Cloud Reaccelerates to +34%, But a Spending Shock Buries the Profit Line: Initiating Alibaba at Hold

Results vs. Consensus

This was a print where the headline you led with determined whether you called it a beat or a miss. The top line and the strategically decisive cloud line came in ahead; every profitability measure came in sharply lower, both year-over-year and versus expectations. The reconciliation between those two facts is the entire quarter: Alibaba is deliberately spending its margin to buy market position in instant commerce and compute capacity in AI.

Year-over-Year Comparison (Q2 FY2026 vs. Q2 FY2025)

Sequential Comparison (Q2 FY2026 vs. Q1 FY2026, June quarter)

Segment Performance

Alibaba reported on its simplified structure adopted in the June quarter: Taobao and Tmall, Ele.me and Fliggy were combined into Alibaba China E-commerce Group, while Cainiao, Amap and the media businesses were reclassified into All others. The four reportable pillars are China E-commerce, AIDC (international), Cloud Intelligence, and All others.

Cloud Intelligence Group — the crown jewel reasserts itself

Cloud revenue of RMB39.8B (+34%) accelerated from +26% in the June quarter, with external-customer revenue up 29% and AI-related product revenue posting triple-digit growth for a ninth consecutive quarter. Segment adjusted EBITA rose 35% to RMB3.6B, holding a roughly 9% margin even as Alibaba pours money into capacity. The competitive position is the strongest in the group: management cited a 35.8% share of China's AI cloud market, ahead of the combined second-through-fourth players, and more than 180,000 Qwen derivative models on Hugging Face.

"Robust AI demand further accelerated our Cloud Intelligence Group business, with revenue up 34% and AI-related product revenue achieving triple-digit year-over-year growth for the ninth consecutive quarter." — Eddie Wu, CEO

Assessment: This is the segment that justifies owning the stock. An accelerating, share-gaining, full-stack AI cloud franchise with a proprietary frontier model (Qwen) is a genuinely scarce asset in China. The only caveat is that the ~9% segment margin means cloud is not yet large or profitable enough to offset the consumption drag; it is a growth and optionality story, not yet an earnings story.

China E-commerce Group — top line up 16%, profit down 76%

Segment revenue grew 16% to RMB132.6B, its fastest in several quarters, but adjusted EBITA fell 76% to RMB10.5B. The split within it is the story: core e-commerce (Taobao/Tmall) grew 9% with CMR up 10% on take-rate gains, while quick commerce revenue surged 60% to RMB22.9B and consumed the segment's profit. Management noted that excluding the quick-commerce loss, segment EBITA would have grown mid-single-digits year-over-year, meaning the base marketplace remains healthy and profitable.

"Excluding loss from our quick commerce business, our Alibaba China E-commerce Group EBITA would have grown at mid-single-digit year-over-year for the quarter. Going forward, this adjusted EBITA may fluctuate quarter-over-quarter due to intense competition and a significant investment in user experience." — Toby Xu, CFO

Assessment: The core marketplace is doing its job: take rate is rising, CMR is compounding, and 88VIP crossed 56 million members. The read-through is that Alibaba is not losing the marketplace war; it is choosing to spend marketplace profit to enter the instant-commerce war. Whether that is value-accretive depends entirely on the quick-commerce unit-economics path.

AIDC (International) — the quiet inflection

AIDC revenue rose 10% to RMB34.8B and, more importantly, the segment turned to a RMB162M adjusted EBITA profit from a RMB2.9B loss a year ago, driven by AliExpress Choice efficiency gains and the new "AliExpressDirect" local-inventory model spanning 30+ countries. Management cautioned that EBITA may fluctuate on tactical investment in select markets.

Assessment: A small but meaningful positive. AIDC swinging to profit removes one of the perennial loss centers from the bear case and shows Alibaba can drive operating discipline when it chooses to. It also demonstrates that the quick-commerce losses are a choice, not an inability to run a tight business.

All others — the disposal drag

All others revenue fell 25% to RMB63.0B, almost entirely on the deconsolidation of Sun Art and Intime, with a wider RMB3.4B EBITA loss reflecting technology-business investment (foundation models, AI apps), partly offset by Hujing DME reaching profitability for a third consecutive quarter.

