BABA: E-com slips again; AI is the only lifeline---

Fighting on two fronts in food delivery and AI, $Alibaba(BABA.US) posted Q3 FY26 results pre-market tonight (Mar 19). From a expectations vs. reality lens, results were broadly in line with the recently lowered guide, except for softer Intl e-com growth and wider losses in 'Others'. In absolute terms, though, the print was weak.

The core e-com engine slowed markedly as the state subsidies tailwind and higher monetization cycle largely ended. The bright spot was Alibaba Cloud, which continued to grow strongly.

Bloomberg consensus was not timely this quarter and has limited reference value. We therefore benchmark against updated major-house estimates below.

1) CMR up less than 1%: The core metric for the traditional marketplace business — CMR rose 0.8% YoY, a sharp decel from ~10% in prior quarters. Similar to JD, the drag came from state subsidies fading in 2025 and a high base in 2024, with a late Lunar New Year also pushing more domestic consumption into Q4 FY26 (calendar Q1 2026).

In addition, the 0.6% service fee and sitewide ad push rolled out since Sep 2024 have lapped their favorable monetization base. The two factors combined drove a rapid slowdown in CMR. The company had communicated this well, so delivery was roughly in line with expectations.

2) Food delivery losses narrowed as expected, but pace underwhelmed: China E-com Group posted adj. EBITA of RMB 34.6bn, down ~RMB 26.5bn YoY. With CMR nearly flat, assuming ex-Flash Sales Taotian saw a slight YoY profit decline, the back-solve suggests Flash Sales net loss was ~RMB 25bn this quarter.

Losses narrowed sharply vs. last quarter (i.e., Flash Sales UE improved), but landed at the upper end of the RMB 20–25bn loss range that the market was braced for. This implies improvement is slower than hoped. By Dolphin Research’s estimate, Taobao Flash Sales loss per order tightened from just over RMB 5 last quarter to ~RMB 3.5 this quarter (non-official, for reference only).

3) Alibaba Cloud re-accelerates, still standout: On the AI front, Alibaba delivered again. Alibaba Cloud revenue grew 36% YoY, continuing a modest acceleration.

It did not beat, and the step-up looks small. But external revenue grew 35% YoY vs. 29% last quarter, a clear re-acceleration; by group mix, faster external growth carries higher quality.

The company also noted token consumption on the Bailian MaaS marketplace rose 6x over the past three months (as the shift from Chatbot to Agent drives much higher token/compute needs). Dolphin Research remains constructive on long-term cloud compute demand.

Meanwhile, Alibaba Cloud’s margin was 9%, flat QoQ. AI mix did not dilute margins, which is a decent outcome.

Capex came in at RMB 29.9bn, down QoQ, likely reflecting the Nvidia chip restriction. Given prior heavy investments and a market increasingly focused on ROI, lower investment may not be a bad thing. Free cash flow turned positive again this quarter.

4) Intl e-com slowed and slipped back into loss: Intl e-com revenue growth slowed to under 4% YoY, below the already-cut ~7% market view, with Lazada’s YoY revenue decline the key drag. In SE Asia, Alibaba faces intense competition from Sea and TT Shop amid their expansion phases.

Adj. EBITA also swung back to a RMB 2bn loss. That said, it is largely seasonal (e.g., Black Friday), and still improved YoY, not a return to a heavy investment cycle. Management will maintain a refined, ROI-driven approach.

5) Other units stepped up spending: This quarter, ‘Others’ loss widened sharply to ~RMB 9.8bn, well above major houses’ already-lowered ~RMB 8.5bn loss view. Higher investment in Gaode’s street-ranking push and AI apps like Qianwen/Quark for user acquisition and development makes the widening understandable, though the magnitude was large.

Given Q4 (Dec qtr) saw limited Qianwen push and there were red-envelope and free-order campaigns around CNY, next quarter’s ‘Others’ loss may also stay elevated.

6) At the consolidated level, total revenue rose ~1.7% YoY, or ~9% on a comparable basis excluding YinTai and Sun Art vs. 15% last quarter. The slowdown mainly reflects materially softer domestic and overseas e-com.

Gross profit, adding back SBC, fell 2% YoY, with GPM down 150bps YoY, with a wider decline. Instant retail delivery costs were likely the main drag, with ‘Others’ losses also weighing.

On opex, sales and marketing hit RMB 70.9bn, up ~RMB 28.8bn YoY, exceeding the food delivery losses. This clearly signals heavier user-acquisition spend across other segments.

