AI Frenzy: Is BABA 'Bleeding'?

$Alibaba(BABA.US) released its FQ4 FY26 results pre‑mkt on May 13. As widely expected, the absolute print was weak, with group profit dragged below even the peak loss quarter during the food delivery war last year due to heavy holiday 'cash handouts' around CNY. Cloud delivered no standout surprise to offset the shortfall, but versus updated sell‑side models the print was within range, albeit near the low end.

In detail: Below are the major takeaways.

1) CMR looks soft, but has likely bottomed: The core far‑field e‑comm metric, CMR rose 1.2% YoY, which looks weak at first glance. The key driver was a change in accounting, as merchant subsidies shifted from marketing opex to a contra‑revenue line, and stripping this effect, underlying CMR growth was Approx. 8%, a clear improvement vs. last quarter.

In line with JD’s print and NBS data, the trough for e‑commerce growth appears to be behind us.

2) On‑demand retail continues to narrow losses, on expected track: Taobao Flash’s revenue was near 20 bn this quarter, slightly down from 20.8 bn QoQ, while our take on delivery suggests Flash losses narrowed to ~17 bn this quarter vs. 23‑24 bn last quarter. In other words, Flash is seeing higher revenue per order on mix improvements, while loss per order is trending lower.

From an expectations lens, China E‑comm Group adj. EBITA was 24 bn, down 15.7 bn YoY, near the low end of the sell‑side range. Assuming Taotian ex‑on‑demand has stabilized with flat to low single‑digit YoY profit growth, Flash losses land mid‑to‑upper within sell‑side ranges, so the pace of loss reduction is not above expectations.

3) Alibaba Cloud keeps accelerating, but no upside surprise: The most watched cloud unit grew revenue 38% this quarter, a modest sequential step‑up. Top brokers were looking for ~40%, so the print was a touch light. External revenue rose 40% vs. 35% last quarter, while internal revenue growth slowed further to ~33%, indicating more capacity was prioritized for monetized external demand over internal usage and R&D.

A small positive: adj. EBITA margin reached 9.1%, up 10 bps QoQ, alleviating concerns of deterioration. Also, management noted AI‑related revenue reached ~9 bn this quarter (annualized 36 bn), about one‑fifth of cloud revenue, and is still growing at triple‑digit rates.

4) Capex guide raised: Cash capex was 26.6 bn this quarter, back on an uptrend, albeit not by much. Management also guided that capex will exceed the prior 3‑yr ~380 bn plan, and part of future compute build will be structured as opex (e.g., leases), not fully reflected in capex.

5) Intl e‑comm back to near breakeven: Intl e‑comm revenue grew 5.5% YoY, improving from ~4% last quarter. Adj. EBITA loss narrowed sharply to 0.14 bn vs. a 2.0 bn loss last quarter, better than expected loss reduction.

This confirms the segment is still in an efficiency‑driven, fine‑tuned ops phase, and last quarter’s large loss was a seasonal blip.

6) New biz posted a 20+ bn loss: The not‑so‑'important' Other segment had revenue of ~65.5 bn, down 21% YoY, still mainly reflecting the deconsolidation of Intime and Sun Art.

However, Other segment losses widened to 21.1 bn, above already raised sell‑side expectations of ~20 bn. With Other losses up by 11+ bn QoQ, and on‑demand losses only narrowed by ~7 bn, group profit fell to the lowest in recent years.

The main driver was CNY‑period user acquisition spending for Qwen app red‑envelopes, tie‑in Flash free‑delivery cards, and stepped‑up R&D and user acquisition for multiple AI models and apps.

7) On consolidated results, total revenue was ~243.4 bn (+2.9% YoY). Ex‑Intime and Sun Art divestitures, comparable growth was 11%, vs. 9% last quarter, indicating some reacceleration. Group adj. EBITA was 5.1 bn, down ~84% YoY, a multi‑year low.

On costs and expenses, note that non‑GAAP GPM was 34.7%, down nearly 400 bps YoY, driving GP to 84 bn, down 7.5% YoY, almost 10 bn below Bloomberg consensus.

By contrast, non‑GAAP marketing spend rose by only ~17.5 bn YoY this quarter vs. ~29 bn last quarter. Hence, similar to the CMR reclass moving some subsidies from marketing to contra‑revenue, we think some Other segment spend was treated similarly this quarter.

