AMZN 19 min read

Amazon.com Inc (AMZN) — Investment Tree v1

Stage 7 final essay. Bilingual companion: tree_v1_zh.md. Date: 2026-05-30 (R2 re-anchor) · Anchor price: $270.64 (2026-05-29 close) · Market cap: ~$2.91T · Forward P/E: ~30× FY2026E ($9.00) · Dividend yield: 0% Archetype: Three-platform conglomerate with AWS RE-ACCELERATING (+28.4% Q1 2026) and the AI-capex FCF compression as the binding question (capex $131.8B, FCF collapsed to ~$1.2B)

SOURCE QUALITY: R2-VERIFIED (2026-05-30). FY2025 financials Tier-A (10-K accn 0001018724-26-000004, filed 2026-02-06) and Q1 2026 Tier-A (10-Q accn 0001018724-26-000014, filed 2026-04-30); price/multiples Tier-B (2026-05-29). Supersedes the 2026-05-13 GENERATE-mode scaffold (which used FY2024 financials and anchored at $220 — a +23% understatement vs the real $270.64). See evidence_2026-05-30.jsonl + update_2026-05-30.md. Only forward scenario targets remain analytical (†).

0. Company Fundamentals — what Amazon is and how it earns

Figures FY2025 (calendar 2025) unless noted.

What it is & how it earns. Amazon is a three-platform conglomerate: a North America + International e-commerce/marketplace + logistics operation (combined ~$588B, ~82% of revenue), AWS cloud & AI infrastructure (~$129B, 18% of revenue but the profit engine at a 35.4% margin and ~57% of consolidated operating income), and a high-margin Advertising business ($68.6B, +22%) embedded inside the retail segments. FY2025 total net sales were $716.9B (+12.4%) with $80.0B operating income (11.2%). The retail flywheel acquires customers; AWS and Advertising convert them to profit — and AWS is now also the centerpiece of an enormous AI-capex build.

Cash-flow anatomy. Operating cash flow is large and growing, but a step-change in capex has all but erased free cash flow:

	FY2023	FY2024	FY2025
Operating cash flow	$84.9B	$115.9B	$139.5B
Capex (gross)	~$53B	~$83B	$131.8B
Free cash flow	$36.8B	$38.2B	~$1.2B
FCF margin (FCF/revenue)	~6.4%	~6.0%	~0.2%

The FY2025 capex surge (AWS AI datacenters/Trainium plus logistics) compressed Amazon-defined free cash flow from ~$37-38B to roughly break-even despite record operating cash flow. (Amazon's reported FCF nets equipment finance-lease principal, so it sits below a simple OCF − capex.)

Balance sheet & capital allocation. Cash and marketable securities ~$123B against total equity ~$411B; Q1 2026 long-term debt rose to $122.6B (+$53.8B in the quarter) to fund the buildout, leaving a roughly net-neutral cash/debt position (A1/AA-). Amazon pays no dividend and buybacks are immaterial (no repurchases under the $10B authorization in the nine months to Sep 2025). Essentially all cash is being plowed into capex (~18% of revenue) rather than returned.

What drives it. The thesis reduces to whether AWS reacceleration (+28.4% Q1 2026) and retail-margin expansion generate an adequate return on the $130B+ capex bill before FCF normalizes — the question the tree resolves.

I. One-sentence verdict

Amazon at $270.64 is a three-platform conglomerate where AWS re-accelerated to +28.4% (Q1 2026, from +19.7% FY2025 — softening the scaffold's "share-erosion-to-Azure" concern), Advertising ($68.6B, +22%) is bigger than modeled, and the SOTP-blindness thesis remains partially falsified (Stage-3 Branch C: Advertising already implied at a Google/Meta peer multiple) — BUT the $131.8B capex (+59%) collapsed FCF to ~$1.2B and forced +$53.8B of new debt in Q1 2026, making the AI-capex IRR the binding question; re-anchored at $270 (the scaffold's $220 was a +23% FY2024-era understatement), the asymmetry is 1.61× favorable (+10.4% EV) with the Street PT $312.79 (+15.6%) — a modest improvement over the scaffold's read, suitable as a Watch-only / opportunistic 1-2% starter (the capex-FCF collapse + still-falsified SOTP-magnitude + 19/25 durability cap conviction), scaling to 3% on FCF recovery + Trainium customer #2 + AWS sustaining 28%. MSFT remains the cleaner cloud-AI exposure at the same date.