Group adj. EBITA was RMB 23.4bn. Loss narrowing in food delivery drove a notable improvement vs. sub-RMB 10bn last quarter. But beyond that, core Taotian profit likely fell YoY, Intl e-com turned loss again, and ‘Others’ loss widened more than expected, leaving overall profit trends weak.

Dolphin Research view:

1) On the quarter, Alibaba’s print was clearly soft.

1) The core marketplace business stalled, as sector-wide Q4 spending was weak and the monetization uplift cycle neared its end. Revenue growth for the most important cash engine flatlined, and profit naturally followed.

Taobao Flash Sales is narrowing losses, but progress looks slow.

2) At the same time, Intl e-com slowed and, despite seasonal factors, posted another single-quarter loss of over RMB 2bn. This suggests limited benefits from refined operations, with neither growth nor profit delivery standing out.

Meanwhile, with Taotian’s cash-generation weakening, other lines re-entered an investment phase. Even as Flash Sales losses narrow, group profit still trended down.

3) The only real support came from AI and cloud. As a domestic leader, Alibaba Cloud’s growth and margin were solid. But unlike last quarter, there was no major beat to overshadow weaknesses elsewhere.

2) Outlook:

1) Marketplace (far-field) e-com: The sector’s ‘ballast’ will likely face a smaller state-subsidy push in 2026 than in 2025, alongside tough comps. In addition, the emerging theme of AI compute replacing labor could weigh on spending by affected workers, leaving Dolphin Research cautious on 2026 domestic e-com growth.

That said, one reason Q4 e-com was so weak was the late CNY timing that delayed consumption. Online goods sales grew over 10% YoY in Jan–Feb per official data, a clear rebound; Q4 CY25 + Jan–Feb CY26 combined online retail rose ~5.4% YoY, slower but not terrible.

Putting it together, Dolphin Research is cautious on full-year 2026 e-com growth, but Q4 CY25 likely marked the trough, and the worst may be behind us.

Also, the 0.6% fee uplift and sitewide ad tools tailwind has largely run its course. And since Oct 2025, tax enforcement for merchants tightened (from self-filing to platform-facilitated disclosures), which implies higher tax/profit pressure, especially for SMEs. This could reduce SMEs’ ability to pay for ads.

Both suggest Taotian’s path to revenue growth via higher monetization will be tougher.

2) Instant retail: While Alibaba and Meituan have dialed back subsidies vs. Q3 CY25, investor focus has shifted from the ‘delivery war’ to the ‘AI war’. Still, order volumes and subsidy intensity remain closely matched.

Management has stated ‘the goal in instant retail is No.1 by scale’, implying competition and subsidy normalization will not end soon. This quarter’s loss at the upper end for Flash Sales validates that view.

Though research suggests Taobao Flash Sales’ order mix is improving (higher-ticket meals and non-meal retail), this quarter’s data did not yet reflect that.

Based on recent checks (for reference), CY26 Taobao Flash Sales budget is RMB 50bn + 20bn, with RMB 50bn fixed and RMB 20bn flexible. Given current trends, spend could land near the upper bound.

If it ends up at RMB 70bn, that would be similar to CY25 in aggregate. But since CY25 spend was spread over ~2.5 quarters, CY26 quarterly run-rate spend and losses should still decline.

3) Chips + Cloud + Models: the best local AI stack story: Similar to GOOG, Alibaba’s Pingtouge (in-house chips), Alibaba Cloud (IaaS/PaaS), and Qianwen models & app family make it one of China’s most complete AI stacks. Within large-caps, Alibaba remains a core domestic AI holding.

This mid-term AI narrative should hold despite a weak quarter. Key recent developments follow.

First, in the AI consumer entry battle, Doubao, Qianwen and Yuanbao competed in CNY ‘red-envelope’ campaigns. Retention was poor, with DAUs dropping from peaks for most apps except Doubao.

Qianwen fared better than Yuanbao, but the gap vs. Doubao widened. So on the C-end entry point, Alibaba achieved some results but with limits.

As OpenClaw spread domestically, the frontier shifted from Chatbot traffic acquisition to Agent R&D and user capture. Dolphin Research sees potential impacts:

a) Agents demand multiples of Chatbot compute/tokens, fueling rapid token/compute growth, which benefits compute suppliers. Alibaba Cloud’s recent price hikes on select products likely reflect tight compute supply.

b) With scarce compute, focus could shift from consumer-facing apps like Qianwen/Doubao (initially cost centers offering free or subsidized use for traffic, with monetization still exploratory) toward Agent use cases (more office/2B oriented, higher willingness to pay, and immediate revenue via token billing). This helps cloud/model vendors realize AI revenue and profit sooner.