Also notable, non‑GAAP R&D hit 17.7 bn (+32% YoY), the fastest since FY21, underscoring increased investment in AI apps and model development.

Dolphin Research view:

1) The quarter was clearly not good

In absolute terms, revenue trends improved, but heavy investment crushed profit. Versus expectations, the print was slightly below the sell‑side midpoint, but without an outright bomb.

By segment: a) Ex the CMR reclass, far‑field e‑comm growth and profit have likely bottomed and are recovering, suggesting the worst is past.

b) Flash order mix kept improving and loss per order kept narrowing as expected, but there was no clear upside surprise.

c) Cloud growth kept accelerating and margins ticked up, yet revenue acceleration still lagged market hopes, capacity constraints or not.

d) Taken alone the above would be neutral to slightly positive, but the AI‑driven heavy loss in Other tilted the overall print negative.

2) Outlook:

The past is past; the focus is the road ahead. Below are our views by track.

1) Far‑field e‑comm: Our big‑picture take remains largely unchanged from last quarter; as state subsidies fade in 2026 and AI compute gradually substitutes labor, we stay cautious on 2026 domestic e‑comm growth, with GMV likely in low single digits for the year.

Based on this quarter, Q4 CY25 was the trough and the worst phase is over. That said, Q1’s rebound was mainly CNY‑driven, and online retail weakened again after Mar, so we expect Taotian (ex‑Flash) to deliver low single‑digit positive growth in revenue and profit in FY27, consistent with guidance.

2) On‑demand retail: Two clear trends: Alibaba and Meituan are likely to keep competing on order volumes and share for quite some time, with no dramatic share gap opening, and UE should keep improving, helped by lower platform subsidies, better order mix and higher operating efficiency.

Alibaba previously guided to halving Flash losses in FY27‑28 each year. With FY26 losses near 90 bn, that implies ~40+ bn in FY27, ~20+ bn in FY28, and profitability in FY29, and management essentially reaffirmed this after the quarter.

On one hand, synchronized loss narrowing is a positive for all platforms. On the other, this assumes share stays broadly stable, and we think Alibaba remains determined to maintain order scale and stay close to Meituan. If any party pushes for a step‑change in share, subsidies and losses could flare up again.

3) Chips + Cloud + Models: Despite being among the strongest integrated AI players domestically and the one most able to mirror Google’s story, Alibaba’s recent stock performance has been underwhelming. Beyond broader e‑comm drag, the key gap is that its base model leadership is not clearly ahead of the pack, and while it has in‑house chips, limited disclosure and lack of external sales make it hard for the market to ascribe standalone value.

Overseas, the two‑plus‑one players at SOTA level—OpenAI, Anthropic and Google—have delivered strong revenue and valuation gains. Domestically, it is still a melee; Qwen is top‑tier, but not clearly ahead of GLM, Kimi, DeepSeek and other unicorns, and may even lag slightly.

Thus, while Alibaba Cloud has scale, it lacks a 'must‑have' unique edge versus other clouds and model vendors. As a result, revenue acceleration is unlikely to match Anthropic or Google. Hence, a key watch item for the Alibaba trade is whether it can launch a base model that can sustainably reach domestic SOTA.

That said, near‑ to mid‑term cloud is not without positives. With severe supply‑demand tightness pushing up average compute rental prices, management expects cloud margins to break above 10% from 9% this quarter, while keeping its long‑term 20% margin target.

4) 2C vs. 2B is not either‑or

With Agent‑style products like OpenClaw and Claude Code gaining traction, the frontier has shifted from C‑side traffic gateways to B‑side productivity tools.

We see two different logics: the former follows a mobile‑internet route where AI is a traffic entry point and monetization is indirect via ads/e‑comm, while the latter is software‑like, monetizing value created by the tool itself. Consequently, sectors 'at risk' shift from search/ads/e‑comm to software/SaaS.

While B‑side direct monetization is easier near term than C‑side, both 2B and 2C models should coexist and prove out long term.

OpenAI’s 2C monetization has been bumpy not only because the path is harder, but also because as a pure model startup it lacks vertical expertise and fulfillment capabilities like e‑comm or travel. For Alibaba, with large existing C‑side franchises, AI should run on both 2C and 2B tracks, with only the near‑term emphasis shifting.