Assessment: Optically ugly but strategically clean. Shedding physical-retail assets sharpens the portfolio toward AI + Cloud and marketplace, and the reported revenue decline overstates any real weakness. The rising investment loss here is where consumer-AI bets (Qwen app, Quark) are being incubated and should be watched as a future cost line.

Key KPIs

Key Topics & Management Commentary

Overall Management Tone: Management was confident and unapologetic, treating the profit collapse as the deliberate cost of a two-front investment cycle in AI + Cloud and consumption rather than something to defend. The posture leaned forward on cloud demand and quick-commerce unit-economics progress, and the one place it was noticeably non-committal was on when profitability inflects and how much capex ultimately runs.

1. Cloud demand is supply-constrained, not demand-constrained

The single most important message of the call was that Alibaba cannot build capacity fast enough to meet AI demand. Wu was emphatic that order growth is outrunning the pace at which new servers can be deployed, reframing the cloud story from "will demand materialize" to "can we supply it."

"We're not even able to keep pace with the growth in customer demand ... in terms of the pace at which we can deploy new servers. So we certainly do see the demand for AI is accelerating." — Eddie Wu, CEO

Assessment: If accurate, this is the most bullish possible framing, because it implies the +34% growth rate is a floor set by supply rather than a ceiling set by demand. It also implies the capex number is going higher, which is bullish for growth and bearish for near-term free cash flow simultaneously.

2. Capex: RMB380B "might be on the small side"

Asked about the three-year RMB380B capex plan (of which ~RMB120B has been spent in the trailing four quarters), management signaled the figure is likely to rise given demand and would only be held back by supply-chain availability of chips, DRAM and other components.

"The RMB380 billion figure we had mentioned previously might be on the small side, certainly in terms of the customer demand that we're currently seeing ... we wouldn't rule out further scaling up that capex." — Alibaba management

Assessment: This is the crux of the investment debate. A rising capex plan against a supply-constrained, share-leading cloud franchise is defensible capital allocation. But it also means free cash flow stays pressured for years, and it exposes Alibaba to the same "AI capex digestion" risk the market is applying to US hyperscalers. Investors are being asked to fund the build now and trust the returns later.

3. Quick commerce: unit economics improving faster than the headline loss suggests

The e-commerce head laid out concrete progress: per-order UE loss cut roughly in half since July/August, non-beverage orders now over 75% of the mix, double-digit AOV growth, and logistics cost per order now below pre-investment levels. The strategic frame is a RMB1 trillion GMV target within three years.

"Since November, the per order UE loss for quick commerce has been cut by 50% compared to July, August ... quick commerce is a core strategic pillar in the Taobao Tmall Group's platform upgrade. Our goal is to generate RMB 1 trillion in GMV for the platform within 3 years." — Jiang Fan, CEO of Alibaba E-commerce Business Group

Assessment: The direction of travel is right and the disclosures were specific enough to be credible. The bull case is that this is a Meituan-style compounding local-commerce franchise being bought at the trough of its loss curve. The bear case is that the RMB1T GMV target implies sustained subsidy intensity in a three-way war where JD.com and Meituan get a vote.

4. The margin collapse is concentrated in one line: sales & marketing

The mechanics of the profit decline are unusually clean. Cost of revenue as a percent of sales was essentially flat (60.8% vs 60.9%) and G&A fell; the entire swing sits in sales & marketing, which doubled to 26.8% of revenue (RMB66.5B) on quick-commerce subsidies and user-experience investment.

Assessment: This is reassuring for a mean-reversion thesis. Because the damage is subsidy-driven rather than structural (gross margin held, overhead fell), profitability can snap back quickly if and when Alibaba dials down subsidy intensity. It also means quarterly EPS will be extremely sensitive to the competitive cadence, so the near-term numbers are low-visibility.

5. CMR resilience and the coming base effect

CMR grew 10% to RMB78.9B on take-rate gains from Quanzhantui penetration and the software service fee introduced in September 2024. The CFO flagged that the lapping of that fee will slow CMR growth beginning next quarter.

