On a par with competitors, Alibaba is entering a new phase internally.

Written by | Lan Dong Business Zhao Weiwei

Alibaba’s Q2 2026 financial report revealed profit pressures amidst heightened AI investments.

AI and cloud businesses continued to be the primary focal points. Compared to Microsoft Cloud and Amazon Cloud, which maintained revenue growth rates of 43% and 37%, respectively, Alibaba Cloud achieved a 45% revenue increase this quarter, keeping pace with international counterparts and positioning itself in the same league as its competitors.

However, this growth comes with increased AI investments. With Tencent’s latest quarterly capital expenditures reaching RMB 52.784 billion, Alibaba’s latest quarterly capital expenditures surged to RMB 67.678 billion, marking a 75% year-on-year increase and outpacing Tencent’s bets. Combined, Alibaba and Tencent’s single-quarter capital expenditures exceeded RMB 120 billion.

The accelerated capital expenditures reflect the rising AI demand faced by cloud providers, making it a race where no one can afford to lag behind.

Over the past year and a half, Microsoft, Google, and Meta’s quarterly capital expenditures have soared from $30 billion to $60 billion, becoming a core metric for the sustained AI narrative. AI demand is far from reaching its peak, with significant market expansion potential remaining.

In the broader cycle of AI’s accelerated commercialization, Alibaba is on a par with its competitors, but internally, it marks the beginning of a new phase relative to its own trajectory.

A year ago, Alibaba announced a three-year, RMB 380 billion AI investment plan. Now, Eddie Wu revealed during the earnings call that total investments reached RMB 190 billion by the end of Q2 this year, meaning over half of the AI investment plan has been completed, with the focus now shifting to the more critical next phase.

Regarding the return cycle for AI capital expenditures, Eddie Wu provided a clear answer during the earnings call: AI computing capital expenditures offer high investment return certainty, with current overall investments expected to recoup within three years. As gross margins for AI-related products continue to rise, the recovery cycle for computing assets is expected to shorten further to around 2.5 years, or even two years.

Funds saved from the food delivery battle are now flowing faster into the AI arena, marking Alibaba’s most critical strategic pivot. The food delivery war is over, but the AI investment battle is only halfway through.

ATH Business Group's Performance Remains Unclear

In this quarter's financial report, Alibaba adopted a new performance disclosure framework, dividing operations into four segments. However, the performance of the newly established ATH (Alibaba Token Hub) business group remains unresolved.

According to Alibaba's latest quarterly financial report, the original Chinese e-commerce, international e-commerce, and other segments, including Hema, were consolidated into Alibaba's e-commerce segment. The "AI Cloud and Computing Services" segment includes Alibaba Cloud and T-Head chip businesses, while the "AI Labs and Applications" segment encompasses AI model R&D, Qianwen C-end business unit, and Qianwen Office business group.

The inaugural division of "AI Cloud and Computing Services" and "AI Labs and Applications" provides a clear view of Alibaba's revenue and investments in AI models and applications. Among them, "AI Cloud and Computing Services" holds the primary position, with revenue of RMB 48.437 billion, a 45% year-on-year increase. In contrast, "AI Labs and Applications" generated RMB 3.34 billion in revenue, a 16% year-on-year increase.

Currently, cloud and computing services remain the largest contributors to Alibaba's AI revenue, while AI application models are still in the investment phase, incurring a loss of RMB 13.9 billion. Meanwhile, the newly established ATH business group, as the core carrier of AI infrastructure, is gradually beginning to show its performance contributions.

AI cloud business has become one of the most critical drivers of Alibaba Cloud's revenue contributions. During the earnings call, Eddie Wu disclosed that Alibaba's AI-related products' annualized revenue (ARR) surpassed RMB 49.5 billion this quarter, accounting for 35% of Alibaba Cloud's external commercial revenue.

This means AI has substantially contributed over one-third of Alibaba Cloud's external revenue. Moreover, due to higher gross margins for AI products, it is driving the entire cloud division's profit margins from breakeven towards a healthy 12% range. AI is not just a growth story but has become a core component of the cloud business revenue structure.

A significant part of this growth comes from the MaaS (Model as a Service) business.

Last quarter, Alibaba expected MaaS's annualized revenue to exceed RMB 10 billion this quarter. By August this year, this figure surpassed RMB 16 billion, significantly exceeding the previous target, while the year-end target of RMB 30 billion remains unchanged.

Another highly anticipated figure comes from T-Head's chip business, which has begun accelerating commercialization, though specific chip revenue was not separately disclosed in the financial report.

Eddie Wu mentioned that T-Head chips have entered a large-scale commercialization phase, with the previous generation of T-Head chips manufacturing and shipping over 500,000 units. Super node instances based on the latest-generation Zhenwu M890 have been launched on Alibaba Cloud and are now available for sale, with volume ramping up in the second half of the year.

As of early August, Zhenwu chips have served over 650 clients. T-Head has formed a full-stack self-developed chip layout covering GPUs, CPUs, and network chips, with Eddie Wu noting, "We are indeed first in the industry."

Currently, Alibaba's large-scale AI data center delivery cycle has been compressed to 100 days. The speed of chip business and underlying data center construction is the biggest assurance for Alibaba to guarantee a three-year return on AI computing investments.

How Will Expenditures Translate into Profits?

Investing without immediate returns—does demand outweigh investment?

