Alibaba's AI-related accounts have been calculated so clearly for the first time.

A financial report with a 76% year-on-year drop in net profit is not necessarily bad news.

This is my first impression after reading the first-quarter financial report for Alibaba Group's 2027 fiscal year released on the evening of August 20. What is really worth pondering is not the decline in profits, but the fact that Alibaba has completely changed its reporting caliber starting this quarter — the company is split into four segments: E-commerce Group, AI Cloud and Computing Power, AI Labs and Applications, and "All Other". The change in the structure of the books indicates that the company wants the market to read it in a new way.

Let's first lay out the full framework of this financial report. For the quarter ending June 30, Alibaba recorded revenue of 2689.53 billion yuan, a year-on-year increase of 9%; net profit attributable to owners of the parent was 105.37 billion yuan, down 76% year on year; operating profit was 151.61 billion yuan, down 57% year on year. After the financial report was released, U.S. stocks fell more than 4% in pre-market trading, but Hong Kong stocks closed up 1.61% the next day — the same set of figures drew two completely opposite reactions from the two markets.

The divergence between the two markets points to the same core: the new segment reporting caliber that Alibaba adopted this time.

Profitable AI: Cloud Business Enters Harvest Period

The first segment that has achieved commercial viability under the new caliber is "AI Cloud and Computing Power Services", the new business segment formed after the Cloud Intelligence Group merged with T-Head Semiconductor.

According to comprehensive information from the financial report and Tianyancha Media, this quarter the segment recorded revenue of 484.37 billion yuan, a year-on-year increase of 45%, hitting a new 22-quarter high in growth rate; among which revenue from AI-related products reached 123.76 billion yuan, marking the 12th consecutive quarter of triple-digit growth. Calculated based on financial report data, the annualized AI revenue has exceeded 495 billion yuan. What is more remarkable is the profit side: the segment's adjusted EBITA surged 133% year-on-year to 56.28 billion yuan, with the profit margin rising from 7.2% to 11.6%. The simultaneous rise in growth rate and profit margin indicates that the scale effect has entered the realization period.

The sustained momentum of this business is supported by two easily overlooked details. The first is chips: T-Head has built a full-stack self-developed system covering GPU, CPU, storage and network chips. Its latest generation of AI processor Zhenwu M890 has served more than 650 external customers across over 20 industries through Alibaba Cloud — Amazon's self-developed chip business has an ARR of more than 25 billion U.S. dollars, which has verified the path that self-developed chips can improve cloud profit margins, and Alibaba is replicating this path. The second is market share: Omdia's "China AI Cloud Market Share Report 2025" shows that Alibaba Cloud ranks first in China's AI cloud market with a 38.1% share.

In short, Alibaba's upstream computing power business has started to make profits, and the profit growth is accelerating.

Money-burning AI: How Costly Is Qwen's Loss Sheet

When we shift our view to the next statement, the situation changes completely.

The newly established AI Labs and Applications segment — which includes the AI model laboratory, Qwen App and Qwen Office — recorded revenue of 33.38 billion yuan this quarter, a year-on-year increase of 16%, but its adjusted EBITA posted a loss of 138.61 billion yuan. The figure in the same period last year was a loss of 32.24 billion yuan, expanding by 330% in a year, and the loss amount is 4.2 times the revenue. Alibaba's explanation is straightforward: the increase in investment in AI capabilities, coupled with the rise in reasoning costs related to the Qwen App.

The logic behind this account is reasonable: 250 million users of the Qwen App have experienced AI shopping. Each additional call from a user consumes more reasoning computing power; the total global downloads of the Qwen series of models have exceeded 3 billion times, with more than 300,000 derivative models. In August, Alibaba just open-sourced Qwen3.8-Max with 2.4 trillion parameters. Open source builds the ecosystem, and the ecosystem diverts traffic back to Alibaba Cloud's API calls — losing money at the application layer and making profits at the computing power layer, that's how Alibaba plans to operate.