"We started charging the payment processing fee in September of last year. And so starting from next quarter, we did expect to see a slowdown in growth due to that base effect." — Toby Xu, CFO

Assessment: A helpful pre-warning that should temper expectations for the December quarter's CMR line. The underlying take-rate story is intact; investors just should not extrapolate 10% CMR growth into next quarter.

6. The consumer-AI push: Qwen app as an ecosystem entry point

Alibaba launched the Qwen app, which surpassed 10 million downloads in its first week, positioning it as a native AI-first assistant that will progressively integrate e-commerce, maps and local services. Management framed the synergy between a leading model and Alibaba's commerce ecosystem as a structural advantage no other China player has.

"Alibaba is the only company in China with both a leading large model and extensive lifestyle and commerce use cases. Qwen will gradually integrate e-commerce, map navigation, local services and more becoming an AI-powered entry point for everyday life." — Eddie Wu, CEO

Assessment: Strategically logical and potentially very valuable, but early and unquantified. Consumer-AI user acquisition is another future cost line (it lives in All others today), and the payoff is a multi-year bet on Alibaba winning the China super-assistant race. Optionality, not a modeling input yet.

7. AI investment priorities: models first, then Bailian inference

On how it allocates scarce compute, management was unusually explicit: the top priority is training its own foundation models, followed by inference-as-a-service on the Bailian platform, then internal use cases, then external GPU rental, with full-stack cloud customers prioritized over simple GPU renters.

Assessment: A rational allocation framework that favors long-term platform value over near-term GPU-rental revenue. It signals discipline in how the capex is being spent, which is exactly what a skeptic of AI capex should want to hear, even if it defers monetization.

8. Free cash flow and the balance sheet

Operating cash flow fell 68% to RMB10.1B and free cash flow swung to a RMB21.8B outflow on quick-commerce investment and cloud capex. Management leaned on the balance sheet, ~US$41B of net cash, as the license to keep reinvesting.

"We are reinvesting our free cash flow to create a winning quick commerce business and to be a leader in AI. Our strong balance sheet backed by USD 41 billion in net cash gives us confidence for this reinvestment strategy." — Toby Xu, CFO

Assessment: The balance sheet is a genuine asset and removes solvency risk from the equation. But negative FCF plus a rising capex plan plus continued dividends and buybacks (US$253M repurchased this quarter, US$19.1B remaining authorization) means the cash pile will be drawn down, and the market will eventually want to see FCF inflect before re-rating the stock on cash returns.

9. AIDC's turn to profit as a proof point

The international segment's swing to a small EBITA profit was cited as evidence that Alibaba's loss-making units can be turned when management prioritizes efficiency, via AliExpress Choice unit-economics gains and logistics optimization.

Assessment: A useful credibility anchor. It supports management's claim that the quick-commerce losses are a deliberate, temporary investment rather than structural, because they have just demonstrated the same playbook working in international commerce.

10. The "AI bubble" question

Pressed on ROIC and whether AI is overbuilt, Wu argued scaling laws have not broken, that a global component shortage (fabs, DRAM, CPUs) will keep AI supply tight for two to three years, and that even three-to-five-year-old GPUs are still running at full capacity, concluding he does not see a bubble in the next three years.

"Looking ahead to the next, say, 3 years, we don't really see much of an issue in terms of a so-called AI bubble." — Eddie Wu, CEO

Assessment: A confident answer that will read as either conviction or complacency depending on one's priors. The observation that even old GPUs run at full utilization is a strong empirical point in Alibaba's favor. Still, "no bubble for three years" is a demand call on a market that has repeatedly surprised, and it is the risk embedded in the rising capex plan.

11. Broader consumption: integrate, don't just acquire

Beyond quick commerce, management pointed to Freshippo, Fliggy, Amap and local services as the consumption matrix, emphasizing that the next phase is driving synergy across existing assets rather than entering new categories. Amap reaching 360M DAU and its new Street Stars feature were cited as engagement proof points.

Assessment: A sensible message of consolidation over expansion, which should limit the risk of new loss-making adventures. The synergy thesis (quick commerce driving MAU that drives CMR that drives ad load) is coherent, but it is a compounding-flywheel argument that needs several quarters of data to validate.

