Alibaba Group Holding Limited (NYSE: BABA · HKEX: 9988) — The Operating Business at a Discount, with a Free AI-Cloud Option

An independent equity research note Date: June 7, 2026 · Fiscal year referenced: FY2026 (ended March 31, 2026) Reporting: U.S. GAAP in Renminbi (RMB) · ADS ratio: 1 ADS = 8 ordinary shares · Structure: Cayman holding company controlling PRC operating entities via VIE/contractual arrangements Price reference: US$121.06/ADS (2026-06-05) · Market cap: ~US$290B

⚡ Claude’s Take

This block is the author’s own independent opinion and general information only — not investment advice. The body that follows (Executive Summary through Source Appendix) deliberately takes no position, sets no price target, and carries no recommendation; that discipline is intact everywhere except here.

Verdict: ACCUMULATE-ON-WEAKNESS / speculative BUY for investors who can underwrite the China tail — a HOLD at today’s ~US$121, an attractive add in the low-US$100s. Directional fair-value zone ~US$150–190/ADS if the core normalizes and Cloud keeps compounding; an EPV/asset floor of roughly US$95–125/ADS limits fundamental (ex-tail) downside. Position-size for the jurisdiction, not the business.

The market is handing you Alibaba’s operating businesses — the most profitable marketplace in China plus the #1 AI-cloud franchise in Asia — for roughly US$185–210B, because ~US$80–105B of net cash, the 33% Ant stake, and an equity portfolio cover a third of the ~US$290B cap. On normalized numbers the China-commerce core alone is worth about what the market assigns to all the operating segments combined, which means you are paying little-to-nothing for a Cloud business growing 34–40% with AI revenue up triple-digits for eleven straight quarters. The reason it’s cheap is real but, I think, over-extrapolated: FY2026 was a deliberately awful year — management halved core EBITA and turned free cash flow negative (−RMB46.6B) to fund a RMB380B AI build-out and a food-delivery subsidy war that regulators have now forced to a ceasefire. That is a chosen trough, not a broken business. The framing is contrarian value with a free AI-cloud option — Greenwald’s EPV (no-growth earnings + assets) brackets the current price, so you’re buying the downside-protected part and getting the growth for a song.

What keeps this a speculative buy and not a table-pounder is that the cheapness is partly earned: Taobao/Tmall is genuinely ceding GMV share to PDD and Douyin (the moat is narrowing, not impregnable), the asset “floor” is partly non-fungible to a foreign holder (onshore-RMB capital controls, illiquid Ant), and the whole claim sits on a VIE structure and a dormant-but-reinstatable delisting tail that no amount of diligence can eliminate. That tail is exactly why a permanent discount is rational — so the edge is in handicapping reversibility, not in denying the risk. Conviction: medium. Flips bullish if China E-commerce adjusted EBITA recovers toward its prior ~US$25B+ run-rate over the next 4–6 quarters now that the subsidy war is ending, and/or Ant moves toward a re-IPO that crystallizes the stake. Flips bearish if core EBITA stays below ~US$20B normalized despite the ceasefire (proving the margin hit is structural, not cyclical) or Cloud growth decelerates below ~20% as chip controls bite. Tag: “Buying the operating business at a discount, with an AI-cloud call option attached — if you can stomach the jurisdiction.”

1. Executive Summary

Alibaba is a high-quality, cash-generative core wrapped in a sprawling, currently cash-consuming conglomerate, trading at a valuation that brackets its no-growth earnings power plus assets — cheap on arithmetic, and the entire debate is whether that cheapness is a mispricing or a correct discount for impaired realization.

The business. For FY2026 (ended March 31, 2026), Alibaba generated revenue of RMB1,023,670M (US$148.4B), +3% reported but +11% like-for-like (the gap is the deliberate disposal of Sun Art and Intime offline retail). The company is radically profit-concentrated: the Alibaba China E-commerce Group (Taobao/Tmall customer-management revenue + quick commerce + 1688) produced RMB107.5B (US$15.6B) of adjusted EBITA — more than the entire group — while Cloud Intelligence Group contributed a modest-but-growing RMB14.3B, and AIDC (international) and “All Others” lost a combined ~RMB38B. The mature Taobao/Tmall take-rate engine funds the whole enterprise, including the AI build-out and the food-delivery war.

What happened in FY2026 — a chosen trough. Group adjusted EBITA fell 56% (margin 17.4% → 7.5%), operating income fell 64% (margin 14% → 5%), and free cash flow swung to a −RMB46.6B outflow (first negative year in Alibaba’s public life), versus +RMB73.9B prior. Two deliberate decisions drove it: (1) the RMB380B (~US$53B), three-year AI + cloud capex program (capex +47% to RMB126B); and (2) the instant-commerce / food-delivery subsidy war with Meituan and JD, which torched RMB tens of billions and is the main reason China-commerce EBITA halved. GAAP net income (RMB102.1B, −19%) is flattered by RMB87.5B of non-operating equity mark-to-market gains; the cleaner read is non-GAAP net income −62%.

The two franchises that matter. (1) CMR (the Taobao/Tmall ad/commission take) grew +7–8% like-for-like — a genuine re-stabilization, though the moat is narrowing: Taobao/Tmall’s GMV share fell from ~50% (2020) to ~40–44% (2024) as PDD rose to ~19% and Douyin built ~US$650B GMV from zero, and a new merchant-subsidy “contra-revenue” program betrays weakened pricing power. (2) Cloud is the bright spot — revenue +34% (Q4 +38%, external +40%), AI products at 30% of cloud revenue with an 11th straight quarter of triple-digit growth, #1 in China at ~36% share. Cloud sits in a structurally attractive re-accelerating oligopoly; e-commerce sits in a structurally deteriorating, fragmenting, price-warring one.

Balance sheet & capital allocation. Fortress: gross cash and liquid investments RMB520.8B (US$75.5B), net cash ~US$37.8B, plus a 33% Ant stake and a large equity portfolio. But capital allocation is at its highest-risk inflection: buybacks collapsed from ~US$12B/year (FY24–25, each shrinking the float ~5.1%) to just US$1.0B in FY2026, with ~US$19B of authorization unused — management pivoted from returning capital to consuming it, and even issued equity-linked debt while pausing repurchases. The recurring dividend (US$1.05/ADS, ~US$2.5B) continues. Insider economic alignment is low (all directors/officers 1.9%) under an Alibaba-Partnership weighted-voting-rights structure.

Valuation. A net-cash, asset-rich, China-discounted value situation, not a growth-multiple story. The ~US$80–105B non-operating floor plus a no-growth EPV (~US$150–200B) brackets the ~US$290B market cap, so the AI-cloud growth is lightly priced. Forward P/E ~13× on depressed earnings; own-history ~31st percentile (cheapish, not a generational low). The catch: the asset floor is partly non-fungible to a foreign holder (onshore-RMB capital controls, illiquid Ant), and the whole position rests on a VIE structure and a dormant-but-reinstatable HFCAA delisting tail.

The investment question. Not “is it cheap” — it demonstrably is on arithmetic — but “is the cheapness earned.” The deciding variables (is the subsidy-war EBITA hit cyclical or structural; can Cloud scale under US chip controls; is the asset floor monetizable to foreign holders; how dormant is the delisting tail; is China commerce ceding GMV permanently) are all currently unresolvable from the outside and will only clarify over the next several quarters of EBITA recovery — or non-recovery. This memo takes no position on that question; the body that follows is rec-free and price-target-free by design.

2. Business Overview

Alibaba Group Holding Limited is a Cayman Islands holding company that, through a network of PRC-incorporated subsidiaries and variable interest entities (VIEs), operates China’s largest e-commerce ecosystem and its largest public-cloud business, alongside international commerce, logistics, local services, and a fast-growing artificial-intelligence franchise. The company reports under U.S. GAAP in Renminbi; its American Depositary Shares each represent eight ordinary shares (NYSE: BABA), with a primary listing in Hong Kong (HKEX: 9988) under a weighted-voting-rights structure controlled by the Alibaba Partnership (Source: FY2026 20-F). For fiscal year 2026 (ended March 31, 2026), Alibaba generated revenue of RMB1,023,670M (US$148,401M), up 3% reported but +11% on a like-for-like basis excluding the disposed Sun Art and Intime offline-retail businesses, with net income of RMB102,127M (Source: FY2026 earnings PR).

The new four-segment architecture. In Q1 FY2026 management restructured reporting into four segments, which materially aids analysis because it isolates the profit engine from the cash-consuming investments:

(Source: FY2026 earnings PR segment tables. Group totals net inter-segment elimination.)

How each segment makes money — the four distinct economic models.

Alibaba China E-commerce Group (the engine). This combines the Taobao and Tmall marketplaces (“TTG”), the Ele.me/Taobao Instant Commerce quick-commerce business, Fliggy (travel), and the 1688 domestic wholesale platform. Its core economic model is customer management revenue (CMR) — a take-rate business. Taobao and Tmall do not buy and resell inventory on the marketplace; instead they monetize third-party merchants through pay-for-performance marketing (advertising, the Quanzhantui ad tool), commissions, and software-service fees levied as a percentage of gross merchandise value (GMV). FY2026 CMR was RMB343,867M, +5% reported (+7% like-for-like excluding a new “contra-revenue” program described below). A second sub-line, Direct sales, logistics and others (RMB105,518M), is the gross-margin first-party business of Tmall Supermarket / Tmall Global, where Alibaba books full inventory revenue and cost. A third sub-line, Quick commerce (RMB78,520M, +47%), is the on-demand delivery business — economically a subsidy-funded model reported net of contra-revenue subsidies, and the principal cause of FY2026’s margin collapse. China commerce wholesale (1688) adds RMB26,312M of largely subscription/value-added-service revenue. CMR is the high-margin, quasi-recurring core; quick commerce is transactional and currently loss-making by design (Source: FY2026 earnings PR).

A critical accounting wrinkle: in Q4 FY2026 Alibaba “upgraded” its merchant business-development program so that platform subsidies, previously booked in sales-and-marketing expense, are now recorded as contra-revenue to CMR. This depressed reported Q4 CMR growth to +1% versus +8% like-for-like (Source: FY2026 earnings PR). The substance — Alibaba paying merchants to spend more on the platform — is itself a competitive-pressure signal (developed in ).