Practically, Alibaba should keep investing in C‑facing Qwen and B‑facing tools like Wukong and Qwen Code in parallel. 2C apps play defense, bolstering the retail moat; despite this quarter’s heavy Qwen red‑envelope losses, MAUs rose sharply to 160 mn+, second only to Doubao, which is not nothing.

B‑side monetization is the offense, adding incremental revenue and profit for AI and cloud now. The trade‑off is sustained high AI spend, though not to the point of a one‑off 20 bn quarterly hit from holiday handouts, and Other segment losses in FY27 will likely remain sizable.

5) Funding pressure is notable; spend mix to be optimized: Due to Other investments, operating cash inflow shrank to just 9.4 bn this quarter, and with heavy capex, FCF turned negative at -17.3 bn.

Looking ahead, Taotian (ex‑on‑demand) can generate ~200 bn of adj. EBITA per year, viewed as a proxy for cash flow. As on‑demand losses fall from ~90 bn by half, more capital can shift to AI, which is a positive.

Without drawing on cash on hand, roughly 150 bn p.a. could be deployed into AI, which should cover current capex and spending on Qwen and other AI apps.

3) Valuation: Using SOTP, for China E‑comm we assume low single‑digit YoY adj. EBITA growth for legacy Taotian (ex‑Flash and Fliggy) in FY27 given a muted outlook in CY26. We also assume ~43 bn full‑year on‑demand losses, or ~120 bn profit after tax for the combined unit, and assign a relatively high 12x PE, implying ~$74/sh.

Ex‑on‑demand, Taotian implies ~7.7x PE, near the upper end of the sector’s 6‑8x range. For Alibaba Cloud, given its full‑stack chips+cloud+software capabilities, on ~220 bn FY27 revenue (+40% YoY), we assign 5x PS; with a long‑term 20% margin target, this equates to 25x PE and ~$67/sh.

For Intl e‑comm, growth is still slow but loss reduction beat; we cut FY27 revenue to ~150 bn, keep 1x PS, implying ~$7/sh.

Summing the three core pillars yields ~$151/sh, our neutral case. In a conservative case, we apply 7x to legacy Taotian + on‑demand losses combined and 4x PS to cloud, implying ~$115/sh.

As long as far‑field e‑comm does not backslide, and Meituan does not use Alibaba’s subsidy fade to grab share and reignite subsidy wars, Alibaba remains a name to watch.

Any positive on cloud, base models or chips could lift the stock, and the odds of such catalysts are not low.

Detailed analysis below:

I. New reporting framework

In FY26, with Ele.me on‑demand retail and Fliggy travel folded into Taotian to form the new China E‑commerce Group, Alibaba again updated its org and reporting. As shown below, the group now has four pillars.

1) China E‑commerce Group: legacy Taotian + Fliggy + Taotian Flash/Ele.me;

2) Legacy Intl e‑comm Group unchanged;

3) Alibaba Cloud Group unchanged;

4) All others, incl. Cainiao, China Local Services, Damai Ent, etc., grouped under Other.

II. Ex reclass effects, e‑comm growth has bottomed

1) Far‑field e‑comm’s core metric, CMR was 73 bn, up 1.2% YoY, a poor print . A key factor was the shift of merchant subsidies from marketing to contra‑revenue, depressing revenue; ex this, comparable CMR growth was Approx. 8%.

As noted last quarter, CY4Q was likely the trough for domestic e‑comm, confirmed by both JD and Alibaba prints.

NBS data also show online physical goods retail growth rebounded from 2.3% in 4Q to 7.5% this quarter. But note the rebound and last quarter’s softness share the same cause—late CNY timing pulled more holiday demand into this quarter.

As CNY tailwinds fade, Mar online physical goods growth slipped back to 2.5%, implying full‑year growth pressure remains elevated.

2) Taobao on‑demand retail revenue was near 20 bn, only slightly down from 20.8 bn QoQ, while we estimate Flash losses narrowed from ~23 bn last quarter to ~17 bn this quarter. Given order volumes fell vs. 4Q, revenue per order is up and loss per order continues to fall, in line with expectations but clearly moving in the right direction.

3) 1P retail (incl. related fulfillment revenue) declined again by nearly 6% YoY, vs. a slight uptick last quarter. With overall e‑comm improving and 1P having no reclass impact, Alibaba is likely still exiting low‑quality 1P businesses.