II. Company snapshot

Amazon.com, Inc. is the world's third-largest company by market capitalization (~$2.91T), operating three reported segments: North America ($426.3B FY2025 / $29.6B operating income, 6.9% margin), International ($161.9B / $4.75B OI, 2.9%), and AWS ($128.7B, 18% of revenue but 57% of operating income at 35.4% margin, +19.7% FY2025 growth accelerating to +28.4% in Q1 2026).

The company is operationally a three-platform conglomerate: AWS cloud + AI infrastructure (the economic engine), Amazon Advertising ($68.6B revenue, +22%, ~70% margin, hidden inside e-commerce segments — the structural-blindness layer), and E-commerce + logistics (the customer-acquisition engine + Prime flywheel anchor). The market applies a single blended multiple (~30× forward PE); the H-0 thesis tested whether SOTP would re-allocate that value — Stage 3 found it would re-allocate but not magnify in aggregate (Branch C, partially falsified).

FY2025 key financials (Tier-A): Total net sales $716.9B (+12.4%), operating income $80.0B (11.2%), net income $77.7B, GAAP diluted EPS $7.17, operating cash flow $139.5B, capex $131.8B (+59%), FCF collapsed to ~$1.2B (from $25.9B) — the AI buildout consumed it. Q1 2026 took on +$53.8B of new debt (total LT debt $122.6B) to fund the capex. No dividend.

III. The five facts that drive everything

AWS at 35.4% OPM contributes 57% of consolidated operating income on 18% of revenue (FY2025 Tier-A). This is the most important sentence about Amazon. Revenue center-of-gravity (e-commerce) diverges from economic-profit center-of-gravity (AWS) by a factor of ~3×. ✅A
Amazon Advertising at $68.6B revenue (+22% FY2025) and ~70% estimated OPM contributes ~$48B operating income equivalent — comparable to AWS in absolute dollars — yet is embedded in segment reporting with NO separate disclosure. The structural-blindness mechanism (disclosure, not valuation) is real. ✅A
HOWEVER, Stage 3 Branch C quantification shows the SOTP discount is NOT 30-50% as H-0 claimed. Reverse-engineering peer multiples (AWS at 25× EV/EBITDA; e-commerce at 0.5× EV/Revenue) yields implied Advertising EV/Revenue of 12-18× — comparable to Google/Meta, NOT discounted. The market appears to be pricing the SOTP in aggregate already; what's mispriced is the attribution, not the total. ✗A
AWS RE-ACCELERATED to +28.4% in Q1 2026 (from +19.7% FY2025) — the "share erosion to Azure" narrative is softening. AWS OI margin expanded to 37.7% in Q1 2026. Azure (+40%) still grows faster, but AWS re-accelerating undercuts the scaffold's structural-share-loss concern. ⚠️A (improved)
FY2025 capex $131.8B (+59%) collapsed FCF to ~$1.2B (from $25.9B) — the IRR test of the decade, INTENSIFIED. Amazon took +$53.8B of new debt in Q1 2026 to fund it. AWS OI margin held 35.4% (FY2025) / 37.7% (Q1 2026) despite the ramp. But the 2022 fulfillment-overshoot precedent (capex over-build by 12-18 months) is the binding caution — this is the bear's core datapoint. ⚠️A

IV. The H-0 thesis

H-0 (one sentence): Amazon is mispriced as a unified e-commerce-and-cloud conglomerate when it is actually three distinct platform businesses (AWS AI-infrastructure, Amazon Advertising, E-commerce logistics) each deserving its own peer multiple — but the market's structural failure to price Advertising separately leaves $300-600B† in value invisible on a blended-multiple basis.

Mispricing taxonomy: Structural blindness × Category (per mispricing.md).

H-0 confidence post-R2: 50% (scaffold 48%; up modestly as AWS RE-ACCELERATED to +28.4%, but the Branch-C SOTP-magnitude falsification still caps it). Stage 3 Branch C quantification falsified the headline magnitude of the SOTP gap; the structural blindness exists at the disclosure level but the valuation gap is materially smaller than the H-0 claim. The capex/FCF compression (FCF ~$1.2B) is the new offsetting concern. Branch A (AWS durability) and Branch E (risk calibration) came in strong (75% and 80% sub-confidence); Branch B (Advertising platform economics) at 60%; Branch C (SOTP quantification) at 30%; Branch D (Trainium optionality) at 45% (pre-inflection).