Cloud Intelligence Group. A classic IaaS/PaaS/MaaS business: compute, storage, networking, databases sold on consumption (public cloud), plus a rapidly growing Model-as-a-Service layer monetizing the Qwen (Tongyi Qianwen) family of large language and multimodal models via Model Studio. FY2026 revenue grew 34% to RMB158,132M with adjusted EBITA of RMB14,265M; growth accelerated through the year to +38% in Q4 (external +40%), with AI-related products reaching ~30% of cloud revenue and posting an eleventh consecutive quarter of triple-digit growth (Source: FY2026 earnings PR). Cloud is the only growth franchise in the group with both scale leadership and positive — if thin (~9% EBITA margin) — economics.

AIDC. AliExpress, Trendyol, Lazada, Daraz, and Alibaba.com — cross-border and Southeast-Asian retail and wholesale. FY2026 revenue +9% to RMB144,170M; the segment narrowed its adjusted-EBITA loss from RMB15,137M to RMB2,051M, approaching breakeven on AliExpress “Choice” unit-economics improvement (Source: FY2026 earnings PR). It is subscale against Amazon, Shopee/Sea, and PDD’s Temu, and has never demonstrated franchise economics.

All Others. Cainiao (logistics), Amap (mapping/local services), Freshippo (grocery), Hujing Digital Media & Entertainment, Alibaba Health, DingTalk, Lingxi Games, and the Qwen consumer app. FY2026 revenue fell 25% (Sun Art/Intime disposals) and the segment’s adjusted-EBITA loss widened to RMB35,737M, driven by user-acquisition spend on the consumer Qwen app and technology investment (Source: FY2026 earnings PR).

Revenue and profit concentration — the central structural fact. Alibaba is not a diversified profit machine; it is one extraordinary profit pool subsidizing several loss-makers. China E-commerce Group’s RMB107,509M of adjusted EBITA exceeds total group adjusted EBITA of RMB76,416M, because AIDC, All Others, and unallocated corporate costs collectively lose ~RMB45B. Cloud contributes a modest RMB14,265M. In effect, the mature Taobao/Tmall take-rate business funds the entire enterprise — including the quick-commerce war, the RMB380B AI/cloud build-out, and the Qwen consumer push. The recurring, defensible core is CMR; nearly everything else is transactional, cyclical, or investment-stage. The ~33% equity stake in Ant Group (Alipay) sits off to the side as an equity-method investment, contributing RMB375M of profit in Q4 FY2026 (one quarter in arrears) and substantial embedded asset value not reflected in operating results (Source: FY2026 earnings PR; alibabagroup.com Ant 33% agreement, 2018).

Verdict. Alibaba is a high-quality, cash-generative marketplace wrapped inside a sprawling, capital-hungry conglomerate. The CMR engine is genuinely excellent; the surrounding portfolio is a mix of one real second franchise (cloud) and a long tail of subscale or subsidy-funded businesses that, in FY2026, turned group free cash flow sharply negative (an outflow of RMB46.6B versus a RMB73.9B inflow the prior year). The business model is sound at the core and undisciplined at the periphery — a structure that demands a sum-of-the-parts lens, never a consolidated multiple.

3. Industry Dynamics

Alibaba operates across three structurally distinct industries — Chinese e-commerce, Chinese public cloud/AI, and international/cross-border commerce — each at a different point in its life cycle and capital cycle. They must be assessed separately.

Chinese e-commerce: the world’s largest market, now mature, fragmenting, and at war on price. China is the largest e-commerce market on earth, with online penetration of retail among the highest globally, but the era of effortless ~20-30% GMV growth ended years ago; the market now grows roughly mid-single to high-single digits and tracks broader (soft) Chinese consumption (Source: trade.gov China eCommerce guide; Statista 2024-25). Three forces define the structure today:

1. Disruption from below (PDD/Pinduoduo). Founded in 2015, Pinduoduo built a hard-discount, manufacturer-direct model that exploded from ~7.2% GMV share (2019) to ~19% (2023), overtaking JD to become the #2 platform and at one point surpassing Alibaba in market capitalization (Source: SCMP; CNBC, May 2024). PDD reset the price expectation of the entire market.
2. Disruption from the side (content commerce). Douyin (China’s TikTok) and Kuaishou converted short-video/livestream attention into transactions, building an estimated ~US$650B GMV from a standing start (Source: industry estimates via WebSearch, accessed 2026-06-07). This is demand-generation Alibaba structurally cannot replicate — Alibaba is a search-and-fulfillment destination; Douyin is impulse-driven discovery.
3. Incumbent erosion. Taobao/Tmall’s GMV share fell from roughly half the market in 2020 to ~40-44% by 2024 (Source: Statista; AIM Group 2024). Alibaba’s GMV growth has at times lagged the industry (e.g., +2.5% vs ~+6% market in a 2024 quarter), the textbook signature of a share-donor incumbent (Source: AIM Group, 2024).

Profit pools have compressed accordingly. Because Taobao/Tmall historically under-monetized GMV relative to PDD and Douyin (lower effective take-rate), Alibaba retains a monetization-gap opportunity (raise take-rate toward peers) — but it is pursuing that into a market where merchants now multi-home and where the firm is simultaneously subsidizing merchants (the contra-revenue program) to defend volume. The structural attractiveness of Chinese e-commerce has unambiguously declined: more credible competitors, lower barriers (merchants face no exclusivity post-2021), and a chronic low-price equilibrium.

The 2025 “instant retail” / food-delivery war — a prisoner’s dilemma in real time. The most important recent industry development is the all-out instant-commerce battle among Alibaba, Meituan, and JD. Each pledged RMB10B-50B subsidy programs; Meituan hit ~150M daily orders (July 2025) and Taobao Instant Commerce surpassed ~80M daily orders within months of launch (Source: thebridgechronicle.com; gabgrowth.com, 2025). Sell-side estimates put the combined damage for the twelve months to June 2026 at ~RMB41B of losses for Alibaba, ~RMB26B for JD, and a ~RMB25B EBIT decline for Meituan (Source: Goldman Sachs via thebridgechronicle, 2025) — figures that reconcile directly to Alibaba’s segment-EBITA collapse and negative free cash flow. This is a textbook Greenwald prisoner’s dilemma: all three players would be better off with pricing discipline, but each defects to defend share, and the only stable no-cooperation equilibrium is mutual subsidy and zero profit. There is no evidence yet of the cooperative structure (most-favored-nation pricing, tacit truce) that would tame it.

Chinese public cloud / AI: the one structurally attractive arena. China’s public-cloud (IaaS/PaaS) market is consolidated and re-accelerating. Alibaba Cloud is the clear #1 at ~36% share (Q3 2025), ahead of Huawei Cloud (~16-17%) and Tencent Cloud (~9-10%) (Source: Omdia/Canalys 2025; Statista). After a flat 2023-24, the market reaccelerated to +21-26% YoY through 2025, driven by AI training and inference demand (Source: Omdia, Q1-Q4 2025). This is a far better industry structure than e-commerce: top-three oligopoly, high fixed-cost barriers, switching costs, and a demand tailwind. The competitive twist is full-stack integration — Alibaba is building “models + chips + platform + public cloud” (Qwen LLMs, proprietary T-Head inference chips, Model Studio MaaS) to differentiate versus Huawei (which leans on its Ascend silicon) and Tencent (Source: FY2026 earnings PR; Omdia 2025). The principal industry risk is U.S. export controls on advanced AI chips, a recurring theme in Alibaba’s own risk factors (Source: FY2026 20-F forward-looking/risk language).

International commerce. A fragmented, low-barrier, capital-intensive arena (Amazon, Sea/Shopee, PDD’s Temu, MercadoLibre, regional incumbents). No participant earns durable franchise economics at scale; it is a market-share land-grab funded by losses. Structurally unattractive.

Regulation — the cloud has lifted but never fully cleared. The 2020-21 platform-economy crackdown was severe: a record RMB18.2B (US$2.8B, = 4% of 2019 revenue) SAMR antitrust fine in April 2021 for the “pick one of two” (er xuan yi) exclusivity practice Alibaba had imposed on merchants since 2015, plus a forced ban on that exclusivity and a three-year rectification-and-reporting obligation (Source: SCMP, Caixin, CNN, April 2021). That rectification was formally completed and closed by SAMR on August 30, 2024, and the broader crackdown has eased (Source: SCMP, CNBC, China Daily, Aug 2024). But the residual structural damage is permanent — the exclusivity ban is precisely what enabled merchant multi-homing and PDD/Douyin’s rise. Ongoing regulatory exposure spans anti-monopoly, anti-unfair-competition, data-security/cybersecurity, content, and consumer protection, layered atop the existential VIE structure (Alibaba’s China internet operations run through contractual VIEs that have never been tested in a PRC court and could be deemed impermissible foreign investment) and HFCAA/PCAOB delisting risk (Source: FY2026 20-F).

Capital cycle (Marathon). China internet experienced a violent supply-side bust in 2021-23 — capital flight, capacity rationalization, headcount cuts, the abandonment of empire-building. That phase improved the e-commerce capital cycle by instilling cost discipline. But two new boom dynamics have since emerged: (1) the multi-player instant-retail subsidy war (an investment surge that is destroying industry profitability in real time), and (2) an AI/cloud capex super-cycle — Alibaba alone committed RMB380B over three years to AI and cloud infrastructure, the largest private computing project in China’s history (Source: alibabagroup.com, Feb 2025). Both are classic late-cycle capital deployment that, per Marathon, tends to erode forward returns. The capital cycle thus cuts two ways: e-commerce is in a price-war bust with no profit recovery in sight, while AI-cloud is in an investment boom whose returns are unproven and whose capex is currently overwhelming free cash flow.

Verdict. Mixed, and to be disaggregated. Chinese e-commerce is a structurally deteriorating industry — large but mature, fragmenting, with a destructive low-price/subsidy equilibrium and credible new entrants; barriers have fallen since the exclusivity ban. Chinese public cloud/AI is structurally attractive — a consolidated, re-accelerating oligopoly with real scale barriers, where Alibaba is the dominant player, though export-control and capex-return risks are real. International commerce is structurally unattractive. A reasonable composite verdict: Alibaba’s largest profit pool sits in a worsening industry, while its best-positioned franchise sits in a genuinely good one — the inverse of what an investor would prefer.

4. Competitive Position

The honest assessment of Alibaba requires resisting both the bull’s “untouchable network effect” narrative and the bear’s “structurally broken” caricature. Applying Greenwald’s framework — name the moat type, then run the market-share-stability and ROIC tests — yields a clear answer: a durable, high-quality scale-plus-captivity advantage in the cloud and in the core TTG monetization engine, eroding visibly at the e-commerce franchise edge, and absent entirely in international and the loss-making periphery.