Guidance & Outlook

Alibaba does not issue formal numeric guidance, but management gave enough forward framing to shape a model. The two load-bearing statements are that the September quarter was the peak quick-commerce investment quarter, with investment scale expected to "size down significantly" from the December quarter, and that cloud capex is likely to rise above the RMB380B three-year plan.

Implied trajectory: If the December quarter does see quick-commerce investment "size down significantly," group adjusted EBITA should begin to recover sequentially even as CMR growth slows on the base effect. The swing factor is entirely the competitive intensity of instant commerce; management explicitly reserved the right to adjust spend dynamically to market conditions.

Our read on guidance style: Management is guiding expectations down on near-term profit (the "profitability is expected to fluctuate" language) while guiding conviction up on cloud and long-term consumption share. That is the right posture if they believe the payoff is real, but it also transfers all the uncertainty onto the investor.

Analyst Q&A Highlights

Cloud growth durability and where demand comes from

The opening question probed whether the cloud reacceleration can continue and, absent a US-style base of large frontier-AI labs, what actually drives external revenue growth in China. Management's answer reframed the constraint as supply rather than demand.

Q: "My question is related to cloud business. How should we look at the growth outlook going forward? Should we continue to expect growth to accelerate? And on the demand side, given we don't have a big AI company like in the U.S., how should we look at the key drivers driving the external revenue growth going forward?"
— Gary Yu, Morgan Stanley

A: "We're not even able to keep pace with the growth in customer demand ... it's really coming from all aspects of enterprise operations as AI adoption continues to not only accelerate but deepen ... our conviction in future AI demand growth is strong."
— Eddie Wu, CEO

Assessment: The exchange established the call's central bullish claim, that cloud growth is capacity-limited. It answered the durability question directly and credibly, and it implicitly told the market to expect the capex number to rise.

Quick-commerce progress and the December-quarter e-commerce outlook

A recurring line of questioning pushed on how far quick-commerce economics have improved and what that means for near-term CMR and segment EBITA. Management gave specific unit-economics metrics and, unusually, a direct steer that the peak investment quarter has passed.

Q: "Can management share some key progress for quick commerce and synergy to our core e-commerce so far? Given the synergy, what's the outlook for December quarter CMR and EBITA for our core e-commerce?"
— Kenneth Fong, UBS

A: "Likely, the September quarter will be the quarter during which the scale of those investments are the highest ... by next quarter a significant sizing down in the scale of those investments ... starting from next quarter, we did expect to see a slowdown in [CMR] growth due to that base effect."
— Toby Xu, CFO

Assessment: The most valuable answer of the call for modeling purposes. It sets up a December quarter of narrowing losses but slowing CMR, a mix that could read as "profit recovery" or "growth deceleration" depending on which line the market anchors to.

Where the unit-economics savings get allocated

One line of questioning drilled into the mechanics of UE improvement: as costs come down, how are the savings split among consumers, merchants and the platform, and how much further can UE improve without cutting consumer subsidies?

Q: "As the efficiency is optimized and we obtain cost savings, what are we going to do with those cost savings? How will the benefits ... be allocated or distributed across the value chain among the different key stakeholders ... what does that mean for UE? And how much scope is there going forward for UE growth?"
— Alex Yao, JPMorgan

A: "Our revenues are proportionate to average order value ... over the past few months it's really been primarily new consumers ... what we're doing is converting those users into users with a higher level of stickiness ... there's considerable potential for further monetization."
— Jiang Fan, CEO of Alibaba E-commerce Business Group

Assessment: Management sees the UE improvement path running through higher AOV and monetization of a rapidly growing (100M+ daily) quick-commerce user base rather than through cutting subsidies. That is the more durable path if it works, but it leaves the timeline to profitability dependent on user-behavior conversion that has not yet been proven at scale.

The RMB380B capex plan and the capex-to-revenue relationship

A pointed question asked management to quantify the correlation between capex and incremental revenue, given ~RMB120B already spent against the RMB380B three-year plan. Management declined to give a ratio and instead signaled the plan is likely too small.

