"Double-Faced Alibaba": Accelerating and Decelerating

Questioning AI · How to Sustain AI Investment During E-Commerce Slowdown?

Text | Xiaojing

Editor | Xu Qingyang

On March 19, Alibaba announced its Q3 FY2026 results (ending December 31, 2025). The quarter’s revenue was 284.84 billion RMB, up 2% year-over-year; excluding the sold Gaoxin Retail and Intime businesses, same-store revenue grew 9%. Net profit was 15.63 billion RMB, down 66% YoY. Non-GAAP net profit was 16.71 billion RMB, down 67%. Adjusted EBITA was 23.4 billion RMB, down 57%. (Financial details)

Revenue mostly met Wall Street expectations, but the profit deterioration exceeded the market’s previous estimate of about 40% decline, with actual net profit dropping 66%. Non-GAAP EPS was $7.09 per ADS, down 67% YoY. The earnings report highlights that instant retail, user experience, technology investments, and rising capital expenditures are key factors pressuring profits and cash flow.

The report also reveals a more fragmented picture of Alibaba: cloud business grew 36%, continuing to accelerate; e-commerce CMR (Customer Management Revenue) growth sharply dropped from 10% last quarter to 1%; flash sales revenue increased 56% YoY, but the report did not disclose losses separately; the significant decline in China e-commerce EBITA indicates ongoing heavy investment in instant retail is still dragging down segment profits. Free cash flow over nine months was negative 29.3 billion RMB.

Three days earlier (March 16), Alibaba announced the formation of the ATH Business Group, with CEO Wu Yongming directly leading it, integrating Logic Lab, Bailian MaaS, Qianwen, Wukong, and AI Innovation into an independent organization. The next day, they launched the enterprise-level Agent platform “Wukong.”

Market analysts call Alibaba “the most Google-like Chinese company,” based on the high structural similarity of their AI tech stacks.

Google’s “trinity” consists of Gemini (models) + Google Cloud (cloud platform) + TPU (custom chips). Alibaba’s internal equivalent is called “Tongyun Ge”: Tongyi Lab (Qwen large model) + Alibaba Cloud + Pangu (including Guang, Yitian, PPU, etc., self-developed chips).

The establishment of the ATH Business Group marks Alibaba’s first organizational move to separate AI from cloud services, positioning it alongside e-commerce and cloud as an independent pillar. For capital markets, this suggests future potential for separate valuation of AI assets. Morgan Stanley noted on the day of ATH’s formation that this unit could add a new valuation component outside the existing SOTP framework. However, organizational structure alone is just a prerequisite for valuation, not the valuation itself.

01. Alibaba Cloud: 36% Growth

This quarter, the Cloud Intelligence Group’s revenue was 43.28 billion RMB, up 36% YoY, continuing to accelerate from 34% last quarter (and 26% the quarter before). External customer revenue grew 35%, also higher than the previous quarter’s 29%. AI-related product revenue has maintained triple-digit growth for ten consecutive quarters. Adjusted EBITA for cloud was 3.91 billion RMB, up 25%.

These figures are positive. The 36% growth, slightly below Goldman Sachs’ forecast of 38%, still confirms a three-quarter acceleration trend (from 26% to 34% to 36%). Goldman Sachs upgraded Alibaba to a Conviction Buy based on cloud growth exceeding 30%, which has been validated.

However, some details need to be unpacked. First, cloud profit growth (25%) is significantly lower than revenue growth (36%), indicating that Alibaba is spending heavily on customer acquisition and capacity expansion, which has compressed margins. Second, the report still does not disclose absolute AI revenue, only noting “triple-digit growth for ten consecutive quarters,” leaving the base size and proportion of AI revenue in total cloud unclear.

Comparing with Google Cloud: in Q3 2025, Google Cloud’s revenue was about $15.2 billion (roughly 110 billion RMB), up 34% YoY, and profitable. Alibaba Cloud’s revenue this quarter was 43.28 billion RMB (~$6.2 billion), about 40% of Google Cloud’s size. Alibaba Cloud’s 36% growth slightly outpaces Google Cloud’s 34%, but the absolute scale gap remains large. The adjusted EBITA of 3.91 billion RMB corresponds to roughly a 9% profit margin, with growth lagging revenue growth.