Taobao/Tmall (TTG): the textbook “economies of scale + customer captivity” moat — being pressure-tested in real time. In Greenwald’s taxonomy, TTG historically held the strongest advantage type: economies of scale combined with customer captivity, manifested as a two-sided network effect (more buyers attract more merchants and vice-versa), reinforced by merchant switching costs (storefront setup, accumulated reviews/ratings, ad-algorithm learning, Alipay/logistics integration) and a proprietary data/scale advantage in ad targeting. For a decade this produced franchise economics — CMR margins and group ROIC well above cost of capital — that no entrant could replicate because customers and merchants were captive to the dominant platform.

The framework’s own diagnostic, though, is the market-share-stability test, and here TTG fails the “formidable barrier” threshold. Greenwald holds that share moves greater than ~5 points over 5-8 years indicate eroding barriers; under 2 points indicates formidable ones. Taobao/Tmall’s GMV share fell from roughly 50% (2020) to ~40-44% (2024) — a 6-10 point loss — while PDD climbed from ~7% to ~19% and Douyin built ~US$650B GMV from zero (Source: Statista; SCMP; CNBC, 2024-25). This is not the signature of a formidable moat; it is the signature of a strong incumbent losing share at the margin to differentiated entrants. Three mechanisms explain the erosion:

• The exclusivity ban broke captivity. The 2021 er xuan yi prohibition forced TTG to allow merchants to multi-home. Merchants that once had to choose Alibaba now list simultaneously on Taobao, PDD, and Douyin — directly dismantling the demand-side captivity that made the scale advantage self-reinforcing (Source: SCMP, 2021; FY2026 20-F competition risk factors).
• PDD attacked on cost structure, not scale. By going hard-discount and manufacturer-direct, PDD reached competitive scale precisely because customers were no longer captive — exactly the Greenwald condition under which scale advantages collapse.
• Douyin attacked on demand generation. Content commerce is a different demand model (impulse discovery) that TTG’s search-and-fulfillment architecture cannot match, fragmenting the buyer pool TTG once aggregated.

The clearest tell that pricing power has weakened is Alibaba’s new contra-revenue business-development program — subsidies to merchants tied to their platform ad spend, which suppressed reported Q4 CMR growth from +8% (like-for-like) to +1% (Source: FY2026 earnings PR). A platform with intact monetization power raises take-rate freely; a platform paying merchants to retain spend is defending, not extracting. That said, the moat is eroding, not gone: TTG still commands ~40%+ GMV share, 88VIP grew double-digits to >62M members (a genuinely captive high-value cohort), and the take-rate-gap-to-peers gives a real, if shrinking, monetization runway (Source: FY2026 earnings PR). The verdict on the core is durable but no longer impregnable — a wide moat narrowing at the edges.

Cloud: a genuine, defensible economies-of-scale advantage — the best franchise in the group. Alibaba Cloud’s ~36% share of Chinese public cloud (versus Huawei ~16-17%, Tencent ~9-10%) reflects a real scale advantage: cloud is a high-fixed-cost business where the largest player amortizes data-center, R&D, and now AI-model investment over the most volume, and customers face meaningful switching costs once workloads, data, and applications are embedded (Source: Omdia/Canalys 2025). The moat type here is economies of scale + switching costs, increasingly augmented by a supply/technology differentiator that — per Greenwald — is usually transient but matters at this moment: the full-stack Qwen + T-Head-chip + Model Studio (MaaS) integration. Customer count for Model Studio grew eight-fold YoY; AI revenue posted eleven straight triple-digit-growth quarters (Source: FY2026 earnings PR). The market-share-stability test passes here — Alibaba has held #1 throughout — and the industry is re-accelerating, the favorable inverse of e-commerce. The risk is U.S. chip export controls constraining the supply side, but on competitive structure this is Alibaba’s strongest position.

AIDC and the periphery: no moat. AIDC (AliExpress/Lazada/Trendyol/Daraz/Alibaba.com) is subscale against Amazon, Sea/Shopee, and Temu, competes on price and logistics in fragmented markets, and only narrowed its loss to ~breakeven in FY2026 (Source: FY2026 earnings PR). The Greenwald tests fail on every axis — unstable shares, sub-cost-of-capital economics, no identifiable barrier. The All Others portfolio (Cainiao, Amap, Freshippo, Qwen consumer) is similarly moat-less and structurally loss-making, with the consumer-Qwen user-acquisition spend a pure capital-cycle bet with no demonstrated captivity.

Direct competitor comparison.

Verdict. Alibaba possesses a durable competitive advantage in its core, eroding at the edges. The cloud franchise is a genuine, defensible economies-of-scale moat in a structurally attractive, re-accelerating market — the highest-conviction advantage in the group. The TTG monetization engine remains a wide moat (scale + a captive high-value cohort + a take-rate gap to exploit), but the market-share-stability test confirms the bears’ central point: the demand-side captivity that made it impregnable was broken by the exclusivity ban and is being steadily competed away by PDD and Douyin, with the contra-revenue subsidy program betraying weakened pricing power. AIDC and the periphery have no moat and consume capital. The correct one-line characterization: a fortress core that is no longer growing its franchise, attached to one excellent emerging franchise (cloud) and a portfolio of capital-consuming, moat-less bets. That is a good business — but a narrower, more contested one than its history implies, and one whose competitive position must be valued segment-by-segment, never as a monolith.

5. Growth History and Forward Opportunities

Revenue — multi-year, by segment (RMB millions, fiscal year ending March 31; FY26 reporting structure):

¹FY2024 segments are not comparable — they predate the FY26 reorganization (Q1 FY26 recombined Taobao-Tmall, Ele.me, Fliggy into one China E-commerce group and pushed Cainiao/Amap/media into “All others”). FY2024 is shown at group level only; Cloud FY2024 ≈ RMB106Bn per the FY2024 20-F.

The headline (+3%) materially understates the underlying business. Two large, deliberate divestitures — Sun Art (hypermarkets) and Intime (department stores), both sold in FY2025 — strip low-margin direct-sales revenue out of the comparison. On a like-for-like basis excluding those businesses, group revenue grew +11% in FY2026, and the same +11% in Q4. The “All others” −25% collapse is almost entirely the loss of Sun Art/Intime plus a Cainiao decline; it is not organic deterioration. So the right way to read Alibaba’s top line is: a low-double-digit organic grower being optically dragged by a deliberate balance-sheet cleanup.

Core commerce CMR — decelerated but stabilizing. Customer management revenue (CMR — the high-margin advertising/commission take on Taobao/Tmall, the true profit center) grew +5% reported in FY2026 (FY25 RMB326,769M → FY26 RMB343,867M), or +7% like-for-like excluding a new accounting reclassification. In Q4 FY26, CMR grew only +1% reported but +8% LFL: a “new business development program” now ties merchant subsidies to marketing spend and reclassifies those subsidies from S&M expense into contra-revenue, optically suppressing reported CMR. Management attributes the FY26 take-rate improvement to monetization of newer ad formats and the 88VIP cohort (>62M members, double-digit growth). This is the single most important number in the entire model: CMR is high-incremental-margin and funds the rest of the company. A 7–8% LFL pace — after years of zero-to-low growth amid Pinduoduo/Douyin competition — is a genuine re-stabilization, but it is not a return to the high-teens growth of the 2018–2021 era.

Cloud — the real re-acceleration, and the highest-quality growth in the model. Cloud Intelligence revenue grew +34% FY26 (RMB118,028M → RMB158,132M), accelerating to +38% in Q4 with external-customer revenue +40%. The driver is genuinely AI: AI-related product revenue posted its 11th consecutive quarter of triple-digit growth, reaching RMB8,971M in Q4 alone, and the Model Studio (MaaS) customer base grew eight-fold YoY. This is externally validated, durable, and structurally advantaged (Alibaba is the #1 cloud provider in Asia-Pacific and a leading China AI-cloud platform). It is the cleanest “second growth curve” in the business. The caveat: cloud segment EBITA is still only RMB14,265M on RMB158,132M of revenue — a ~9% margin — and the AI capex required to sustain it is enormous (below). High-quality revenue growth; not yet high-quality returns.

Quick commerce / instant retail — the land-grab, and the source of the FY26 margin destruction. Quick commerce grew +47% FY26 (and +57% in Q4) after the late-April-2025 rollout of “Taobao Instant Commerce.” This is a direct response to Meituan and JD’s instant-retail push. It is subsidized GMV: quick-commerce revenue is reported net of contra-revenue subsidies, and the spending behind it is the principal reason group adjusted EBITA fell 56% for the year and 84% in Q4. Management claims sequential unit-economics and average-order-value improvement, “with increasing focus on high-value food orders” — but this is a hypothesis, not yet visible in segment EBITA, where the China E-commerce group’s adjusted EBITA fell 44% (RMB193,223M → RMB107,509M). This is the lowest-quality growth in the portfolio: incremental revenue that currently destroys, rather than creates, operating profit.

AIDC (International) — scaling toward breakeven, the genuine bright spot on the cost side. AIDC revenue grew +9% to RMB144,170M, and — critically — the segment’s adjusted-EBITA loss narrowed 86%, from −RMB15,137M (FY25) to just −RMB2,051M (FY26), with Q4 essentially at breakeven (−RMB138M). The improvement is AliExpress Choice unit-economics and logistics optimization. AIDC moving from a ~RMB15B annual drag to roughly breakeven is a meaningful, high-quality margin tailwind that partially offsets the quick-commerce bleed.

Forward opportunities. (1) Qwen / AI monetization — full e-commerce integration into the consumer Qwen app and the Qwen Shopping Assistant create a new AI-native funnel, but consumer Qwen is today a cost (user-acquisition spend is the main driver of the −RMB35,737M All-Others loss). (2) Cloud capex cycle — the RMB380B three-year AI/cloud buildout is a bet that AI-inference demand sustains 30%+ cloud growth and eventually lifts cloud margins. (3) CMR take-rate normalization — software-service-fee and ad-load increases give a durable lever as competitive intensity stabilizes. (4) International — AIDC at breakeven can pivot from defense to profitable scaling.

Verdict — mixed-quality, transitional growth. The durable, high-quality engines (CMR +7–8% LFL; cloud +34–40% AI-led) are intact and arguably strengthening. But FY2026’s reported growth profile is dominated by a deliberate trade: management is sacrificing near-term profit to (a) win the instant-retail war and (b) capture AI-cloud demand. The cloud growth is high-quality; the quick-commerce growth is subsidized and currently value-destructive; the AIDC turn is genuinely high-quality. Net: this is a company whose best business (CMR + cloud) is being asked to fund two large bets, one proven (cloud) and one unproven (quick commerce). Growth quality is therefore good at the core, speculative at the margin.

