صافي الأرباح للربع المالي الأخير لشركة علي بابا يتراجع بنسبة تقارب 70% على أساس سنوي، هل يمكن لإعادة تشكيل خريطة الذكاء الاصطناعي أن تدعم هدف المئة مليار دولار خلال خمس سنوات؟| تفسير التقرير المالي

Cailianshe March 19 News (Reporter Fu Jing) After going through a series of events that have sparked industry-wide discussion since the beginning of the year—such as the AI Spring Festival marketing battle, the resignation of Lin Junyang, the former technical head of the Qianwen team, the establishment of the Token Hub business group, the release of “Wukong,” an enterprise-level lobster Agent, and Alibaba Cloud’s price increases—today, after the market close, Alibaba released its latest quarterly financial report.

According to the financial data, Alibaba’s third quarter of fiscal year 2026 (for the quarter ended December 2025) achieved revenue of RMB 284.84 billion, up 2% year over year. Net profit attributable to ordinary shareholders was RMB 16.32 billion, down 67% year over year. Non-GAAP net profit was RMB 16.71 billion, also down 67% year over year.

In addition, net cash flow from operating activities in the quarter was RMB 36.03 billion, down 49% year over year. Free cash flow was RMB 11.35 billion, down 71% year over year. For the nine months ended December 31, 2025, free cash flow was RMB -29.31 billion, compared with RMB 70.13 billion in the same period last year. It is said that the decline in free cash flow is mainly attributable to investments in instant retail.

Cloud business brings improvement in net profit Yesterday announced a price increase

Looking at each business segment in detail, in the quarter, instant retail revenue increased 56% year over year, making it the business with the largest year-over-year revenue growth among the company’s many segments. Cloud Intelligence Group revenue was RMB 43.28 billion, up 36% year over year. Meanwhile, “All Other” revenue—including Hema, Cainiao, Alibaba Health, Tiger Whale Entertainment Group, Amap, Qianwen C-end business group, Lingxi Game, DingTalk, and other businesses—declined 25% year over year, mainly due to revenue declines resulting from the disposal of Xinxi Retail and Yintai business, as well as Cainiao’s revenue decline.

Alibaba said that the overall decline in profitability was mainly attributable to investments in instant retail, user experience, and technology, while the growth of the cloud business also brought improvements. The Cloud Intelligence Group’s revenue growth was mainly driven by the growth of public cloud revenue, including an increase in the adoption of AI-related products.

Just yesterday, Alibaba Cloud’s official website published an announcement stating that, due to the global AI demand surge and supply-chain price increases, Alibaba Cloud’s AI computing power, storage, and other products saw the highest price increases of up to 34%. Among them, services related to compute cards such as the “Pingtouge Zhenwu 810E” saw increases of 5%-34%, and the file storage product CPFS (Zhicsuan version) increased by 30%.

A Cailianshe reporter, upon checking Alibaba Cloud’s official website, noted that the GPU-related services subject to this price adjustment include cloud servers ECS, Intelligent Computing Lingjun, the artificial intelligence platform PAI, the large-model service platform Bailian, and more. According to Alibaba, it has begun mass production of its self-developed Pingtouge GPUs, supporting end-to-end AI workloads spanning training, fine-tuning, and inference.

Earlier this year, when multiple industry insiders interviewed with Cailianshe reporters, they analyzed that, considering factors such as rising upstream compute costs, growth in market demand, relatively low unit prices of Token (the smallest semantic unit of text) at the current stage, and intense market competition, domestic cloud vendors are generally under pressure and are expected to see a wave of phased price increases in the cloud services sector. However, as of now, price-increase expectations in the cloud computing field are only present in AI-related areas.

Reality is consistent with the above judgment. Not only Alibaba Cloud announced price increases. Yesterday, Baidu Intelligent Cloud also released an announcement: AI compute-related products were raised by about 5%-30%, and parallel file storage and similar products increased by about 30%.

Last week, Tencent Cloud officially announced price increases for its AI models. First, GLM 5, MiniMax 2.5, and Kimi 2.5 ended free public testing on March 13 and were converted into正式 commercial services, with charges based on model calls. Second, the Hunyuan series models Tencent HY2.0 Instruct and Tencent HY2.0 Think increased in price. Regardless of input or output, the price increases are generally more than 4 times.