4) What ATH means:

Alibaba announced a new Alibaba Token Hub (ATH) by bundling Tongyi Lab, Cloud’s MaaS, Qianwen, Wukong, and AI Innovation. This reflects the post-Chatbot, Agent-led shift where token becomes the key growth vector.

Dolphin Research believes the key benefit is aligning upstream model R&D, midstream compute/model sales, and downstream 2C/2B apps within one org, improving tech-product fit and internal coordination. It avoids R&D chasing frontier metrics while apps chase near-term DAUs in isolation.

A second-order but near-term earnings-relevant change: as Cloud moves from selling raw compute (‘bare metal’) (customers self-build) to MaaS/Token (pre-configured models atop infra), revenue and margins should be higher. This could show up as faster Cloud revenue and potentially better-than-expected margins.

3) Valuation: Using SOTP, for China E-com Group, with a cautious CY26 outlook for traditional e-com, we model Taotian ex-Flash & Feizhu adj. EBITA for FY27 up low single-digit percent YoY and assume RMB 70bn instant retail losses for the year, implying ~RMB 110bn after-tax group profit.

As Taotian is among the most resilient in e-com and instant retail has meaningful loss-narrowing potential that can lift China E-com growth, we assign 12x PE, implying ~$80 per share. Ex-food delivery losses, Taotian is ~6x PE, akin to JD and Vipshop at cyclical troughs.

For Alibaba Cloud, leveraging the chips + cloud + software stack, we model ~RMB 220bn revenue in FY27 (+40% YoY). At 4x PS or 40x PE on 10% margin, that equates to ~$53 per share.

For Intl e-com, on ~RMB 156bn revenue (+7% conservative), given weaker momentum and a return to quarterly loss, we cut to 1x PS, or ~$9 per share.

Sum-of-parts yields ~$146 per share as a neutral case. A conservative case, valuing Taotian plus food delivery losses at 8x, gives ~$115 per share.

Downside to the bottom looks limited. Upside likely hinges on Cloud’s growth/margin beats, or clearer listing and external commercialization of Pingtouge, unlocking incremental segment value.

Detailed results analysis below:

I. New reporting structure

From FY26, with Ele.me instant retail and Fliggy folded into Taotian to form China E-com Group, Alibaba updated its org and reporting. The group now has four pillars:

1) China E-com Group: original Taotian + Fliggy + Taotian Flash/Eleme;

2) Intl E-com Group unchanged;

3) Alibaba Cloud Group unchanged;

4) All others, including previously standalone Cainiao, China Local Services, Damai Ent, etc., now in ‘Others’.

II. Core e-com growth still resilient

1) For the core marketplace, CMR was RMB 102.7bn, +0.8% YoY, a sharp slowdown. Similar to JD, the slowdown reflects weaker state-subsidy effects in 2025 and a high 2024 base, plus a late CNY that left Q4 online physical goods retail up just 2.3% YoY, with Taotian GMV a bit below the sector.

Also, the favorable base from the 0.6% fee and sitewide ad tools (since Sep 2024) has rolled off, so CMR’s outperformance vs. GMV narrows. That is another drag.

While absolute trends were weak, the company had reset expectations, and delivery matched that reset.

2) Taobao Flash + Ele.me: instant retail revenue was ~RMB 20.8bn, +56% YoY, but down 9% QoQ, a bit worse than our own expectation, given seasonal order declines (colder weather).

We had expected order-mix upgrades (more high-ticket meals and non-meal retail) to lift revenue per order. That uplift looks smaller than hoped.

3) 1P retail (incl. related fulfillment) grew 0.3% YoY, a marked slowdown similar to CMR. The main driver was broad Q4 domestic e-com softness.

1688.com wholesale, as a key pillar of the ‘value-for-money’ play, grew a relatively faster 5.3%, driven by member add-on services per management. All in, with both CMR and other lines slowing with the market, instant retail was the key offset, and China E-com revenue rose 5.8% YoY.

III. Food delivery loss-narrowing fell short of hopes

While attention shifted from the ‘delivery war’ to AI, Flash Sales spend/loss still has an outsized impact on profits.