6. Financial Quality

Margins — a sharp, deliberate compression (RMB millions / % of revenue):

Gross margin is stable at ~40% — the COGS pressure from quick-commerce/cloud expansion (cost of revenue 60.0% → 60.2%) is offset by the Sun Art/Intime exit. The damage is entirely below the gross line: sales & marketing exploded from RMB144,021M (14.5% of revenue) to RMB245,023M (23.9%) — a +RMB101B, +9.4pt increase — driven by quick-commerce subsidies and Qwen user acquisition. That single line is the story of FY2026. Operating margin collapsed from 14% to 5%, and adjusted EBITA margin from 17.4% to 7.5%.

Segment adjusted EBITA — TTG is the cash engine funding everything else (RMB millions):

The structure is stark: China E-commerce (TTG) generates RMB107.5B of adjusted EBITA — more than the entire group — while every other segment except cloud loses money. Cloud contributes a still-modest RMB14.3B. AIDC has nearly fixed its losses (−RMB2.1B). But “All others” — which now houses the consumer Qwen user-acquisition spend, plus Freshippo/Cainiao/media — widened its loss 276%, from −RMB9.5B to −RMB35.7B. And TTG’s own EBITA fell 44% because the quick-commerce subsidy lives inside it. So even the cash engine is being run hot to subsidize growth. The blunt read: one profitable mature business (TTG) and one profitable-but-sub-scale business (cloud) are funding tens of billions of RMB in deliberate annual losses elsewhere.

Free cash flow vs. the capex surge — the defining FY2026 fact. Alibaba generated positive FCF every year of its public life until now. In FY2026 it reported a free-cash-flow OUTFLOW of RMB46,609M (−US$6.8B), versus +RMB73,870M in FY2025. Operating cash flow fell 53% to RMB76,213M, while capex surged 47% to RMB126,063M (FY25 RMB85,972M). Q4 alone was −RMB17,300M FCF. The capex is the RMB380B / three-year cloud + AI infrastructure program announced February 2025 — management’s own framing was that the next-three-years investment “will exceed the total investment made in the past decade.” Depreciation already rose 27% (the adjusted-EBITDA D&A addback went from RMB29,260M to RMB37,067M), and property & equipment, net jumped from RMB203,348M to RMB282,699M (+39%) — the datacenter buildout is now visibly inflating the asset base, which will pressure ROIC for years until the AI revenue scales into it.

SBC and dilution — a relative bright spot. Total share-based compensation was RMB14,821M FY26 (RMB11,180M non-cash), down 5% YoY and only ~1.4% of revenue — disciplined for a megacap tech, and management is deliberately shifting toward long-term cash incentives. Diluted share count actually fell slightly (19,318M → 19,235M). SBC is not a quality flag here; dilution is well-contained. (Note the equity-linked convertibles/exchangeables issued in FY26 create future dilution optionality, offset by capped-call hedges on the convert.)

ROE / ROIC — structurally depressed, and worth understanding why. FY2026 ROE is only ~9–10% (net income RMB102,127M on ~RMB1,098B average equity). This is not a sign of a low-return operating business — TTG’s incremental returns on capital are very high. It is depressed by (i) a massive, partly idle equity base: RMB520.8B of cash/liquid investments, RMB206.8B in equity-method investees (mostly the 33% Ant stake, which earns a single-digit yield and whose contribution fell to RMB5.0B), RMB449.9B in equity securities, and RMB247.4B of goodwill — together roughly RMB1.4 trillion of assets earning well below the operating ROIC; and (ii) FY26’s depressed operating profit. Strip out the cash and investment carrying values and the operating business earns far higher returns; the consolidated ROE understates economic quality but accurately reflects a balance sheet bloated with low-yield assets and goodwill.

Quality-of-earnings flags — significant. (1) GAAP net income is increasingly an investment-portfolio artifact. FY26 “interest and investment income, net” was +RMB87,512M (+322% YoY), driven by mark-to-market gains on the equity portfolio plus the Trendyol GO disposal gain. This is why GAAP net income fell only 19% while non-GAAP net income fell 62% (RMB158,122M → RMB60,658M) — the cleaner read of operating reality. Investors anchoring on GAAP EPS (diluted RMB44.00/ADS) are partly anchoring on volatile portfolio marks. (2) Goodwill impairments distort GAAP operating income: RMB9,515M FY26 (All-others), +54% YoY — the residue of the Sun Art/Intime/Youku/retail mis-investments of prior years. (3) The contra-revenue reclassification of merchant subsidies optically suppresses CMR growth (reported +1% Q4 vs +8% LFL) — a presentation choice that flatters the S&M line and muddies the true take-rate trend.

Verdict — do economics improve with scale? Yes at the core, not yet at the margin, and not at the consolidated level right now. TTG/CMR demonstrates classic operating leverage and high incremental margins; AIDC is proving that international scale eventually flips to breakeven. But FY2026 is a deliberate de-scaling of consolidated profitability: management is spending RMB100B+ of incremental S&M and RMB40B+ of incremental capex to buy growth options. The balance sheet is fortress-like (net cash ~US$37.8B), so the company can absolutely afford the bet. But on the current trajectory, economics are deteriorating with scale because management has chosen to make them so. The honest verdict: a high-quality core profit machine wrapped in a self-inflicted, balance-sheet-funded investment cycle whose returns are unproven.

7. Capital Allocation

Capital returns — multi-year (the central signal):

The buyback collapse is the most important capital-allocation fact in this report. Alibaba ran one of the largest repurchase programs in China internet — US$12.5B in FY2024 (1,249M shares) and US$11.9B in FY2025, each shrinking the share count ~5.1%, executed into a depressed valuation (Marathon-correct behavior: buying a real asset below intrinsic value). Then in FY2026 buybacks collapsed to US$1.046B (just 73M shares, all on the NYSE, and only in the April–August 2025 window — nothing after August). Remaining authorization is ~US$19.7B (the US$40B program upsized US$25B in Feb 2024, effective through March 2027), so this is a choice, not a constraint. Management redirected the cash from shareholder returns to (a) the RMB380B AI/cloud capex and (b) the quick-commerce subsidy war. The pivot is rational if those bets clear the cost of capital — but it reverses the prior, clearly-good policy of shrinking the float cheaply, and it does so at a still-modest valuation. Through a Marathon lens, halting buybacks at a low price to fund unproven growth capex is the weaker of the two uses of capital unless cloud/AI ROIC is demonstrably above WACC, which it is not yet.

Dividends. Alibaba initiated its first-ever dividend in FY2024 (US$4B annual + special). It has been sustained and is now a recurring annual program: for FY2026 the board declared US$0.13125/ordinary share (US$1.05/ADS), ~US$2.5B; total cash paid during FY26 was RMB33,732M (≈US$4.9B, including the prior year’s special). The dividend is small relative to cash generation and is dwarfed in importance by the buyback decision.

M&A and — more importantly — disposals: a genuine, credible refocusing. The 2023–2026 story is a retreat from the empire-building of the prior decade. Sun Art (hypermarkets) and Intime (department stores) — capital-intensive, low-return physical retail bought at peak-conglomerate ambition — were both sold in FY2025, taking disposal losses but recovering capital. Trendyol GO (Turkish local services) was 85%-sold in FY2026 for ~US$0.7B (RMB5B). This is disciplined, value-accretive housekeeping: exiting balance-sheet-heavy, low-ROIC assets to “double down on core businesses and invest strategically in AI” (FY25 20-F). The recurring goodwill impairments (RMB9.5B FY26) are the cost of admitting the prior mis-investments — painful but honest.

The abandoned 2023 “six-unit split.” In March 2023 Alibaba announced a six-business-group split with multiple intended spin-offs/IPOs. By November 2023 the centerpiece — the full Cloud spinoff — was cancelled, explicitly citing US chip-export-control uncertainty, and the Cainiao IPO was withdrawn. By Q1 FY2026 management had effectively recombined the consumer businesses (“strategic combination of Taobao-Tmall, Ele.me and Fliggy into Alibaba China E-commerce Group”). The whiplash — announce a radical breakup, then reverse it within a year — is a mark against strategic consistency, though the reversal (keeping cloud in-house as the AI flywheel’s core) looks correct in hindsight given the AI pivot.

Debt issuance — a capital-structure tension worth flagging. With net cash of ~US$37.8B, Alibaba nonetheless issued in FY2026: zero-coupon convertible senior notes due 2032 (US$3.2B, Sept 2025, with capped-call) and zero-coupon exchangeable bonds due 2032 referencing its Alibaba Health stake (HK$12B, July 2025). Total debt rose to ~RMB260B and total-debt/adjusted-EBITDA jumped from 1.14x to 2.29x (the EBITDA denominator halved). Issuing equity-linked debt — i.e., selling cheap optionality on its own (arguably undervalued) shares — while simultaneously halting buybacks is internally inconsistent capital allocation. The charitable read is liquidity/tax optimization (offshore USD/HKD funding for international and cloud). The skeptical read is that management is both a seller (converts) and a non-buyer (paused repurchases) of its own equity at the same low valuation.

Compensation / incentive alignment. FY2026 D&O cash compensation was a modest RMB250M (US$36M) plus 28.1M shares granted. The annual cash-bonus pool is formulaically tied to “adjusted pretax operating profits” — a reasonable, profit-linked metric (though “adjusted” gives latitude). The Alibaba Partnership + WVR structure is the dominant governance fact: the Partnership nominates a simple majority of the board, so public shareholders have limited ability to discipline capital allocation. Insider ownership is low — all directors/officers as a group own just 1.9% (349.5M shares); Joe Tsai 1.5% (largely via Parufam/PMH/foundation entities, with no pecuniary interest in ~158M of those shares); Eddie Wu 0.1%. Founder Jack Ma is a “continuity partner” but no longer a director and is no longer a separately-disclosed ≥1% holder — a notable change from the founder-dominated cap table of a decade ago.

Ant Group. The 33% (fully diluted) equity-method stake is a major, separately-valuable asset (carrying value embedded in the RMB206.8B investee line) whose earnings contribution fell to RMB5.0B (from RMB12.6B) as Ant reinvests. Alibaba paid Ant RMB18,019M in payment-processing fees in FY26 (related party) and received a RMB3,293M dividend. Ant has conducted its own buybacks at depressed valuations in recent years (a partial liquidity/value-realization avenue for Alibaba’s stake), but no IPO is on the table.