Falsification conditions (any one breaks the long-term thesis):

FF1 AWS growth decelerates below 14% CC for 2 consecutive quarters → infrastructure thesis broken
FF2 AWS OPM compresses >300 bps YoY for 2 consecutive quarters → capex ROI failing
FF3 FTC remedy scope expands to include Advertising self-preferencing → ad-platform impairment
FF4 Trainium remains single-customer (Anthropic only) through FY2027 → AI silicon thesis failure
FF5 Azure AI workload share gains accelerate dramatically (Azure >$25B AI annualized while AWS <$20B) → structural share-loss

V. Tree — five branches

H-0: Amazon is mispriced as a unified conglomerate when it's three distinct platforms;
     the SOTP-blindness on Advertising leaves $300-600B† in hidden value
     [REVISED post-Stage 3: aggregate SOTP ≈ market cap; mispricing is attribution, not magnitude]
│
├── L1A — AWS Platform Durability  ✅A supported with Trainium-customer caveat
│   ├── A.1.1  AWS ≥15% sustained ≥3 quarters             ✅A strongly supported (17%+ FY24 + Q1 25)
│   ├── A.1.2  AWS OPM ≥33% through capex cycle           ✅A 37% FY24 with buffer
│   ├── A.1.3  Trainium hyperscale customers ≥2          ⊗A Anthropic confirmed; #2 unverified
│   └── A.1.4  AI TAM growing faster than AWS              ⚠️A TAM growing; Azure taking AI mix share
│
├── L1B — Amazon Advertising Platform Economics  ⚠️A growth confirmed; platform quality opaque
│   ├── B.1.1  Ads ≥15% growth ≥3 quarters               ✅A 18-20% sustained
│   ├── B.1.2  Ads ROAS superiority confirmed             ⚠️A Amazon-marketing-claim only (M4)
│   ├── B.1.3  Ads CPM trending up                        ⚠️A directionally plausible; not verifiable
│   └── B.1.4  Ads advertiser diversity                   ⚠️A structurally multi-side; not quantified
│
├── L1C — SOTP Mispricing Quantification  ✗A HEADLINE THESIS FALSIFIED
│   ├── C.1.1  Implied Ads EV/Revenue <7×                ✗A lands 12-18×; SOTP gap ≤20% not 30-50%
│   ├── C.1.2  Sell-side SOTP models exist                ⚠️A likely no major SOTP; weak test
│   └── C.1.3  Management discloses Ads OPM               ✗A no disclosure (structural blindness intact at reporting layer)
│
├── L1D — Captive AI Silicon Optionality  ⊗A pre-inflection; option value real but not crystallized
│   ├── D.1.1  Trainium customers ≥2                     ⊗A Anthropic only confirmed
│   ├── D.1.2  External Trainium pricing published        ✅A EC2 Trn1/Trn2 publicly priced
│   └── D.1.3  Trainium cost-per-FLOP ≥30% below H100    ⚠️A Amazon-claim-only (M4)
│
└── L1E — Risk Calibration (Regulatory + Capex)  ✅A strongly supported
    ├── E.1.1  FTC scope excludes AWS/Ads                ✅A complaint scoped to marketplace conduct
    ├── E.1.2  Management AI demand backlog explicit      ✅A Jassy: "$100B+ demand backlog"
    └── E.1.3  AWS OPM compression <300 bps YoY           ✅A OPM expanding, not compressing

Verdict tally: 7 ✅ · 6 ⚠️ · 2 ✗ · 2 ⊗

VI. Market consensus and the orphan assets

What consensus says: AWS is the AI infrastructure backbone; Amazon Advertising is a high-margin "free upside" inside the e-commerce flywheel; capex is demand-driven; FTC trial 2026 is a known overhang. Sell-side mostly Buy/Overweight (~80% bullish†). 12-month consensus target ~$240-260†.