Another noteworthy signal: Pangu’s GPUs have been scaled up for mass production, supporting Qianwen and cloud infrastructure, providing high-cost-performance AI services to external clients and contributing meaningfully to cloud infrastructure supply. The report states “scaled rapidly and now contributes meaningfully,” suggesting self-developed chips could reduce inference costs, which is beneficial long-term but still involves heavy investment short-term.

Immediately after ATH’s formation, Alibaba released the enterprise Agent platform “Wukong.”

Wukong is positioned as a “B-end AI native work platform,” leveraging DingTalk’s enterprise user base to embed model capabilities into workflows. Alibaba emphasizes “security, compliance, and auditability” as differentiators. These features are essential in enterprise scenarios, but whether they command a premium depends on whether companies are willing to pay more for “Agents working compliantly within enterprises” than for raw API calls.

Alibaba’s logic: the same tokens sold via API to developers only cost the call fee; embedding models into core business processes via Wukong could increase token value through higher per-customer prices and longer cooperation cycles. Theoretically plausible, but actual customer conversion and renewal rates are key.

Wu Yongming previously said internally that “a large number of digitalization tasks will be supported by hundreds of billions of AI Agents,” but this remains a distant goal.

A related data point in the earnings report: Qianwen App integrated with Taobao Flash Sale (instant retail) on January 15; by February, MAU across the platform exceeded 300 million, with 140 million users completing shopping, food delivery, flight booking, etc., via Agent functions. This indicates B-end Agent deployment within Alibaba’s ecosystem is starting to work, but real-world data for Wukong’s B-side remains to be seen.

By comparison, Google’s Gemini embedded in Google Workspace follows a similar path, supported by a billion-level paid user base. DingTalk, while sizable in China’s enterprise market, has lower paid penetration and ARPU than Google Workspace. Whether Wukong can succeed in the “Agent as a Service” model remains more a vision than reality.

02. CMR Growth Plummets to 1%, the Cost of Flash Sales Continues

This quarter, Alibaba China E-commerce Group’s revenue was 159.35 billion RMB, up 6% YoY. Breaking down: e-commerce revenue was 131.58 billion RMB, up just 1%; CMR (Customer Management Revenue) was 102.66 billion RMB, up 1%. Last quarter’s CMR growth was 10%, the previous quarter also 10%, but this quarter’s growth sharply dropped to 1%. The explanation: “weaker transaction activity and gradual fading of software service fees.” Taobao Flash Sale (instant retail) revenue was 20.84 billion RMB, up 56%. Adjusted EBITA for China E-commerce was 34.61 billion RMB, down 43% YoY.

The plunge of CMR growth from 10% to 1% is one of the most concerning data points in this report. It indicates that core monetization—advertising and commissions from merchants—has essentially stalled. 88VIP members exceeded 59 million, with double-digit YoY growth; Taobao app MAU also grew double digits YoY, but user growth did not translate into revenue growth, implying declining monetization per user or that incremental user spending from flash sales is minimal.

In instant retail, revenue from flash sales was 20.84 billion RMB, up 56%, but slower than last quarter’s 60%. Management said that this quarter’s flash sales continued to improve unit economics (UE), with average order value increasing month-over-month, and order structure shifting toward high-value dining and non-food categories, with strong user retention.

These are positive signals, but the report did not disclose flash sale losses separately. We can roughly estimate that flash sales still consume profits, as total China E-commerce Group’s adjusted EBITA decreased by about 25.8 billion RMB (from 60.4 billion to 34.6 billion). Last quarter, Goldman Sachs estimated flash sale losses at about 36 billion RMB for one quarter. Management promised “significant contraction” this quarter. The narrowing of EBITA decline from 76% last quarter to 43% this quarter suggests losses are decreasing, but the absolute amount remains substantial.

A deeper issue: can the user growth driven by flash sales feed back into e-commerce profitability? The drop in CMR growth from 10% to 1% offers a pessimistic short-term outlook. At least this quarter, the active users attracted by flash sales have not translated into merchant ad and commission revenue.