Verdict — historically good, currently in question. For FY2024–FY2025, Alibaba’s capital allocation was genuinely good: massive buybacks at a low valuation (Marathon-correct), a sensible initiation of dividends, and a disciplined exit from value-destructive physical retail. FY2026 introduces real doubt. The pivot from ~US$12B/year of buybacks to ~US$1B — redirecting capital into an unproven RMB380B AI capex cycle and a margin-destroying quick-commerce subsidy war — is defensible only if those investments earn above the cost of capital, which is not yet demonstrated. Issuing equity-linked debt while pausing buybacks at the same valuation is internally inconsistent. The fortress balance sheet (net cash ~US$37.8B) means the company can afford to be wrong; the WVR/Partnership structure means shareholders can’t force a different path. Capital allocation: a strong multi-year track record now entering its highest-risk, lowest-visibility phase.

7a. Filings Sweep — Insider / Beneficial-Ownership & Material-Event Read

The trailing 36-month SEC corpus comprises 4× 20-F annual reports and ~171× 6-K furnishings, plus a thin set of Form 3/4 filings — the latter a direct consequence of foreign-private-issuer status: most Alibaba insiders are not required to file Section 16 Form 4s, so the usual open-market-buy-vs-10b5-1-sale signal is essentially unavailable and should not be over-read. The authoritative insider data is therefore the 20-F beneficial-ownership table, which shows strikingly low insider economic alignment: all directors and executive officers as a group own just 1.9% (349.5M of 18,669,888,147 ordinary shares as of May 18, 2026). Group Chairman Joseph Tsai holds 1.5% (272.5M shares) — but overwhelmingly through controlled entities (Parufam Limited 147.4M, PMH Holding 113.5M, and the Joe & Clara Tsai Foundation 10.7M), with Tsai disclaiming pecuniary interest in ~158M of them; his direct holding is just 0.8M shares. CEO Eddie Wu holds only 0.1% (19.7M shares). Founder Jack Ma is a designated “continuity partner” of the Alibaba Partnership but is not a director and no longer a separately disclosed ≥1% holder — the founder’s economic footprint has shrunk markedly. Crucially, control does not flow from economics here: the Alibaba Partnership’s weighted-voting-rights structure lets it nominate a simple majority of the board regardless of share ownership, so the low insider stake does not translate into weak control — but it does mean management/insiders have limited direct financial skin in the per-share outcome, a modest negative for alignment.

There were no notable discretionary insider open-market purchases in the period (none would necessarily surface given FPI rules). The most material capital-markets actions in the 6-K timeline were: the FY25 Sun Art and Intime disposals; the July 2025 Alibaba-Health-referenced exchangeable bond (HK$12B) and September 2025 convertible senior notes (US$3.2B); the late-April-2025 Taobao Instant Commerce launch (the quick-commerce land-grab); and the May 2025 special + annual dividend declaration. One-time items distorting run-rate are significant and must be normalized: (i) FY26 GAAP net income is flattered by RMB87.5B of interest-and-investment income — largely non-operating equity mark-to-market gains plus the Trendyol GO disposal gain (excluded from non-GAAP NI); (ii) FY26 carries RMB9.5B of goodwill impairment (legacy retail) in GAAP operating income; (iii) FY25 G&A included a one-time shareholder-class-action provision that inflated the prior-year base. The clean, run-rate earnings figure is the non-GAAP net income of RMB60,658M (−62% YoY) — not the RMB102,127M GAAP figure — and the clean cash signal is the −RMB46.6B free-cash-flow outflow, both of which the headline GAAP EPS obscures.

8. Major Changes — Last Two Years

The two years to March 2026 are the most consequential strategic period in Alibaba’s history since the 2020 Ant IPO collapse, and they cut in opposing directions: a decisive recentralization and AI pivot on one hand, and a self-inflicted margin collapse from the instant-commerce subsidy war on the other. The thread connecting them is a management team (Eddie Wu, CEO since September 2023; Joe Tsai, Chairman) deliberately trading near-term profit for two long-duration options — leadership in China AI/cloud, and defense of the consumption ecosystem — while the external regime (US chip controls, China deflation, the dormant-but-live VIE/delisting tail) constrains both.

1. The “1-to-6” reorganization and its reversal. In March 2023 Alibaba announced the most radical restructuring of its life: splitting into six independently-governed business groups, each potentially to raise outside capital and IPO. By late 2023–2024 the plan was substantially unwound. The full spinoff of Cloud Intelligence Group was cancelled on November 16, 2023, with management explicitly citing US semiconductor export-control uncertainty — a spun-off, capital-raising cloud unit could not credibly promise investors a stable AI-compute supply when its access to advanced chips was a function of Washington’s licensing regime. The Cainiao Hong Kong IPO was withdrawn in March 2024; the Freshippo/Hema IPO was shelved. By mid-2025 the structure was recentralized: Taobao/Tmall, Ele.me and Fliggy were combined into “Alibaba China E-commerce Group” (effective the June 2025 quarter), with Cainiao, Amap and Digital Media reclassified into “All others.” Interpretation: the reversal is an admission that the 2023 break-up thesis (unlock value via independent listings) was overtaken by a harder reality — capital markets were shut to Chinese tech IPOs, and the chip-supply constraint made a standalone cloud uninvestable. The new strategy (“user-first + AI-driven”) is a coherent, if expensive, retreat to integration.

2. The RMB380B (~US$53B) AI + cloud capex commitment. On February 24, 2025, Wu committed at least RMB380 billion over three years to cloud and AI infrastructure — a figure that, by the company’s own framing, exceeds Alibaba’s total AI/cloud spend of the prior decade, and the largest such private-enterprise commitment in China. Management has since signaled it may spend above this number. The early evidence is in the cash flows: FY2026 capex pressure and the cloud build drove free cash flow to a −RMB46.6B outflow (vs +RMB73.9B the prior year), and Q4 capex alone was RMB26.9B. Interpretation: this is the single most important capital-allocation decision in the file. It is a bet that China’s AI-cloud market is a winner-take-most race and that Alibaba’s full-stack position (Qwen models, Model Studio MaaS, T-Head inference chips, hyperscale capacity) wins it. It is also the bet most exposed to the chip-control tail (see ). The validation signal is real — Cloud revenue +38% in Q4 (external +40%), AI-related product revenue at 30% of cloud and an eleventh consecutive quarter of triple-digit AI growth — but the capex is being spent ahead of the FCF that would justify it.

3. The instant-commerce / food-delivery subsidy war. The largest swing factor in the last 12 months. After JD formally entered food delivery (February 2025) and Meituan defended, Alibaba launched “Taobao Instant Commerce” (end-April 2025) and a ~RMB50B subsidy program (July 2025). The cost was severe and industry-wide: the three platforms incurred >RMB100B (~US$14B) in delivery subsidies/sales expense across 2Q+3Q CY2025 (TechNode, Dec 1 2025); Meituan swung to a RMB23.4B (~US$3.4B) FY2025 net loss (Caixin, Mar 27 2026); Alibaba’s relevant operating profit fell from RMB35.2B to RMB5.4B. Inside Alibaba’s own segments, China E-commerce Group adjusted EBITA fell 44% for the year (RMB193.2B → RMB107.5B) and “All others” EBITA loss widened from −RMB9.5B to −RMB35.7B, with quick-commerce and Qwen user-acquisition the named culprits. By early 2026 the war was abating — regulators (SAMR) publicly branded it a “race to the bottom” and on April 17, 2026 fined seven platforms RMB3.6B (US$528M) for “ghost delivery”/food-safety abuses, and management reported improving quick-commerce unit economics and AOV. Interpretation: this is the clearest case in the file of profit deliberately sacrificed for defensive market-share reasons. Whether it created durable value (a stickier 88VIP-anchored ecosystem, now >62M members) or simply destroyed RMB tens of billions defending against Meituan/JD will not be known for several quarters. The regulatory-forced ceasefire is the bull’s near-term catalyst.

4. Portfolio pruning. Alibaba exited two large physical-retail positions — Sun Art (Auchan/RT-Mart, sold to DCP) and Intime (department stores) — in FY2025, plus Trendyol’s local-services business in FY2026. These disposals are the entire explanation for the optical disconnect between +3% headline revenue and +11% like-for-like. Interpretation: a clean, correct narrowing toward the high-return core (commerce monetization + cloud) and away from capital-heavy, low-margin offline retail that never fit. Directionally thesis-strengthening.

5. Shareholder-return inflection and the Hong Kong primary listing. Alibaba declared its first ordinary dividend in FY2024 and has institutionalized it — FY2026 regular dividend of US$1.05/ADS (~US$2.5B aggregate), payable July 2026, alongside continued buybacks. Separately, the August 28, 2024 upgrade to a dual-primary HK listing and September 10, 2024 inclusion in Stock Connect (Southbound) opened the register to mainland Chinese investors — Goldman estimated ~US$15–16B of potential inflow. Interpretation: both moves matter. The dividend + buyback signal a maturing-into-capital-return mindset; the HK primary listing is a partial structural hedge against the US delisting tail because the economic interest survives intact on HKEX even in a forced ADR delisting, and it broadens demand to a buyer base less sensitive to HFCAA risk.

Verdict (Changes): Net, the period is thesis-clarifying but value-ambiguous. The strategic direction is sound — recentralize, double down on AI/cloud, prune offline retail, return cash, hedge the listing — but the execution cost (a 56% collapse in adjusted EBITA, negative FCF, a subsidy war that torched RMB tens of billions) means the last two years destroyed measurable near-term earnings power in exchange for option value that is not yet proven. The bull reads this as a trough engineered by management; the bear reads it as evidence that the core franchise can no longer earn its keep without burning cash to defend it. Both are defensible on the current evidence — which is precisely why the stock is contested.

9. Risk Analysis — Risk Matrix

Alibaba’s risk profile is dominated by a category most operating businesses never face: a structural, partly-unquantifiable overhang on the value-realization mechanism itself — the VIE structure, ADR delisting risk, and capital mobility — which sits on top of ordinary competitive and cyclical risk. The correct framing is to treat the China/VIE/geopolitical cluster as a discount to economic value, not a competitive flaw: the underlying businesses can be excellent while the claim a foreign shareholder holds on them remains contractually and politically fragile.