This is the biggest concern regarding tech giants' increased AI bets. Amazon AWS, Google Cloud, and Microsoft Cloud each have distinct advantages, with Alibaba Cloud still in a catch-up position.

In terms of revenue scale, Amazon AWS remained the largest in the latest quarter, with revenue of $42.2 billion. Microsoft Intelligent Cloud followed with $39.3 billion, while Google Cloud remained in the third tier, with revenue of $24.8 billion, about 40% behind the leader in absolute terms.

However, Google excels in growth, with an 82% year-on-year increase in the latest quarter, far outpacing peers. It is the only top cloud provider achieving near-doubling high-speed growth, driven by AI-customized computing power and TPUs, forming its biggest differentiated capability.

From a profitability perspective, Microsoft Cloud leverages strong synergies between software subscriptions and cloud services, with high-margin AI products like Copilot maintaining high profitability, achieving a 40.6% profit margin, a target Amazon and Google Cloud continue to pursue.

Temporary pressure on free cash flow has become an industry norm. In terms of capital expenditures, Amazon is the most aggressive, followed by Google, with both experiencing temporary free cash flow pressures. Only Microsoft maintains a slightly lower investment intensity, preserving positive free cash flow and optimal return efficiency.

Every change in cloud providers' metrics attracts outside attention and is amplified. Google Cloud currently has the fastest revenue growth among the four major cloud providers. Google CEO Sundar Pichai defined the current investments as the "early stages of a long-term structural shift" and plans to expand third-party cloud computing power procurement to fill gaps during the transition period of limited self-owned computing capacity delivery.

Demonstrating that demand outweighs current investments is the primary way cloud providers address doubts.

The backlog of orders for the four major U.S. cloud service providers has exceeded $2.3 trillion, with Google's existing orders surpassing $514 billion, paving the way for Google's performance certainty. The revenue realization cycle for Google's self-developed TPU chip external sales will also become a key variable for its 2027 performance growth.

The most significant similarity between Google and Alibaba is their full-stack AI closed loop. Both have complete chains from self-developed AI chips, public cloud infrastructure, native large models to proprietary business scenarios. Especially in terms of business similarities, both use cloud infrastructure as their profit base, with large models and AI applications shouldering high investments, while self-developed chips undertake both cost reduction and commercialization tasks. They are increasing capital expenditures while relying on order backlogs to address capital market concerns about ROI.

Regarding the return cycle for AI capital expenditures, Eddie Wu also provided a clear answer: AI computing capital expenditures offer high investment return certainty, with current overall investments expected to recoup within three years. As gross margins for AI-related products continue to rise, the recovery cycle for computing assets is expected to shorten further to around 2.5 years, or even two years.

To maintain a leading position in the next-generation AI competition, cloud providers are willing to sacrifice short-term free cash flow to build long-term infrastructure barriers.

The Food Delivery War Shifts Gears

The food delivery war has given way to the AI war, marking the end of the era of subsidies for scale in instant retail. This is the clear message from Alibaba's financial report.

Looking back at the past few quarters, instant retail subsidies from Taobao Flash Sales once became the biggest bleeding point on Alibaba's profit statement, with quarterly losses in the billions, dragging the entire group's EBITA into a sharp decline. The rapid narrowing of losses this quarter indicates Alibaba's withdrawal from the food delivery competition, with resources shifting focus.

During the earnings call, Alibaba explicitly stated that instant retail would "accelerate the integration of Hema and Tmall Supermarket" and "particularly focus on expanding front warehouses."

Notably, "in the next fiscal year, non-food category transaction volume in instant retail will surpass food categories, driving growth across numerous physical goods categories in the e-commerce business. Instant retail is expected to achieve overall profitability by FY2029. In the long term, instant retail is expected to contribute 30% of overall transaction volume, becoming Alibaba E-commerce's second growth curve."

In other words, the war is not over. The food delivery war, dominated by dining, has given way to non-dining instant retail, which clearly has a larger market space. The battlefield is no longer about subsidies for market share but about UE (Unit Economics) for profitability.

Reflected in the financial report, total revenue reached approximately RMB 269 billion, an 8.6% year-on-year increase. Overall adjusted EBITA was RMB 27.3 billion, with the year-on-year decline narrowing significantly from 84% last quarter to less than 30%, slightly better than market expectations. It is evident that the phase of high food delivery subsidies nearly depleting group profits is largely over.

Currently, China's instant retail business revenue is RMB 53.3 billion, a 45% increase, driven by Hema and Taobao Flash Sales. Alibaba's instant retail business includes the original Taobao Flash Sales, Hema, and Tmall Supermarket's one-hour delivery services.

Importantly, high AI investments offer more imagination than food delivery war investments and better align with current market competition demands.

Capital expenditures this quarter reached RMB 67.678 billion, a 75% year-on-year surge. The AI Labs and Applications segment's losses expanded to RMB 13.861 billion, primarily from large model R&D and C-end Qianwen inference costs. Meanwhile, T-Head's large-scale chip delivery, intelligent computing cluster expansion, and Tongyi large model iterations all require heavy investments. The freed-up resources are now flowing even faster into the AI arena.

The food delivery war is not over, but it is no longer Alibaba's main battlefield, releasing profit margins. The AI war's baton pass tests Alibaba's patience for long-term investments and its ability to deliver commercialization.