Alibaba Cloud's Zhang Yong: Demand for AI Cloud Services is Very Strong, Incremental Opportunities are Just Starting to Unfold


Tongyi Yunqi (Hangzhou) Information Technology Co., Ltd., a company related to Tongyi Qianwen, recently completed a business registration change, with its registered capital increasing from 100 million yuan to 200 million yuan, a 100% increase. At the same time, Hangzhou Tongyi Lab Technology Co., Ltd. was added as a shareholder, co-owning shares with Alibaba Cloud Computing Co., Ltd. The company was established in November 2018, with Yu Wenyuan as its legal representative. Its business scope includes advertising production, wholesale of computer software and hardware, and sales of electronic products.
According to Alibaba's Q4 2025 financial report, the group's revenue reached 284.843 billion yuan, with a year-over-year growth of 9% after excluding disposed businesses, demonstrating strong operational resilience. Group CEO Wu Yongming proposed a 'strategic military order,' outlining a plan to vigorously promote the commercialization of cloud and AI in the next five years, including MaaS (Model as a Service) business.
Alibaba Cloud launches an extremely cost-effective AI programming subscription plan, lowering the entry barrier for developers through a bundled model and accelerating the implementation of the AGI strategy. The service integrates four leading domestic AI models, breaking the limitations of single models, and competing in the market through downward disruption.
Surging global demand for AI has driven changes in China's cloud computing market. Alibaba Cloud's market share increased from 34% to 36% in Q3 2025, growing for three consecutive quarters, with AI serving as its core growth driver.
Alibaba Cloud launched the new image generation foundation model Qwen-Image-2.0 on February 10, 2026, which integrates image generation and editing capabilities. The model adopts a 7B lightweight architecture, maintaining fast inference while offering four core advantages: professional text rendering capability, supporting ultra-long and complex instructions up to 1k tokens, and demonstrating excellent performance in multiple blind testing benchmarks.