Discussion of the dominant tail. The VIE-plus-delisting cluster is the defining feature of the BABA risk profile and the reason the stock structurally trades below a “clean” DCF of its cash flows. A US shareholder does not own Alibaba’s Chinese e-commerce or cloud operations directly; they own equity in a Cayman holding company whose control over the PRC operating entities rests on a web of contracts (powers of attorney, exclusive service agreements, equity-pledge agreements) that have never been stress-tested in a hostile PRC court. The HFCAA delisting risk is currently dormant — the 2022 PCAOB access agreement removed BABA from the immediate-delisting path — but it is explicitly reinstatable the moment Beijing obstructs inspections, and the 20-F preserves this as a live risk factor. The correct analytical posture: this is not a flaw in the business; it is a flaw in the claim. It should be priced as a discount to economic value (a probability-weighted haircut for impaired realization), not netted against the moat.

Cyclical vs. structural. Of the operating risks, the subsidy war and chip controls most distort current earnings. The subsidy war appears to be a cyclical trough that management and regulators are actively ending (the SAMR fines and “race to the bottom” rhetoric function as a de-escalation mechanism), which means current depressed EBITA likely understates normalized earning power. The chip-control constraint is more structural: even with H200 clearance in May 2026, no deliveries had occurred, and China’s frontier AI compute is migrating to Huawei/Cambricon ASICs where Alibaba holds only ~5% share — a genuine, durable ceiling on how fast Alibaba Cloud’s AI franchise can scale.

Verdict (Risk): The risk matrix is bimodal, not gradient. The day-to-day operating risks (consumption, competition, subsidy economics, capex discipline) are real but manageable and partly self-correcting — the subsidy war is easing, the core is profitable, the balance sheet is net-cash. The defining risk is the low-probability / severe-impact geopolitical-structural tail (VIE unenforceability, forced delisting + capital-repatriation block), precisely the kind of fat tail that justifies a permanent valuation discount and cannot be diligenced away. An investor is not paid to underwrite the business — that is sound — but to underwrite the jurisdiction and the claim. The HK primary listing meaningfully de-fangs the delisting leg; nothing de-fangs the VIE/repatriation leg.

10. Valuation Discussion — Embedded Expectations

No price target and no recommendation follow. This section frames what the ~US$121 ADR price embeds, decomposes the enterprise into parts and into a net-cash-plus-stakes floor, and reverse-engineers the growth/margin path the market is underwriting. The defining feature of Alibaba’s valuation is that a large share of the market capitalization is covered by non-operating assets — net cash, the Ant stake, and a marketable/private equity portfolio — so the question is unusually literal: how much are you paying for the operating businesses, and what must they do to be worth it?

The setup. At US$121.06/ADS, market cap is ~US$290B (≈2.4B ADS / ~19.2B ordinary shares). Reported FY2026: net income US$14.8B (−19%), adjusted EBITA US$11.08B (−56%), operating income US$7.27B (−64%), and a free-cash-flow OUTFLOW of US$6.76B. Trailing P/E ≈18.6×; forward P/E ≈13×. On the firm’s own valuation-index, BABA sits at a composite ~31st percentile vs. its own ~10-year history — i.e., cheapish against itself, but not a generational low (the 2022 trough was cheaper). Note the asymmetry: FY2026 earnings are visibly depressed (subsidy war + capex), so trailing multiples overstate the price relative to normalized power, and the ~13× forward multiple is the more honest anchor.

1. Sum-of-the-Parts (the “you get the core cheap” math)

The bull’s central claim is that net cash + Ant + the investment portfolio cover most of the market cap, leaving the operating businesses for a fraction of fair value. The balance-sheet floor:

So the market is pricing the entire operating engine — China commerce (the cash machine), Cloud (the growth/AI franchise), AIDC and the rest — at roughly US$185–210B.

Valuing those operating parts independently:

• (a) China E-commerce / TTG core. FY2026 segment adjusted EBITA US$15.6B (RMB107.5B) — but this is depressed by the subsidy war (the prior year was RMB193.2B / ~US$28B). On a normalized EBITA of ~US$22–26B and a 10–13× EV/EBITA (defensible for a dominant, cash-generative, but slow-growing, regulated marketplace), the core is worth ~US$220–340B alone — i.e., the core’s normalized value by itself approximates or exceeds the entire ~US$185–210B the market assigns to all operating businesses combined.
• (b) Cloud Intelligence Group. FY2026 revenue US$22.9B (+34%), EBITA US$2.07B, AI revenue compounding triple-digit for 11 straight quarters. At a deliberately conservative China-discounted 3–5× EV/Sales (vs. US hyperscalers far higher), Cloud is worth ~US$70–115B — and arguably more if the AI inflection is real.
• © AIDC / Cainiao / local services / quick commerce. AIDC approaching breakeven (FY26 EBITA −US$0.3B, much narrowed); quick commerce a strategic option, not yet a profit center. Treat as option value / small positive — call it US$0–20B, conservatively zero-to-low.

The SOTP punchline: (a) normalized core (~US$220–340B) + (b) Cloud (~US$70–115B) + © ~US$0–20B + the ~US$80–105B non-operating floor sums to a gross-asset range that materially exceeds the ~US$290B market cap even before any premium — i.e., on a parts basis the bull math holds: the core is being valued at or below a single, conservative segment, with Cloud and the cash/stakes thrown in for little. The bear’s rebuttal is twofold: (i) the “normalized” core EBITA may be permanently lower if the subsidy war is structural and CMR keeps ceding share to PDD/Douyin; and (ii) the non-operating floor is partly non-fungible — most cash is onshore RMB under capital controls, Ant is illiquid and politically constrained, and the portfolio marks are soft. A SOTP that sums conglomerate parts a foreign holder cannot freely monetize is worth less than its arithmetic.

2. Reverse-DCF / Embedded Expectations

At ~US$290B market cap and ~US$185–210B of implied operating value, against depressed FY2026 operating income of US$7.3B and a negative current FCF, the market is plainly not capitalizing trailing FCF — it is underwriting a normalization plus modest growth. Backing into it: to justify ~US$190B of operating value at a ~9–10% discount rate and ~3% terminal growth, the operating businesses must generate roughly US$17–22B of steady-state FCF within a few years — i.e., the subsidy war ends, capex normalizes off the RMB380B peak, China commerce EBITA recovers toward its prior US$25B+ run-rate, and Cloud scales to FCF-positive at scale. Interpretation: the ~13× forward multiple and the implied operating value embed a recovery to roughly mid-cycle earnings power, with low-single-digit terminal growth — i.e., the market is pricing stabilization, not re-acceleration, and is not paying for an AI-cloud re-rating. That is the crux: consensus appears to price a permanent China/structural discount and terminal stagnation, leaving the AI-cloud optionality (Cloud +34–40%, AI 30% of cloud) effectively un-underwritten. If that optionality is real, the price embeds too little; if China commerce is in secular share-loss, the price embeds too much “normalization.”

3. Greenwald EPV Floor (strip growth)

Capitalize normalized operating earnings with zero growth. Take a mid-cycle adjusted EBITA of ~US$22–26B for the whole group (above the depressed US$11.08B FY2026 print, below the US$24B+ FY2025 group peak), tax-effect at ~15–20% PRC-effective → ~US$18–22B normalized after-tax operating earnings. Capitalized at an 11–12% unlevered cost of capital (a China-risk-inflated rate, deliberately punitive) → an EPV of operating earnings of ~US$150–200B. Add the non-operating floor (~US$80–105B) → a no-growth, China-discounted EPV of ~US$230–305B. Interpretation: the EPV floor brackets the current ~US$290B market cap. Unlike TSMC (where EPV was ~⅓ of price, i.e., two-thirds growth value), Alibaba’s price is largely covered by no-growth earning power plus assets — there is comparatively little growth value embedded. That is the mathematical statement of “value, not momentum”: the downside is cushioned by an EPV/asset floor close to the price, and the AI-cloud growth is a free-ish call option, provided one accepts the normalized-EBITA assumption and the fungibility of the asset floor (the two load-bearing “ifs”).

4. Own-history and peer context

BABA trades at a discount to Amazon on every metric (EV/Sales, EV/EBITDA, P/E) — defensible, given AWS’s far larger, higher-margin, jurisdiction-clean cloud and the absence of a VIE/delisting tail. More tellingly, BABA trades at or above PDD (~6.9× fwd) and JD (~6.6× fwd) on forward P/E despite slower commerce growth. Interpretation: the market is not treating BABA as the cheapest China name — PDD and JD are cheaper on earnings. BABA’s relative premium to its e-commerce peers is the market paying up for (i) the Cloud/AI franchise PDD and JD lack, and (ii) the net-cash + Ant + portfolio optionality. The bear reading is that BABA should trade at a discount to faster-growing PDD and isn’t, so the commerce franchise is the rich part; the bull reading is that the cloud/AI/asset stack legitimately warrants the premium.

Verdict (Valuation): Alibaba is a net-cash, asset-rich, China-discounted value situation, not a momentum or growth-multiple story. The defining feature is that the EPV/asset floor brackets the price — roughly US$80–105B of (partly non-fungible) non-operating assets plus a no-growth EPV that, summed, approximates the ~US$290B cap, leaving the AI-cloud growth as a lightly-priced option. The embedded expectations are modest — stabilization and low-single-digit terminal growth, with a permanent structural discount — which is the bull’s whole case (the bar is low) and the bear’s whole worry (the bar is low because CMR share loss, capital-trapping, and the VIE/delisting tail may be permanent, not cyclical). The valuation is cheap on arithmetic, expensive on realizability — the gap between the two is the China discount, and whether it is too wide or correctly wide is the entire investment question. No price target; no recommendation.

11. Variant Perception

Consensus view. The Street is bifurcated and oscillating between two labels for the same stock: “value trap” versus “AI/cloud re-rating story.” The shared, uncontested facts are that Alibaba is statistically cheap (forward ~13×, own-history ~31st percentile, large net cash and stakes) and that its near-term earnings have collapsed (adjusted EBITA −56%, negative FCF). The disagreement is entirely about why it is cheap: is the discount a correct pricing of permanent structural impairment (China consumption malaise + share loss + VIE/delisting tail + trapped capital), or a behavioral/geopolitical mispricing of a still-dominant ecosystem at the trough of a self-inflicted, regulator-ending margin war? The consensus has no stable answer — which is why the stock ranged from US$103.71 to US$192.67 over the trailing 52 weeks (an ~86% high-to-low swing on a US$290B company).