For 1688.com wholesale, growth slowed further to 2.6%, marking a plateau after 2+ years of 10‑20% growth, as the benefit from higher member add‑on fees likely ran its course.

Overall, despite the CMR reclass drag and weak ex‑Flash growth, China E‑comm Group revenue still grew 6%, a mild sequential improvement.

III. Far‑field profit bottomed; on‑demand loss cut in line

China E‑comm Group adj. EBITA was 24 bn, near the low end of the range and down ~15.7 bn YoY. Since the CMR reclass does not affect profit and underlying growth improved, Taotian ex‑Flash likely went from a YoY decline last quarter to roughly flat or low single‑digit growth this quarter. This backs into a Flash net loss of ~16‑17 bn, near the upper half of sell‑side ranges.

Versus our 23‑24 bn loss estimate last quarter, losses narrowed by ~7 bn. Based on channel checks, loss per order likely improved from just above 3.5 last quarter to below 3 this quarter, better than some UE expectations.

IV. Cloud growth still accelerating but a touch light; capex to exceed prior guide

On AI/cloud, momentum remains positive but slightly below expectations. Alibaba Cloud grew 38% vs. 36.4% last quarter, yet the Street was at ~40%, so a minor miss.

Disclosures show external revenue up 40% vs. 35% last quarter, while internal revenue slowed to ~33%, highlighting prioritization of external monetization over internal usage and R&D.

Profitability: adj. EBITA margin reached 9.1%, up 10 bps QoQ, defying fears of deterioration. With capex ramping into AI and compute rental prices moving up from deflationary trends, plus a richer mix of higher‑margin MaaS vs. bare‑metal, cloud margins look set to trend higher near term rather than fall.

Cash capex was 26.6 bn this quarter, back on an uptrend but not sharply higher.

On the call, management said massive AI demand will push capex above the prior ~380 bn/3‑yr guide (~130 bn/yr). Part of compute build will also be done as opex via leasing and may not show in capex.

V. Intl e‑comm re‑tests breakeven

Intl e‑comm improved: revenue up 5.5% YoY vs. ~4% last quarter, driven by AliExpress, while Lazada still declined YoY.

Adj. EBITA loss narrowed to 0.14 bn vs. 2.0 bn last quarter, beating expectations for loss reduction.

This supports management’s view that last quarter’s loss was a temporary promo‑season effect, and that the unit remains focused on disciplined operations with a push toward breakeven.

VI. Other posted a 20+ bn loss, taking group profit to a new low

Outside the three pillars, including Cainiao, Damai Ent, AutoNavi, Qwen and legacy Other, segment revenue was ~65.5 bn (-21% YoY), still mainly due to Intime and Sun Art deconsolidation.

Other segment loss widened to 21.1 bn, above the ~20 bn that the Street had already raised. As Other losses rose by 11+ bn QoQ, while on‑demand only narrowed by ~7 bn, and with Q1 (calendar) being seasonally soft, group profit hit a multi‑year low.

The market anticipated much of this: Qwen’s red‑envelope campaign and Flash’s free‑delivery cards over CNY (some foreign brokers estimate >5 bn for the latter alone). In addition, multiple model initiatives and B‑side AI apps required R&D and customer acquisition, jointly driving the spike in Other losses.

VII. Growth improving, profit deteriorating

Overall, Alibaba revenue was ~243.4 bn (+2.9% YoY), and ex‑Intime/Sun Art, comparable growth was 11%, up from 9% last quarter.

Group adj. EBITA was 5.1 bn (-~84% YoY), a multi‑year low and below the profit level during last year’s delivery war peak, mainly due to Qwen and other AI investments.

From a cost/expense view, non‑GAAP GPM was 34.7%, down nearly 400 bps YoY. Alongside modest marketing growth, beyond the CMR subsidy reclass to contra‑revenue, we think a sizable portion of Other (e.g., Qwen) spend was booked in cost of revenue or contra‑revenue, not marketing.

As a result, GP was 84 bn (-7.5% YoY), nearly 10 bn below Bloomberg consensus. Non‑GAAP marketing was 53 bn, up ~17.5 bn YoY this quarter vs. ~29 bn last quarter, suggesting part of spend shifted from opex to cost of revenue and impacted GPM.

Outside marketing, non‑GAAP R&D reached 17.7 bn (+32% YoY), a post‑FY21 high, clearly signaling rising AI app and model investment.