What consensus misses (partially — Stage 3 revised this):

Miss 1 (revised) — The SOTP attribution, not magnitude, is what's mispriced. Pre-Stage 3, the H-0 claim was that Advertising's $39B OI-equivalent at platform multiple represents $300-600B in hidden value. Stage 3 quantification showed that the aggregate market cap approximately equals the aggregate SOTP — the market is intuitively pricing the platforms correctly, just without explicit attribution. The disclosure catalyst (Amazon voluntarily breaking out Advertising) would reallocate value across segments and reduce uncertainty discount (estimated +5-15% re-rating), not unlock $300-600B of hidden value. This is the most important finding in the tree and revises the upside-magnitude expectation downward.

Miss 2 — AWS share is slowly eroding to Azure. AWS at 32% global IaaS share down from 33% in 2019, with Azure +1-2 pp/yr. For AI workloads specifically, Azure's OpenAI exclusivity gives MSFT a structural advantage in enterprise AI procurement (Azure → M365 → Teams → Azure OpenAI Service is a coherent enterprise AI bundle). AWS's response (Trainium + Bedrock + Anthropic stake) is real but the second-customer inflection (D.1.1) has not yet confirmed. The "AWS AI dominance" narrative may already be wrong.

Miss 3 — The 2022 fulfillment-overshoot precedent. Bears focus on capex magnitude ($100B+) and discipline questions; bulls focus on AI demand backlog. The 2022 pattern (capex ahead of demand absorption, 18 months of margin compression) is the historical analog the market is partially discounting but underweighting. The Q3 durability score (capital allocation, 3/5) reflects this concern.

The orphan asset math (revised after Stage 3 Branch C):

Amazon Advertising standalone at META-equivalent 24-25× PE on ~$39B OI = ~$940-975B implied; embedded in current market cap somewhere between $600-1,030B depending on AWS multiple assumption — the "hidden value" is at most $50-300B (not $300-600B as H-0 claimed)
Trainium optionality: $50-150B† if 2nd customer confirms; $0-50B if it doesn't — genuine optionality, not core value
Kuiper / Alexa+ / healthcare: $0-50B† combined option value; pre-scale

VII. Scenario analysis

Scenario	Prob	Multiple	FY2027E EPS†	Target price	Return from $270.64
Bull — AWS sustains 28%+; advertising scales; capex IRR proves	25%	~33×	$11.00†	$360†	+33.0%
Base — Compounder continues; capex weighs; SOTP attribution unchanged	50%	~30×	$10.25†	$310†	+14.5%
Bear — Capex IRR disappoints; AWS decel; FTC scope expands	25%	~24×	$9.00†	$215†	−20.5%

Expected value: $298.75† (+10.4% from $270.64) Asymmetry: +$89.36 bull / −$55.64 bear = 1.61× FAVORABLE

The re-anchor: the scaffold's $220 anchor (FY2024-era) was +23% low; FY2025 actuals are a year further along ($716.9B net sales). At the real $270, AWS re-accelerating to +28.4% improves the read (EV +10.4%, Street PT $312.79 +15.6%) — but the $131.8B capex / ~$1.2B FCF collapse is the binding risk. The bear is a genuine −20.5%, not a benign floor. Still a Watch-only / opportunistic-starter, not a conviction buy — MSFT remains the cleaner cloud-AI exposure.

VIII. Risks

Valuation risk (Medium): At 31× forward PE, AMZN is priced for continued AWS growth + Advertising margin maintenance + AI capex absorption. Not bubble territory but not quality-floor pricing either. Bear scenario implies ~28× PE on slightly lower EPS — meaningful compression.

Execution risk (Medium-High): $100B+ AI capex cycle is unprecedented in scale. Historical precedent (2022 fulfillment overshoot) argues for caution. AWS RPO disclosure gap (vs. MSFT's $250B+ committed backlog) is an information asymmetry that increases uncertainty.

Competition risk (Medium): Azure share gains in AI workloads specifically. Google Cloud at 12% remains a strong third. TikTok Shop + Temu + Shein at e-commerce low-price tier.

Technology risk (Medium): Trainium vs. NVIDIA CUDA-ecosystem moat is the long-term competitive question. AMD MI400 enters market; alternative AI silicon (Groq, Cerebras) commoditization risk.