The strongest bull case. Alibaba is a dominant, net-cash, cash-generative ecosystem trading near its no-growth EPV with a free AI-cloud call option. The math: ~US$80–105B of net cash + Ant + portfolio covers a quarter-to-a-third of the cap; the normalized China-commerce core alone is worth roughly the entire implied operating value; and Cloud is inflecting — +34–40% revenue, AI at 30% of cloud revenue, eleven straight quarters of triple-digit AI growth, full-stack (Qwen models, MaaS, T-Head chips). The margin collapse is a cyclical trough that management chose and that regulators are now ending (SAMR’s “race to the bottom” rebuke and April 2026 fines function as an industry ceasefire), so normalized earnings are materially above the FY2026 print. Layer in disciplined capital return (recurring dividend, US$19.1B buyback authorization), take-rate headroom (CMR LFL +7–8% on rising monetization), the HK primary listing de-risking the delisting tail, and Ant optionality (a re-IPO would crystallize a stake carried far below its 2020 peak). The bull’s framing: quality compounder at a trough price, with the downside cushioned by an EPV/asset floor.

The strongest bear case. The cheapness is correct, not anomalous, because the franchise is in structural, not cyclical, decline of its profit pool. China e-commerce is a share war Alibaba is losing at the margin — PDD’s low-price model and Douyin’s content-commerce are taking GMV, and Alibaba had to light RMB tens of billions on fire (subsidy war) merely to defend the ecosystem, the opposite of a moat that prices power. The instant-commerce expansion is margin-dilutive even when it “works.” China’s macro is a persistent deflationary, weak-consumption, property-dragged regime with stimulus fatigue — a structural ceiling on GMV and CMR. The RMB380B AI/cloud bet is capex ahead of returns, constrained by chip export controls (no H200 deliveries, ~5% domestic accelerator share vs. Huawei’s ~62%) that cap how fast the one growth engine can scale. And the asset “floor” is partly a mirage to a foreign holder: most cash is onshore RMB under capital controls, Ant is illiquid and politically leashed, the portfolio marks are soft — all sitting under a VIE structure and a dormant-but-live delisting tail that can impair the claim regardless of the business. The bear’s framing: a melting-ice-cube core dressed up by non-fungible assets, where the “cheap” multiple is the market correctly pricing impaired realization.

The 3–5 assumptions that actually decide it.

1. Is the subsidy-war margin collapse cyclical or structural? (Does normalized China-commerce EBITA recover toward US$22–28B, or is the prior peak gone permanently because defending share now costs cash?) — swings the SOTP core by US$100B+.
2. Is Alibaba winning China AI-cloud, and can it scale despite chip controls? (Cloud’s +34–40% and AI inflection vs. Huawei’s compute dominance and the H200 delivery limbo.) — decides whether the embedded “free option” is worth anything.
3. Is the non-operating floor fungible to a foreign shareholder? (Onshore-RMB capital controls, Ant illiquidity, VIE enforceability.) — decides whether the US$80–105B “floor” is real downside protection.
4. Is the VIE/HFCAA delisting tail correctly dormant, or one geopolitical shock from reactivation? — decides the size of the permanent discount.
5. Is China commerce ceding GMV share permanently to PDD/Douyin? — decides whether the core is a compounder or a melting cube.

Falsification tests.

• Falsifies the BULL: China E-commerce Group adjusted EBITA fails to recover above ~US$20B normalized over the next 4–6 quarters despite the subsidy-war ceasefire (margin hit proves structural); and/or Cloud growth decelerates below ~20% as chip constraints bite — either confirms the bear’s “melting core + capped growth” thesis.
• Falsifies the BEAR: China-commerce EBITA recovers toward the prior US$25B+ run-rate and Cloud sustains 30%+ growth with improving EBITA margin into FY2027 — confirming a cyclical trough; and/or a concrete Ant re-IPO or large offshore-funded buyback acceleration demonstrates the asset floor is monetizable to foreign holders.

Verdict (Variant Perception): The genuine variant question is not “is it cheap” (it demonstrably is on arithmetic) but “is the cheapness earned.” Consensus is paralyzed between value-trap and re-rating because the deciding variables — subsidy-war reversibility, AI-cloud scalability under chip controls, and asset fungibility under capital controls/VIE — are all currently unknowable from the outside and will only resolve over the next several quarters of EBITA recovery (or non-recovery) and any Ant/listing catalyst. The variant edge available to a patient investor is in handicapping reversibility: the bull is right that the bar embedded in the price is low and the war is ending; the bear is right that a low bar is rational when the downside includes permanent share loss, trapped capital, and a fat geopolitical tail.

12. Fact vs. Interpretation Table

13. Open Questions

1. Is the China-commerce EBITA reset cyclical or permanent? FY25 RMB193.2B → FY26 RMB107.5B. With the subsidy war easing, does it recover toward US$25B+, or has defending share against PDD/Douyin permanently raised the cost of monetization?
2. When does quick commerce reach segment-level EBITA breakeven? Management claims sequential unit-economics improvement, but the loss is currently buried inside the China-commerce and All-Others segments. What is the standalone burn rate and the path to profit?
3. Will the RMB380B AI/cloud capex earn its cost of capital? Cloud EBITA margin is still only ~9%; depreciation is rising 27%; ROIC will be pressured for years. Does AI-inference demand sustain 30%+ cloud growth and margin expansion?
4. Can Alibaba Cloud scale AI under US chip controls? H200 was “cleared” (May 2026) but with no deliveries; Huawei holds ~62% of China’s accelerator market. How binding is the compute ceiling, and how good are the in-house T-Head chips really?
5. How much of the US$75.5B gross cash is offshore and freely usable for ADR buybacks/dividends versus trapped onshore under capital controls? Disclosure is not granular.
6. What is Ant Group worth today, and is a re-IPO realistic? The last hard mark is the 2023 buyback (~US$78.5B implied); is there a catalyst to crystallize it?
7. Why issue equity-linked debt (converts/exchangeables) while pausing buybacks at the same valuation? Is this tax/liquidity optimization or a signal about management’s own view of the share price?
8. Does the contra-revenue merchant-subsidy program ever reverse, allowing reported CMR to re-converge with like-for-like, or is paying merchants now a permanent cost of competing?

14. What Must Be True (Bull vs. Bear, each with a falsification test)

For the BULL case to be right (the cheapness is a mispricing of a chosen trough):

1. The subsidy-war margin collapse is cyclical — China E-commerce adjusted EBITA recovers toward its prior ~US$25B+ run-rate as the regulator-forced ceasefire holds.
2. Cloud sustains 30%+ growth with expanding EBITA margin, validating the RMB380B capex and the AI franchise as a real, scaling second engine.
3. The CMR core holds ~40%+ GMV share and continues +7–8% like-for-like, i.e., the moat is narrowing but not breaking.
4. The asset floor proves at least partly monetizable to foreign holders (offshore-funded buyback re-acceleration, an Ant catalyst, sustained dividend).

Falsification test: If China-commerce EBITA is still below ~US$20B normalized 4–6 quarters after the subsidy war ends, or Cloud growth falls below ~20%, the “chosen trough” thesis is wrong and the cheapness is earned.

For the BEAR case to be right (the cheapness is a correct discount for structural impairment):

1. The core profit pool is in secular decline — defending GMV share against PDD/Douyin permanently costs cash, so normalized EBITA never returns to prior levels.
2. Chip controls cap Cloud below ~20% growth, neutralizing the only structurally attractive franchise.
3. The asset floor is largely non-fungible to ADR holders (trapped onshore cash, illiquid/politically-leashed Ant), so the SOTP “floor” doesn’t accrue to foreign equity.
4. The VIE/delisting tail reactivates (a geopolitical shock), imposing a step-change discount or impairment on the claim.

Falsification test: If China-commerce EBITA recovers toward US$25B+ and Cloud holds 30%+ with margin expansion into FY2027 and a concrete Ant/offshore-cash catalyst emerges, the structural-decline thesis is wrong and the discount was a gift.

15. Source Appendix

A full source appendix — primary filings (FY2024–FY2026 20-Fs, the FY2026 earnings 6-K/press release, beneficial-ownership and risk-factor sections), the trailing 36-month SEC corpus, third-party market-share/industry data, regulatory and chip-control sources, and macro data — is provided as Appendix B below. Primary sources (Alibaba SEC filings) are the authority for all financial figures; third-party estimates (GMV share, cloud share, subsidy-war losses, Ant valuation) are labeled as such throughout and reconciled to filings wherever possible. Because BABA is an ADR (1 ADS = 8 ordinary shares) reporting in RMB, aggregator EV/EBITDA, cash/debt, and per-share figures are RMB-unit-distorted for the ADR and were discarded — every material number is anchored to the 20-F.

APPENDIX A — Standard Diligence Questionnaire

Companion to the research note on Alibaba Group Holding Limited (NYSE: BABA · HKEX: 9988), FY2026 (ended March 31, 2026). Fact / Interpretation / Assumption labels applied where they matter. Figures in RMB unless noted; 1 ADS = 8 ordinary shares.

General

What thoughtful questions have other investors asked about this company? The recurring serious questions: (1) Is the FY2026 margin/FCF collapse a cyclical investment trough or structural franchise decline? (2) Can Alibaba Cloud scale its AI franchise under US chip-export controls? (3) How much of the US$75.5B gross cash and the Ant stake is actually fungible to a foreign ADR holder given onshore RMB capital controls and VIE structure? (4) Is the VIE/HFCAA delisting tail genuinely dormant? (5) Is Taobao/Tmall permanently ceding GMV share to PDD and Douyin? (6) Why did management slash buybacks ~90% (from ~US$12B to ~US$1B) while the stock was cheap, and issue equity-linked debt? These are the right questions; this memo addresses each.

Cyclicality & Earnings Nature

Are earnings at a cyclical high or low? Low — a deliberately chosen trough. (Fact) FY2026 adjusted EBITA −56%, operating income −64%, and free cash flow turned negative (−RMB46.6B) for the first time in Alibaba’s public life. (Interpretation) This is self-inflicted — the food-delivery subsidy war plus the RMB380B AI/cloud capex — not demand collapse. Trailing multiples overstate the price vs. normalized earning power.

Driven by the external environment or internal actions? Predominantly internal (the subsidy war and capex are management choices), amplified by a weak external backdrop (China deflation/soft consumption). The regulator-forced ceasefire in the delivery war is an external de-escalation now under way.

How stable are revenues? Group revenue is stable-to-slow-growing (+3% reported, +11% like-for-like). The high-margin CMR core (+7–8% LFL) is reasonably stable and quasi-recurring; quick-commerce and direct-sales revenue are more transactional/cyclical; Cloud is fast-growing (+34%).