But whether this plan can succeed depends on whether the model capabilities can keep up with the industry, as well as the paid conversion rate of Qwen Office and Qwen App. This financial report has not given answers to either of the two points.

Where Did the Profits Go: Let's Break It Down Ourselves

Rather than generally saying that AI has dragged down profits, it is better to break down the 76% year-on-year decline. The funds are mainly used in four directions.

The first part is the AI investment itself: the AI Labs and Applications segment recorded an additional loss of more than 100 billion yuan in a single quarter. The second part is that the "All Other" segment turned from profit to loss — the segment that includes Alibaba Health, Youku, and Amap saw its adjusted EBITA change from a profit of 6.87 billion yuan to a loss of 33.43 billion yuan, resulting in a total reduction of 40 billion yuan in profit. The third part is one-off items: goodwill impairment reached 44.58 billion yuan, compared with zero in the same period last year; in addition, there is a provision of about 43 billion yuan for EU fines, which directly pushed the general and administrative expense ratio from 3.0% to 4.7%. The fourth part is hidden below the income statement: net interest and investment income was cut in half from 173.76 billion yuan to 90.04 billion yuan, while income tax increased by 44% to 127.98 billion yuan, and the effective tax rate jumped from 17.6% to 57.4% — the profits in the same period last year included a large amount of low-tax investment income, and when such income receded this year, the tax burden structure changed immediately.

The divergence in the cash flow statement is also worth noticing. The net cash flow from operating activities was 229.45 billion yuan, a year-on-year increase of 11%, indicating that the basic business still maintains strong profitability; however, the net free cash flow outflow was 446.70 billion yuan, with the gap more than doubling compared with the same period last year. Capital expenditure reached 676.78 billion yuan, a year-on-year increase of 75% — a quarter of the single-quarter revenue was invested in AI infrastructure.

The action of converting profits into computing power is not unique to Alibaba.

Just a week ago, Tencent released a financial report with almost the same trend: capital expenditure in the second quarter reached 528 billion yuan, a year-on-year surge of 176%, and free cash flow turned negative to 138 billion yuan. The AI competition among large domestic tech companies has entered a stage of competing for heavy assets from competing for model performance.

The Conclusion Is Left for Next Quarter

Putting the profits of the four business segments together, Alibaba's capital circulation is clear at a glance: the E-commerce Group's adjusted EBITA is 397.49 billion yuan, which is the only large profit pool of the entire group. The money it earns is being systematically transferred to both ends of AI — the cloud and computing power segment earned 56.28 billion yuan with a rising profit margin, the AI Labs and Applications segment burned 138.61 billion yuan, and the "All Other" segment lost another 33.43 billion yuan.

Wu Yongming put forward a judgment three months ago: Alibaba's AI has crossed the initial investment stage and entered a positive cycle of large-scale commercial returns. This financial report can be said to partially verify this judgment — the growth and profit margin of the cloud business have indeed entered the return cycle; but the application layer is obviously still in the deep water zone of the investment period.

Therefore, it is too early to draw any conclusions now. Three indicators are really worth tracking: whether the proportion of AI-related product revenue in the external revenue of the cloud segment can continue to rise, whether the EBITA profit margin of the cloud segment can stay above 12%, and when the loss of the AI Labs segment will start to narrow. If these three points reach an inflection point, Alibaba's full-stack AI story will truly complete a closed loop; if these inflection points do not come for a long time, the 76% decline in profit will not be growing pains during transformation, but a pricing that needs to be re-evaluated.

The significance of the revised financial report caliber lies here: Alibaba has separated the AI accounts from the vague "All Other" category, listed them separately, and made them responsible for their own profits and losses.

The courage to make the accounts clear is in itself a statement of attitude.

This article is from the WeChat official account "LingTai LT" (ID: LingTai_LT), written by Zhang Qian, edited by Hu Zhanjia, and published with authorization from 36Kr.