China E-com Group adj. EBITA was RMB 34.6bn, down ~RMB 26.5bn YoY. With CMR nearly flat, assuming ex-Flash Sales Taotian profit slightly declined YoY, Flash Sales net loss likely ~RMB 25bn (for reference).

That is a sharp improvement vs. our ~RMB 36bn loss estimate last quarter, confirming better UE. However, at the upper end of the RMB 20–25bn expected range, the pace of improvement seems slower than hoped.

Based on research notes, we estimate Flash Sales loss per order improved from just over RMB 5 to ~RMB 3.5 this quarter (non-official, for reference).

IV. Highlight — Alibaba Cloud continues to re-accelerate

On AI/Cloud, Alibaba shone again. Cloud revenue grew 36% YoY, in line with expectations, modestly faster than last quarter’s 34.5%.

While the headline beat was absent and the step-up small, external revenue grew 35% YoY vs. 29% last quarter, a clear acceleration. External demand is rising meaningfully, and AI-related revenue remains triple-digit growth per the company.

As China’s AI evolves from Chatbot to Agent (multiplying token/compute needs), token consumption on the Bailian MaaS marketplace rose 6x in the past three months. MaaS is expected to become Cloud’s largest product revenue stream, supporting a constructive multi-year compute outlook.

On profit, Cloud adj. EBITA margin was 9%, flat QoQ. Despite higher equipment/energy needs for AI, which theoretically compress margins, tight compute supply (Cloud just announced price hikes on select products) and mix optimization (exiting lower-margin work, emphasizing higher-margin MaaS) helped keep margins stable.

Capex was RMB 29.9bn this quarter; cash capex was RMB 25.3bn, both down QoQ, continuing a FY26 downtrend that may reflect Nvidia chip restrictions.

Given investors now favor ROI over sheer capex scale and prefer limited pressure on earnings and cash flow, slower spend after a more aggressive phase may not be viewed negatively. Watch for capex cadence commentary on the call.

V. Intl e-com slows and turns loss; refined ops intact

Intl e-com underperformed: revenue growth slowed to under 4% YoY, below the trimmed ~7% market view and one of the few misses this quarter. Per the company, Lazada’s YoY revenue decline was the key drag.

With Sea in a ‘spend for growth’ phase, competitive intensity in SE Asia is likely high for Alibaba Intl.

Adj. EBITA swung back to a RMB 2bn loss. However, this is seasonal (e.g., Black Friday), still better than the ~RMB 5bn loss in Q3 FY25. Management reiterated this does not signal a renewed investment cycle and that refined operations continue.

VI. Gaode push & Qianwen/Quark: wider losses in new biz

Outside the three pillars, including Cainiao, Damai Ent, Gaode, Qianwen and other prior ‘Others’, revenue was ~RMB 67.3bn, down 25% YoY, mainly due to the YinTai/Sun Art disposals and some logistics functions moving into front-end units like 1P retail and Intl e-com.

With higher spend on Gaode’s street-ranking, and on user acquisition and development for Qianwen/Quark and other AI apps, ‘Others’ loss widened to ~RMB 9.8bn, worse than the -RMB 8.5bn street view.

Given limited Qianwen push in the Dec qtr and heavy CNY promos (red envelopes/free orders), next quarter ‘Others’ losses may remain high.

VII. Broadly in line with trimmed guide, but weak in absolute

Overall, total revenue was ~RMB 284.8bn, +1.7% YoY, or ~9% ex-YinTai/Sun Art (vs. 15% last quarter), mainly reflecting a broad e-com slowdown onshore and offshore.

Gross profit (ex-SBC) fell 2% YoY, and GPM fell 150bps YoY with a larger decline, mainly dragged by instant retail delivery costs, with wider ‘Others’ losses also weighing.

On opex (ex-SBC), sales and marketing was RMB 70.9bn, up ~RMB 28.8bn YoY. That exceeds the estimated Flash Sales losses, clearly signaling heavier user-acquisition spend in other segments.

Elsewhere, R&D rose nearly 4% YoY, while G&A fell ~25% YoY (likely reflecting YinTai and other disposals). Alibaba remains disciplined on internal costs.

At the group level, adj. EBITA was RMB 23.4bn, down ~43% YoY. Thanks to narrower delivery losses, profit improved materially vs. sub-RMB 10bn last quarter, but core Taotian profit likely fell YoY, Intl e-com turned loss again, and ‘Others’ loss widened more than expected, leaving overall profit trends weak.