Q: "I'm wondering what your thinking is ... regarding the RMB380 billion figure ... over the past 4 quarters RMB120 billion has already been spent. How should we be thinking about CapEx going forward and the incremental revenue being driven by that CapEx?"
— Ronald Keung, Goldman Sachs

A: "The RMB380 billion figure ... might be on the small side ... I don't think it's really possible to make that kind of estimate at least for the time being because overall, the AI sector is still in the early phases ... what we care about more is that our infrastructure is serving high-quality tokens and providing good cost effectiveness."
— Alibaba management

Assessment: The refusal to provide a capex-to-revenue ratio is honest but is precisely the disclosure gap that will keep valuation-sensitive investors on the sidelines. "Trust us, demand is huge" works while cloud accelerates; it will be tested the first quarter cloud growth slips.

AI ROIC and allocating scarce compute

A follow-up asked how to think about return on invested capital across the AI stack (models, infrastructure, consumer apps) amid supply instability. Management answered with its internal prioritization framework and a defense of the demand outlook.

Q: "How should we evaluate the return on invested capital, ROIC, in respect of AI, and including both training and inferencing?"
— Ellie Jiang, Macquarie

A: "The most critical ... first thing that we need to ensure is that we are able to continually train our own foundation models ... Another priority is around inference ... on Bailian ... So looking ahead to the next 3 years, we don't really see much of an issue in terms of a so-called AI bubble."
— Eddie Wu, CEO

Assessment: The prioritization framework (models > Bailian inference > internal > external rental) is disciplined and reassuring on how the money is spent. The "no bubble for three years" conclusion is the confidence that underwrites the rising capex, and it is the single assumption on which the bull case most depends.

Consumption opportunities beyond quick commerce

The closing question asked where else in China consumption Alibaba sees investable opportunities beyond instant commerce. Management pointed to integration of existing assets rather than new category entry.

Q: "Apart from quick commerce, apart from Instant Commerce, what are the other subsectors in the consumption market that you see as good opportunities for investment where you will consider scaling up your investments?"
— Jialong Shi, Nomura

A: "Apart from quick commerce ... we also have Freshippo ... Fliggy ... Amap and local services ... what we need to be doing now really is working to integrate, connect those businesses and to drive more synergies across those existing businesses."
— Jiang Fan, CEO of Alibaba E-commerce Business Group

Assessment: A welcome signal of restraint. Management is prioritizing synergy over new loss-making land grabs, which limits the risk of the investment cycle broadening further. It reinforces that quick commerce and AI are the two, and only two, big bets.

What They're NOT Saying

1. No timeline to group profitability inflection. Management said quick-commerce investment sizes down from the December quarter but never committed to when group adjusted EBITA or FCF turns back up. That omission is deliberate and is the reason the stock cannot be underwritten on near-term earnings.
2. No hard capex number. "RMB380B might be small" is a direction, not a figure. Refusing to re-anchor the three-year capex plan leaves the FCF trajectory unmodelable, which is a problem for a stock the market wants to value on cash generation.
3. No capex-to-revenue return metric. Management explicitly declined to quantify the return on AI investment, resting the case on demand anecdotes. Fine while cloud accelerates; a vulnerability if it decelerates.
4. Quick-commerce absolute loss undisclosed. We got the per-order loss trend (down ~50%) and that segment EBITA ex-quick-commerce would have grown mid-single-digits, but not the absolute quarterly loss, which makes it hard to size the drag independently.
5. Muted on competition specifics. Management referenced "intense competition" but said little concrete about JD.com and Meituan's subsidy response, the variable that most determines how long the loss phase lasts.

Market Reaction

• Pre-print setup: BABA closed at $160.73 on Nov 24, up 89.6% year-to-date and 87.8% over the trailing twelve months, but down 8.0% over the trailing 30 days (off an October high near $174.70). The 52-week closing range was $80.53-$189.34; the S&P 500 was +14.0% YTD.
• Reaction (Nov 25 session, before-open print): Shares gapped up 2.8% to open at $165.20 (traded up ~4% pre-market on the revenue/cloud beat), then reversed across the session to close at $157.01, down 2.3% ($160.73 to $157.01). Intraday range $156.15-$166.37. Volume was 27.1M shares, 1.9x the 30-day average.