Rapidly rebuild the AI landscape Set a five-year business target

In its latest financial report, Alibaba CEO Wu Yongming mentioned that in this quarter, Alibaba continued to invest in two core areas: AI and consumer. AI is one of the main growth engines. The Qianwen App and Alibaba’s consumer ecosystem are accelerating integration across application scenarios, significantly driving new user growth and higher transaction activity.

At tonight’s earnings call, Wu Yongming said that, in the face of the long-term AI market’s growth momentum, Alibaba Group’s commercial goals for its AI strategy are very clear: in the next five years, cloud and AI monetization revenue including MaaS will exceed 100 billion USD.

At present, Alibaba is rapidly reshaping its own AI blueprint.

On March 16, Alibaba established the Token Hub business group, including Tongyi Lab, the MaaS business line, the Qianwen division, the Wukong division, and the AI Innovation division. It covers a complete layout from foundational model research and development, to model service platforms, to AI applications for both individual and enterprise users. As a result, Alibaba became a global technology giant organized around “Token”—the most basic unit of measurement for processing text, code, and images with large language models.

Recently in the industry, “Token” has drawn significant discussion. In this week’s GTC2026, during Huang Renxun’s keynote speech, the term “Token” reportedly appeared more than 70 times in total. The Sullian report shows that in the Chinese market, the average daily Token consumption surged from 100 billion at the beginning of 2024 to an astonishing 30 trillion by the end of June 2025. And in the second half of 2025, China’s enterprise-level large models had already reached an average daily usage of 37.0 trillion tokens.

Media reports citing sources close to Alibaba Cloud said that another important reason Alibaba Cloud announced the price increase yesterday is the “surge in Token calling volume.” Alibaba Cloud is channeling scarce AI compute resources toward the Token business.

One noteworthy point in this adjustment is the first exposure of the mysterious “Wukong division,” positioned to build a B-end AI-native work platform. In other words, besides the C-end Qianwen App, Alibaba also plans to deeply integrate large-model capabilities into enterprise work flows, with a key focus on the B-end.

Wu Yongming also said in the financial report that in the future, leveraging full-stack AI capabilities of “large models + cloud + chips,” along with comprehensive integration with Alibaba’s business ecosystem, Alibaba will continue to push both AI to B and AI to C.

A Cailianshe reporter noted that the “earthquake” in the AI field earlier this month was triggered by the resignation turmoil involving Lin Junyang, the “technical soul figure” behind the Qianwen large model. At that time, right after the “host the guests for the Spring Festival” marketing battle had just come to a close, the Qianwen App arguably fought a great battle. QuestMobile data shows that during the Spring Festival, three AI applications—Doubao, Qianwen, and Yuanbao—hit new highs for DAU. Their peak values were 145 million, 73.52 million, and 40.54 million respectively, with Qianwen achieving the highest increase of 940%.

According to media reports citing messages from people close to Alibaba Cloud, the flagship model Qwen3.5 performed worse than expected. On the overseas well-known large-model “blind test arena” LMArena, as of March 3, 2026, in the overall ranking, Qwen3.5-397B ranked only 18th.

Perhaps it is also because of this that external doubts have arisen: that Alibaba’s focus is to operate large models like a “super app,” that the C-end is “rushing forward” while the B-end/model-end is slowing down, and that this change is related to compute resources being tilted toward C-end products rather than the model-side, causing imbalance.

Whether Alibaba’s establishment of Token Hub this week and its emphasis on pushing both AI to B and AI to C is indeed a “counterattack” to the aforementioned rumors remains unknown. However, this week, media reported that the idea of setting up Token Hub had already been under consideration by the end of 2025, when Alibaba was actively promoting the C-end product Qianwen App.

Since the beginning of the year, multiple domestic models have been launching new releases in close succession, and the battle for the Spring Festival film season has further promoted the popularization of AI applications. The “lobster” craze is still ongoing. Amid intense market competition, what surprises Alibaba will bring in the AI space next, Cailianshe reporters will continue to watch.

（Cailianshe reporter Fu Jing）