maintain strong demand and upward trends for Pop Mart [5] New Listings - New companies like Hu Shang A Yi and Green Tea Restaurant are under observation following their approval for listing on the Hong Kong Stock Exchange [3]

Group 1 - Nomura's report indicates that Baidu (BIDU.O) recently launched two upgraded versions of its LLM models, 4.5Turbo and X1Turbo, focusing on cost reduction and multi-model performance [1] - Baidu expects the average cost of LLMs to continue decreasing, with the capabilities of top-tier models like 4.5Turbo being sufficient to meet many user needs, laying a solid foundation for the rise of AI applications [1] - The report highlights Baidu's exploration of AI application potential through various self-developed AI applications, including the AI-driven programming tool Miaoda, which allows coding using natural language [1] Group 2 - The most notable application mentioned is the Robotaxi, which utilizes a significant amount of Baidu's AI technology [1] - Despite geopolitical risks, Baidu remains optimistic about the future of AI [1] - Nomura maintains a target price of $98 for Baidu's US stock and a "neutral" rating [1]