The intraday round-trip tells the story better than the close. The market's first instinct was to reward the cloud reacceleration and the revenue beat; its second, on digesting a 78% EBITA collapse, a 71% non-GAAP EPS miss, and a rising capex plan, was to sell. After a near-doubling year-to-date, positioning was crowded and the bar was high, so an "everything the bulls wanted on cloud, everything the bears feared on spending" print produced a classic sell-the-news fade. The reaction is about valuation and expectations, not about the business trajectory, which is precisely why we are initiating at Hold rather than Underperform.

Street Perspective

Debate: Is the profit collapse a red flag or a buying signal?

Bull view: The margin damage is entirely subsidy-and-capex driven, gross margin held and overhead fell, so profitability can snap back the moment Alibaba throttles quick-commerce spend; the "real" business grew revenue 15% like-for-like with cloud accelerating.

Bear view: A 78% EBITA decline and negative FCF are not abstractions; management refuses to timeline the recovery, subsidies could persist for years in a three-way war, and "temporary" investment cycles in China internet have a habit of becoming permanent.

Our take: The bulls are right on the mechanics, gross margin and overhead confirm this is spend, not decay, but the bears are right that the absence of a profitability timeline makes the snap-back unownable on faith. This is why the stock is a Hold and not a Buy today.

Debate: Does cloud alone justify the multiple?

Bull view: A share-leading, accelerating, full-stack AI cloud with a top-tier proprietary model is a scarce China asset; on a sum-of-the-parts basis the cloud franchise could be worth a large fraction of the current market cap, with the marketplace and net cash thrown in cheaply.

Bear view: Cloud is still only ~16% of revenue at a ~9% segment margin; it is not yet big or profitable enough to carry group earnings, and the market is paying a re-rated multiple today for a payoff that is years out and capex-hungry.

Our take: The SOTP logic is sound and is the core of the long-term bull case, but "scarce asset" arguments work best before a 90% run, not after one. Cloud justifies owning Alibaba over a multi-year horizon; it does not by itself justify chasing the stock at these levels into a negative-FCF year.

Debate: How long does the instant-commerce war last?

Bull view: Unit economics are already improving fast (per-order loss halved), Alibaba has the deepest ecosystem to cross-sell into, and a RMB1T GMV franchise built at the trough of its loss curve is a generational local-commerce prize.

Bear view: JD.com and Meituan are equally committed and better-established in delivery; a war of attrition among three well-capitalized players can persist far longer than any of them forecasts, and Alibaba is the incumbent with the most marketplace profit to lose.

Our take: The prize is real and Alibaba's execution so far is credible, but the duration of the loss phase is a function of competitor behavior Alibaba does not control. We want two or three more quarters of UE data before underwriting the RMB1T ambition.

Model Update Needed

Valuation impact: Higher cloud growth is a positive to long-term value; sharply lower near-term EBITA and negative FCF are a negative to any cash-flow-based framework in FY26. The net effect pushes value out in time, raising the weight of terminal cloud economics and lowering the weight of FY26-27 earnings. We would frame fair value on a sum-of-the-parts basis (cloud + marketplace + net cash + stakes) rather than a near-term multiple, and note the stock now embeds a meaningful portion of the cloud optionality after its year-to-date run.

Thesis Scorecard Post-Earnings

As initiation, this establishes the standing thesis we will track each quarter.

Overall: The long-term thesis (cloud/AI + marketplace + optionality, backed by a fortress balance sheet) is intact and in places strengthened; the near-term thesis (earnings and cash flow) is impaired by design. The two are in tension, which is the definition of a Hold.

Action: Initiate at Hold. Own it for the cloud/AI franchise over a multi-year horizon, but wait for either a better entry price or hard evidence of the quick-commerce loss narrowing on schedule (and FCF troughing) before upgrading. We would move to Outperform on two consecutive quarters of quick-commerce UE improvement with cloud holding above 30%; we would move to Underperform if cloud decelerates while losses persist.