A stark comparison: Google’s search advertising contributed over $260 billion in revenue in 2025, with high profit margins, supporting about $90 billion in annual capital expenditure. Alibaba’s e-commerce should play a similar role, but Tmall’s CMR growth has fallen to 1%. If Tmall cannot “maintain profits” and “absorb flash sale losses,” Alibaba’s AI investments could be constrained, and the “Chinese Google” narrative would lack a key pillar.

03. Alibaba Still Has Sufficient Cash, but Spending Accelerates

This quarter, net cash outflow related to capital expenditure was 29 billion RMB, slightly below last year’s 31.8 billion RMB. However, cumulative capex over nine months reached 99.2 billion RMB, up 62% from 61.4 billion RMB last year.

More concerning is cash flow: operating cash flow was 36 billion RMB, down 49% YoY (from 70.9 billion RMB); free cash flow was only 11.3 billion RMB, down 71%. Over nine months, cumulative free cash flow was negative 29.3 billion RMB—last year, it was positive 70.1 billion RMB.

These figures directly answer “where the money went”: heavy spending on AI infrastructure and flash sale subsidies squeezed operating cash flow, turning free cash flow negative over nine months. On Alibaba’s balance sheet, net cash position as of December 31 fell from 366.4 billion RMB at the start of the year to 297.4 billion RMB, a decrease of about 69 billion RMB over nine months. Alibaba still has ample cash reserves, but the rate of consumption is increasing.

Goldman Sachs previously raised Alibaba’s FY2026–FY2028 capex forecast to 513 billion RMB, expecting FY2027/FY2028 EPS to grow 31% and 36% YoY respectively. This model assumes AI revenue will ramp up quickly. But this quarter’s data shows cloud revenue growth accelerating, while profit growth remains slow, indicating that the efficiency of investment converting into revenue still needs observation.

04. “Dual-Faced Alibaba”: What’s Missing for the “Chinese Google” Narrative?

The report provides mixed answers. On the positive side: 36% cloud growth (outpacing Google Cloud’s 34%), AI product revenue for ten consecutive quarters with triple-digit growth, Pangu GPUs beginning mass delivery, and Qianwen MAU surpassing 300 million. The “Tongyun Ge” system is indeed aligning with Google in terms of technology and growth momentum.

But on the other hand, the group’s net profit fell 66%, adjusted EBITA down 57%, and nine-month cumulative free cash flow is negative 29.3 billion RMB. CMR growth plummeted from 10% to 1%, and the “blood supply” e-commerce business is slowing. Flash sales, though narrowing losses, still consume large amounts of cash. Cloud profit growth (25%) lags behind revenue growth (36%), indicating that scale expansion has not yet improved margins proportionally.

Analysts covering Alibaba have an average 12-month target price of about $195, with the highest at $237 (Nomura). Goldman Sachs has a “Conviction Buy” at $186, Morgan Stanley an “Overweight/Top Pick” at $180, and Jefferies a “Buy” at $225. Overall, the consensus is bullish, but all these targets assume: cloud growth remains above 30%, AI commercialization accelerates, and e-commerce maintains cash flow. The first is confirmed, the third is problematic.

The formation of ATH, the launch of Wukong, and the “Tongyun Ge” system are all aimed at supporting the “Chinese Google” narrative.

Alibaba’s market cap is about $326 billion, compared to Google’s roughly $3.75 trillion—more than ten times larger. Alibaba’s TTM P/E ratio is just over 18, while Google’s is nearly 29. From a valuation perspective, Alibaba’s valuation isn’t high.

But narrative does not equal reality. Behind Google’s “trinity” are $400 billion annual revenue, 33% profit margins, and over a decade of cloud computing accumulation. Alibaba’s “Tongyun Ge” can be compared in architecture and cloud growth, but in profitability, cash flow quality, and e-commerce fundamentals, this report shows a widening gap.

Alibaba is becoming two very different faces: one of high-growth cloud and AI, and another struggling through flash sale subsidies and weak consumer spending.

How much the market is willing to pay for the former depends on the ability to sustain the latter’s “blood supply.”