Outlook for products/services? Bifurcated. E-commerce: low-to-mid-single-digit GMV growth in a mature, fragmenting market. Cloud/AI: 30%+ growth with an 11-quarter triple-digit AI streak. International (AIDC): scaling toward breakeven.

How big will this market be — growing, shrinking, domestic or international? China e-commerce is the world’s largest but decelerating and structurally more competitive. China public cloud is re-accelerating (+21–26% YoY on AI) — the better market. Revenue is predominantly domestic (China); AIDC adds an unprofitable international leg.

Business Quality & Competitive Moat

Is the industry getting more or less competitive? More in e-commerce (PDD’s low-price disruption, Douyin content commerce, the multi-player instant-retail war; merchant exclusivity banned since 2021). Consolidated/oligopolistic in cloud (Alibaba ~36% #1, Huawei, Tencent).

How profitable is the business (ROIC, ROE)? Consolidated ROE is only ~9–10% (Fact) — but depressed by a ~RMB1.4T balance sheet bloated with cash, low-yield associates (Ant), an equity portfolio, and goodwill, plus FY26’s trough operating profit (Interpretation). The operating core (TTG/CMR) earns very high incremental returns on capital; the consolidated figure understates the core’s economic quality.

How profitable is the industry — how many competitors, what barriers to entry? E-commerce: low barriers post-exclusivity-ban, multi-homing merchants, chronic price competition → compressed profit pools. Cloud: high fixed-cost barriers, switching costs, top-3 oligopoly → structurally more profitable, though Alibaba’s cloud EBITA margin is still only ~9%.

Can the business be easily understood? Mostly — a take-rate marketplace + an IaaS/PaaS/MaaS cloud + international retail + a long tail. The complications are the VIE structure, the equity-investment portfolio that drives GAAP earnings volatility, and segment reclassifications.

Can it be undermined by foreign low-cost labor? Not directly (it is a domestic platform/services business). The relevant analog is domestic low-cost disruption — PDD’s manufacturer-direct model — which has already undermined Taobao/Tmall’s share.

Do brands matter? Yes — Taobao, Tmall, and the 88VIP membership (>62M, double-digit growth) are powerful consumer brands and a genuine captive high-value cohort; Alibaba Cloud and Qwen are increasingly strong enterprise/AI brands. But brand has not prevented GMV share loss to cheaper rivals.

What is the nature of competition? Price and selection in e-commerce (now subsidy-driven in instant retail); scale, performance, and AI-stack differentiation in cloud; price and logistics in international.

Customers’ switching costs? Moderate and falling on the merchant side (multi-homing is now standard); moderate-to-high on the cloud side (embedded workloads/data); high stickiness for the 88VIP consumer cohort. The new contra-revenue merchant-subsidy program is evidence that switching costs/pricing power have weakened.

Financial Condition & Balance Sheet

Assets not fully recognized on the balance sheet? Yes — the 33% Ant Group stake (carried within the RMB206.8B equity-method line, plausibly worth ~US$21–26B and historically more) and a large equity-securities portfolio (RMB449.9B carrying) contain embedded value/optionality. Brand and the data/network asset are not on the balance sheet.

Off-balance-sheet liabilities? The principal “off-balance-sheet” exposure is structural, not financial: the VIE contractual arrangements (the mechanism by which the Cayman holdco controls the PRC operating entities) and related guarantees/commitments, plus the RMB380B multi-year capex commitment. Operating leases and purchase commitments are disclosed.

How conservative is the accounting? Mixed. Conservative in some respects (recurring goodwill impairments are taken, not avoided; SBC is modest at ~1.4% of revenue). Less helpful in others: GAAP net income is flattered by large equity mark-to-market gains (RMB87.5B in FY26), so GAAP EPS overstates operating reality — non-GAAP net income (−62%) is the cleaner read. The contra-revenue reclassification of merchant subsidies optically suppresses reported CMR.

How CapEx-hungry is the business? Historically asset-light (a marketplace), now temporarily CapEx-heavy: capex +47% to RMB126B in FY26 for the AI/cloud build-out (RMB380B over three years). Property & equipment jumped +39% to RMB282.7B. This is the swing factor turning FCF negative.

Capital Allocation & Management

How much FCF does the business generate, and how does management use it? Historically a strong FCF generator (+RMB73.9B FY25); in FY26 it was a −RMB46.6B outflow as management redirected cash (and then some) into capex and the subsidy war. Philosophy has pivoted from returning capital to consuming it.

Significant acquisitions recently? No major acquisitions — the story is disposals: Sun Art and Intime (offline retail, FY25) and Trendyol GO (FY26), a deliberate retreat from balance-sheet-heavy, low-ROIC retail.

Buying back shares? Yes historically — US$12.5B (FY24) and US$11.9B (FY25), each shrinking the float ~5.1%, executed into a low valuation (Marathon-correct). But FY2026 buybacks collapsed to US$1.0B (73M shares, Apr–Aug 2025 only), with ~US$19B of authorization unused — the single most important capital-allocation signal.

Issuing large amounts of new shares to insiders? No — SBC is modest (~1.4% of revenue) and the diluted share count actually fell slightly. However, Alibaba issued equity-linked debt (US$3.2B convertible, HK$12B exchangeable) in FY26 — selling cheap optionality on its own shares while pausing buybacks, an internal inconsistency.

Compensation policy of directors/management? Modest cash (RMB250M D&O) plus equity; the annual bonus pool is tied to “adjusted pretax operating profits” (a reasonable profit-linked metric). Insider economic alignment is low — all directors/officers own just 1.9%; control flows from the Alibaba Partnership’s weighted-voting-rights structure, not from economics.

Motivations of management? (Interpretation) The Wu/Tsai team appears genuinely focused on long-term AI/cloud leadership and ecosystem defense, having de-emptied the prior conglomerate ambition. The concern is not empire-building (that has reversed) but whether the RMB380B AI bet and the subsidy war earn their cost of capital — and that shareholders cannot force a different path under the WVR structure.

Valuation & Market Data

Is the stock an ADR, MLP, or K-1 issuer? ADR — 1 ADS = 8 ordinary shares, NYSE-listed, with a dual-primary HKEX listing (9988). Cayman holding company; PRC operations controlled via VIE. Not an MLP; no K-1 (ADR holders receive 1099/standard dividend reporting).

Dividend policy? Recurring annual dividend since FY2024; FY2026 declared US$1.05/ADS (US$0.13125/ordinary share), ~US$2.5B aggregate (~0.9% yield), with occasional specials. Small relative to cash generation.

How profitable is the business? At the consolidated level, currently depressed (operating margin 5%, ROE ~9–10%). At the core level, highly profitable (China E-commerce adjusted EBITA RMB107.5B = the entire group’s profit). Cloud is modestly profitable and growing; AIDC and “All Others” lose money.

Is net income diverging from cash from operations? Yes, materially and in both directions. GAAP net income (RMB102.1B) is above a clean operating read because of RMB87.5B of non-operating investment gains; meanwhile operating cash flow fell 53% and FCF went negative. Net income is a poor proxy for cash generation this year — use FCF and non-GAAP NI.

Risks & Downside

What factors would cause the stock to decline? Core EBITA failing to recover after the subsidy-war ceasefire (proving structural decline); Cloud decelerating under chip controls; renewed China-internet regulation; China consumption deteriorating further; RMB depreciation; reactivation of the HFCAA delisting tail; or a broad China-risk-off / geopolitical shock.

Risk of a catastrophic loss? (Interpretation) Low-probability but real. The genuine tail is a US–China rupture that forces ADR delisting and blocks capital repatriation and/or renders the VIE claim unenforceable — a scenario in which a foreign holder’s economic claim is severely impaired regardless of business performance. This is the reason a permanent valuation discount is rational.

Chance of a total loss? Very low in any normal scenario (the business is profitable, net-cash, and dual-listed in HK, which preserves the economic interest if ADRs are forced off the NYSE). A total loss requires the catastrophic geopolitical/VIE tail above — low probability, non-zero, and not diligenceable away.

Recent News & Events

Has the business environment changed recently? Yes. (1) The 2025 instant-commerce/food-delivery subsidy war (the dominant swing factor) is easing under SAMR pressure (RMB3.6B fines, April 2026). (2) The RMB380B AI/cloud capex program (Feb 2025) is in full swing, driving Cloud’s +38% Q4 re-acceleration. (3) Chip-export controls tightened (H20, April 2025) then partially eased (H200 clearance, May 2026 — but no deliveries). (4) China macro remains soft (deflation, weak retail, property drag). (AZI news feed returned no coverage for the ADR; this timeline is built from the 6-K/press releases and trade press.)

Significant acquisitions? No — net disposals (Sun Art, Intime, Trendyol GO).

Change in accounting policies? Yes — a Q1 FY2026 segment reorganization (Taobao/Tmall + Ele.me + Fliggy combined into “Alibaba China E-commerce Group”; Cainiao/Amap/media moved to “All Others”) and a reclassification of merchant subsidies into contra-revenue (depressing reported CMR vs. like-for-like). Both require care in period-over-period comparison.

Recent changes — new markets, facilities, management? Eddie Wu (CEO since Sep 2023) and Joe Tsai (Chairman) consolidated control and reversed the 2023 “1-to-6” split (Cloud spinoff cancelled Nov 2023; Cainiao IPO withdrawn). Major new “facility” investment is the nationwide AI/datacenter build-out. The Hong Kong listing was upgraded to dual-primary (Aug 2024) and added to Stock Connect (Sep 2024), broadening the shareholder base to mainland investors.

APPENDIX B — Source Appendix

Target: Alibaba Group Holding Limited (NYSE: BABA · HKEX: 9988) · FY2026 (ended March 31, 2026) Authority hierarchy: SEC filings (20-F / 6-K) are primary for all financial figures. Third-party estimates (GMV/cloud market share, subsidy-war losses, Ant valuation, peer multiples) are labeled as such and reconciled to filings where possible. Because BABA trades as an ADR (1 ADS = 8 ordinary shares) while reporting in RMB, market-data-aggregator EV/EBITDA, cash/debt, and per-share figures are RMB-unit-distorted for the ADR and were discarded; every material number is anchored to the 20-F.

A. Primary sources — Alibaba SEC filings (CIK 0001577552)

B. Company corporate sources

C. Market data (convenience data — reconciled to filings)

D. Third-party industry / market data (estimates — labeled in text)

E. Regulatory, geopolitical & chip-control sources

All figures in this note are anchored to Alibaba’s SEC filings; third-party market-share, subsidy-war-loss, and Ant-valuation estimates are labeled as such in the text and reconciled to the filings wherever possible.