Alibaba's AI Accounting Starts to Make Sense This Quarter

Over the past year, Alibaba's financial reports have consistently revolved around one core question: When will large-scale capital investments translate into substantive business returns?

The first-quarter results for fiscal year 2027, released on August 20, showed that Alibaba generated revenue of RMB 268.953 billion for the quarter, a 9% year-over-year increase; adjusted EBITA was RMB 27.329 billion, down 30% year-over-year; and net profit was RMB 10.444 billion, down 75% year-over-year.

Alibaba's capital expenditures (CapEx) further rose to RMB 67.678 billion this quarter, a 75% year-over-year increase, while net outflow of free cash flow expanded to RMB 44.670 billion.

According to previously announced plans, Alibaba intends to invest RMB 380 billion in AI infrastructure over the next three years. To date, cumulative investment has reached approximately RMB 190 billion, meaning more than half of the planned spending has been completed.

However, this quarter, two of Alibaba's core investments—AI infrastructure and instant retail—began to show clearer paths to returns.

This quarter, Alibaba Cloud's AI cloud and computing power services revenue grew 45% year-over-year, while adjusted EBITA surged 133% year-over-year, with the profit margin rising to 11.6%.

Management provided a recovery model for AI computing power investments for the first time: Based on current gross margin levels for AI products, related CapEx is expected to break even within three years.

In instant retail, Alibaba readjusted the reporting scope of its e-commerce business, explicitly stating that the instant retail segment is expected to achieve overall profitability in FY2029 and could contribute 30% of the platform's total gross merchandise volume (GMV) in the long term.

As computing power investments begin to align with capital recovery cycles, and instant retail starts to clarify its profitability timeline, Alibaba's heavy-investment strategy appears to be entering a new phase of validation.

The Three-Year Break-Even Calculation for Computing Power

The 45% growth rate of Alibaba Cloud itself does not represent a significant expectation gap. Compared to revenue, changes on the profit side are more noteworthy.

This quarter, revenue from AI cloud and computing power services amounted to RMB 48.437 billion, up 45% year-over-year; adjusted EBITA reached RMB 5.628 billion, up 133% year-over-year, with the profit margin rising to 11.6%.

More importantly, Alibaba provided a relatively clear capital recovery period for AI computing power investments during its earnings call for the first time.

Wu Yongming, Chairman of the Board and CEO of Alibaba Group, stated during the earnings call that based on the current average gross margin level of AI products, AI-related CapEx "can achieve break-even within three years."

"If we maintain our current AI product gross margin levels and aim for CapEx break-even within three years, theoretically keeping our growth rate within 33% would allow us to achieve positive cash flow," Wu said.

The "three-year break-even" mentioned here primarily refers to the capital recovery of AI computing assets. This can be expressed by the formula: Break-even period ≈ Unit computing power investment ÷ Annual gross profit contributed per unit of computing power.

Management believes that, given current AI product gross margin levels, this asset-heavy model theoretically meets the conditions for achieving cash flow balance; however, Alibaba is not prepared to do so at this stage.

AI is still in the early stages of industry development. Compared to releasing cash flow prematurely, it is more important to continue increasing computing power supply and maintaining high-speed growth.

Therefore, the company has chosen to reinvest new cash into infrastructure rather than actively suppressing business growth rates to near the cash flow balance line.

A key premise for Alibaba's belief that this model holds is the long actual usage cycle of computing assets.

Wu cited examples, noting that the V100 chips purchased by Alibaba's data centers in 2018 and the A100 chips purchased in 2020 are still operating at near full capacity. The actual service life of AI computing assets may be significantly longer than their theoretical depreciation period, allowing them to continue generating cash flow after initial capital recovery.

Of course, physical operation does not imply that GPUs of different generations possess the same economic value. As chip efficiency iterates rapidly, the unit price and profitability of older computing power may still decline.

Therefore, for Alibaba, the more critical task remains continuously improving the investment efficiency of new computing power.

Currently, there are three main paths.

First is increasing the gross margin of AI products. Management explicitly stated during the earnings call that continuing to advance frontier model capabilities and expanding high-margin Model-as-a-Service (MaaS) businesses are important ways to boost AI product gross margins.

As of August, the annualized recurring revenue (ARR) of MaaS businesses, including Bailian, had exceeded RMB 16 billion, with Alibaba maintaining its year-end target of surpassing RMB 30 billion. Higher-margin revenue streams are rapidly scaling up.

Second is reducing the capital expenditure required to obtain equivalent computing power.

T-Head's current self-developed chip portfolio already covers GPUs, CPUs, and network chips. As production capacity for self-developed chips ramps up, Alibaba hopes to increase the proportion of self-developed chips within its data centers, reducing reliance on external commercial chip purchases, thereby lowering the initial CapEx needed to acquire the same amount of computing power.

Third is reducing the upfront capital Alibaba needs to bear.

In addition to co-building AI computing centers with partners, the company is supplementing computing power through operating expenses such as leasing and has begun collecting prepayments from customers for certain cloud computing services.

If customer prepayments can cover or even exceed the prepayments Alibaba pays to suppliers, the occupation of free cash flow by new business will further decrease.

Ultimately, Alibaba aims to create a state where revenue growth outpaces the growth of new production capacity.

Communications following the earnings call, widely circulated in the market, indicated that management believes sell-side estimates of the current return on invested capital (ROIC) for AI investments may still be too conservative. Alibaba Cloud's revenue growth in the September quarter may further exceed 50%, with potential for continued acceleration in the subsequent December and March quarters.

Is Instant Retail Becoming the "Second Curve"?

As the investment that has most impacted Alibaba's profit performance over the past few quarters, the accounting for instant retail is also becoming clearer.

This quarter, Alibaba readjusted the disclosure scope of its e-commerce business, integrating China Commerce, China Instant Retail, International Digital Commerce, and Global B2B into the newly formed Alibaba E-Commerce Group.

The group generated revenue of RMB 205.862 billion this quarter, a 4% year-over-year increase; adjusted EBITA was RMB 39.749 billion, down 1% year-over-year.

Among these, China Instant Retail was disclosed under a relatively complete business scope for the first time, with revenue reaching RMB 53.295 billion, a 45% year-over-year increase. This includes Taobao Flash Purchase, Hema, and Tmall Supermarket's instant delivery services, with growth primarily driven by Taobao Flash Purchase and Hema.

Traditional e-commerce, which serves as Alibaba's profit foundation, continues to face severe challenges.

This quarter, Customer Management Revenue (CMR) decreased 7% year-over-year under the reported basis. After excluding the impact of marketing subsidies treated as revenue reductions, CMR on a comparable basis increased 1% year-over-year.

As shelf-based e-commerce enters a phase of stock competition, instant retail is bearing Alibaba's expectations for finding new growth drivers.

Jiang Fan, CEO of Alibaba's E-Commerce Business Group, provided a clearer long-term goal for the first time during the conference call: "The instant retail segment is expected to achieve overall profitability in FY2029. In the long term, instant retail is poised to contribute 30% of the platform's total GMV, becoming the second curve for the e-commerce sector."

Hema, newly included in the E-Commerce Group's disclosure scope, is itself a relatively mature instant retail asset that continues to expand.

In FY2026, Hema's GMV exceeded RMB 107 billion, with online transactions accounting for over 60% of Hema Fresh's GMV. It has achieved positive adjusted EBITA for the full year for two consecutive years.

Within Alibaba's broader instant retail system, Hema's value lies in complementing the product offerings and fulfillment capabilities required for non-food instant retail.

This quarter, Hema maintained strong year-over-year growth in both orders and revenue, benefiting on one hand from expansion into emerging counties and cities, and on the other from deepened cooperation with Taobao Flash Purchase, providing the platform with differentiated supplies such as daily necessities and fresh produce.

Jiang Fan stated that the company would "accelerate the integration of relevant segments of Hema and Tmall Supermarket to develop non-food instant retail, especially accelerating the development of forward warehouses."

He expects that non-food instant retail GMV will surpass food categories in the next fiscal year, further driving physical goods transactions within the broader e-commerce landscape.

Compared to food, orders for fresh produce, daily necessities, and fast-moving consumer goods typically have higher average order values, supplementing consumption demand outside of lunch and dinner hours, thereby improving the utilization efficiency of warehousing, rider, and fulfillment networks.

At the same time, the value of instant retail is not limited to profitability per order. Alibaba still hopes that the high-frequency consumption scenarios brought by instant retail will further enhance Taobao user activity and feed back into the platform's commercialization.

The synergy between Hema, Tmall Supermarket, and Taobao Flash Purchase will not automatically lead to profitability.

The FY2029 target ultimately depends on whether the unit economics model of Taobao Flash Purchase can continue to improve, and whether order growth, optimization of product structure, and improved fulfillment efficiency can cover new costs such as warehousing and delivery.

But compared to the previous situation where the market could only see the scale of investment, this quarter Alibaba began to outline a clearer path for the instant retail account.

AI still requires continued data center construction, and Qwen still bears high inference costs; Taobao Flash Purchase is also several years away from overall profitability.

But at least for now, Alibaba's two most watched long-term investments have, for the first time, clear return coordinates: AI corresponds to a capital recovery cycle of about three years, and instant retail corresponds to a profitability node in FY2029.

Moving forward, what the market truly focuses on is no longer just how much Alibaba is willing to invest, but whether these investments can be recouped step by step according to the timeline provided by management.

Risk Warning and Disclaimer

The market carries risks; investment requires caution. This article does not constitute personal investment advice, nor does it take into account the specific investment objectives, financial status, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article align with their specific circumstances. Responsibility for investments made based on this content rests solely with the investor.