China AI: Starting Point, Verification Point, Acceleration Point

Core Viewpoint

On February 20, 2025, Alibaba announced its FY2025Q3 quarterly results, with capital expenditures of 31.8 billion yuan, a year-on-year increase of +259% and a quarter-on-quarter increase of +82%; the total capital expenditure for the first three quarters of FY2025 reached 61.4 billion yuan, a year-on-year increase of +193%. Strong capital expenditures serve as a powerful testament to China's AI entering a new phase. China’s AI has bid farewell to the "mapping" market that benchmarks against US stocks, and the entire AI industry chain is expected to welcome a value reassessment. We believe that three major factors—large enterprises' CapEx, commercialization of AI applications, and iteration of model capabilities—are likely to drive the continuous acceleration of China's AI industry, and we are optimistic about the value reassessment and performance realization of Chinese AI assets in 2025.

Starting Point: DeepSeek Reshapes the Value System of China's AI Industry through Model Equality

At the end of 2022, ChatGPT emerged, ushering the global technology industry into the era of large models. From 2023 to 2024, the global AI industry is dominated by American tech companies, while China's AI has not shown unique advantages in computing power, models, and applications. In early 2025, the domestic large model DeepSeek achieved a breakthrough on the model side, rapidly advancing the reconstruction of China's AI industry value with its advantages of being open-source, commercially viable, and low-cost. 1) Computing Power: Capital expenditures of tech companies are expected to enter a new investment cycle; 2) Models: Model capability iterations are expected to occur at a higher frequency; 3) Applications: The productization and commercialization of AI applications are expected to accelerate. We believe that DeepSeek's model equality is not only a single-point change on the model side but also the starting point of the entire AI industry transformation, which is expected to leverage model changes to unlock greater industrial value in China's AI.

Validation Point: Alibaba's CapEx Expands Significantly, AI Acceleration is Strongly Verified

Alibaba's FY25Q3 cloud business revenue was 31.7 billion yuan, a year-on-year increase of +13%, showing significant acceleration (the year-on-year growth rates for cloud revenue in FY25Q1 and Q2 were 6% and 7%, respectively), mainly benefiting from the growth in AI demand, with reasoning demand accounting for 60%-70% of the new demand. The CapEx for FY25Q3 was approximately 31.8 billion yuan, a year-on-year increase of +259% and a quarter-on-quarter increase of +82%; the total CapEx for the first three quarters of FY25 was approximately 61.4 billion yuan, a year-on-year increase of +193%. The substantial expansion of capital expenditures verifies that China's AI has entered a new upward cycle. According to Alibaba's conference call, in the next three years, Alibaba's CapEx investment may exceed the total of the past ten years (from CY2015 to CY2024, the company's non-real estate CapEx totaled over 320 billion yuan). We believe that subsequent CapEx from ByteDance, Tencent, Xiaomi, and other tech companies is also expected to expand simultaneously, and the demand for model hosting and application reasoning will accelerate after model equality is achieved.

Acceleration Point: Focus on Three Catalysts: Large Enterprises' CapEx, Commercialization of AI Applications, and Iteration of Model Capabilities

1. Large Enterprises CapEx: The CapEx investment of major American technology companies is an important indicator of the progress in the AI industry. Based on internal/external and training/inference classifications, the computing power demand of large enterprises can be divided into four major scenarios, fully reflecting the maturity of the AI industry. Currently, Alibaba has taken the lead in initiating a new round of CapEx investment cycle, and it is recommended to pay attention to the subsequent CapEx situations of other large enterprises; 2) AI Application Commercialization: AI applications are the final link in the commercial closed loop of large models, and the ability to achieve large-scale implementation of AI applications is crucial. It is recommended to focus on the application progress in the 2B and 2C scenarios; 3) Model Iteration: Driven by DeepSeek, global large models are accelerating iteration, with breakthroughs in model capability limits and upgrades in multimodal model capabilities still being key areas of industry focus.

AI Industry Chain Company Overview

Hardware Segment: 1) GPU; 2) Servers; 3) Power Supply; 4) Cooling; 5) PCB; 6) Optical Modules; 7) IDC, for specific company names, please refer to the original research report.

Software Segment: 1) 2B Applications; 2) 2C Applications; 3) Data Middle Platform, for specific company names, please refer to the original research report.

Authors of this article: Xie Chunsheng S0570519080006, Yue Boxiong S0570524080004, Lin Hailiang S0570524070001, Source: Huatai RuiSi, Original Title: "China AI: Starting Point, Verification Point, Acceleration Point"

Risk Warning and Disclaimer

The market has risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk