通义万相 Wan2.5 Preview
Search documents
阿里巴巴-W(09988):定位全栈人工智能服务商,加大AI基础设施建设投入
Guoxin Securities· 2025-09-27 15:25
Investment Rating - The investment rating for the company is "Outperform the Market" (maintained) [2][4][18] Core Views - The company positions itself as a full-stack AI service provider, significantly increasing its investment in AI infrastructure [3][5][6] - The CEO of the company highlighted three stages of AI development: intelligent emergence, autonomous action, and self-iteration, predicting a future shift towards super artificial intelligence (ASI) [3][5] - The company plans to enhance its global data center energy consumption by tenfold by 2032, indicating sustained capital expenditure (CAPEX) investments [4][6] - The company has launched several new models, including Qwen3-Max, which ranks third globally in performance, and introduced the next-generation Qwen model architecture [3][11][12] - The company aims to accelerate its international expansion by establishing data centers in Brazil, France, and the Netherlands, while also expanding operations in five other countries [3][4] Summary by Sections AI Infrastructure and Models - The company is committed to providing leading large models and cloud computing networks, with a focus on self-developed core storage systems and computing chips [4][6][13] - The company has upgraded its AI infrastructure, including high-performance servers and networks, to meet the demands of large model training and inference [13][14] Financial Projections - The revenue forecasts for FY2026, FY2027, and FY2028 are adjusted to 1,066.1 billion, 1,188.8 billion, and 1,301.4 billion respectively, reflecting a more optimistic outlook on cloud revenue growth [4][18] - Adjusted net profits for the same periods are projected at 126.7 billion, 167.2 billion, and 196.0 billion [4][18] Market Position and Strategy - The company’s full-stack self-developed capabilities and increased overseas presence are expected to improve profit margins in the long term [4][18] - The company’s strategy includes a focus on open-source models, aiming to create significant value in the LLM era [5][6]