Regulatory risk (Medium): FTC trial 2026; remedy scope is the key variable. Marketplace-only remedy is most likely (bounded impact ~$3-5B OI); scope expansion to Advertising is tail risk (~15-20% probability) with high magnitude.

Hype risk (Low-Moderate): At $270 (~30× forward), AMZN re-rated +23% from the scaffold's $220 read but is not in clear hype territory — the multiple is supported by AWS +28.4% + advertising +22%. The crowded-AI-capex-trade risk is the watch: a capex-IRR disappointment (FCF already ~$1.2B) would compress the multiple.

IX. Historical analogues

Analogue 1 — Amazon itself, 2022 fulfillment overshoot. Amazon doubled fulfillment capacity in 2020-2021 to serve COVID-era demand. By Q2 2022, demand had normalized but the capex was committed. Stock fell ~50% peak-to-trough. The 18 months of margin compression and FCF disappointment finally absorbed by 2023-2024 as e-commerce volume caught up. Lesson: large capex cycles can over-build by 12-18 months even when management commitment is sincere; the J-curve recovers but the path is volatile.

The current analog: $100B+ AI capex in FY2025 may follow a similar pattern if AI workload absorption is slower than the demand backlog suggests. Cautionary precedent for sizing.

Analogue 2 — Microsoft 2014-2018 cloud transition (positive case). MSFT spent 4 years rebuilding around Azure + Office 365 subscription before the market re-rated. ARR grew under the surface while GAAP revenue lagged. Patient investors who understood the cloud thesis were rewarded 5×+ over the following decade. Amazon's AI infrastructure investment may follow a similar pattern — AI ARR (committed compute contracts) growing ahead of GAAP revenue recognition.

The difference: MSFT was building from a sub-leader position; AMZN is defending a leader position against a faster-growing challenger (Azure). The leader-defending-vs-share-gainer dynamic is structurally less attractive than the share-gainer position.

Analogue 3 — Alphabet 2019 YouTube disclosure (catalyst comparison). Google began disclosing YouTube ad revenue separately in Q4 2019 / FY2020 10-K. The disclosure event triggered a modest one-time re-rating of ~5-10% as analysts updated models to value YouTube against video-advertising peers. Note: the re-rating was modest, not transformative. This is the closest empirical anchor for the magnitude of the Amazon Advertising disclosure catalyst — supportive of the Stage 3 Branch C finding that the SOTP "hidden value" is in the +5-15% re-rating range, not +25-30%.

X. The structural blindness — disclosure vs. valuation

The Stage 3 finding deserves its own section because it overturns the headline H-0 thesis.

H-0 claimed: Amazon Advertising is a $56B revenue / $39B OI business hidden inside e-commerce segments, deserving a Google/Meta-equivalent platform multiple — implying $300-600B† in hidden value the blended multiple obscures.

Stage 3 Branch C tested this quantitatively. Reverse-engineering the current market cap (~$2.35T) under three plausible peer-multiple combinations:

AWS at 25× EV/EBITDA + e-commerce at WMT-style 0.5× EV/Revenue → residual Advertising EV ~$1.03T → 18× EV/Revenue
AWS at 25× EV/EBITDA + e-commerce at Amazon-historical 1.0× → residual Ads ~$700B → 12.5× EV/Revenue
AWS at 30× EV/EBITDA + e-commerce at 0.7× → residual Ads ~$730B → 13× EV/Revenue

Every plausible decomposition lands Advertising's implied EV/Revenue at 12-18×. Compare to:

GOOGL implied advertising EV/Revenue: ~10-12×
META implied advertising EV/Revenue: ~8-10×

Amazon's implied Advertising multiple is ALREADY at or above the peer set. The market, intuitively, is not applying a deep discount to Advertising — it's just not making the attribution explicit.

What this means for the H-0 thesis:

The disclosure structural blindness exists (Advertising IS NOT a reportable segment — this is factual)
The valuation structural blindness does NOT exist at the magnitude H-0 claimed
The catalyst (segment separation) would reallocate value across segments and reduce uncertainty discount, generating an expected +5-15% re-rating, not +25-30%

What's still mispriced (revised thesis):

Attribution uncertainty — analysts cannot model AWS vs. Advertising vs. e-commerce separately, generating a higher cost of equity for AMZN than fully-disclosed peers (MSFT, GOOGL)
Trainium optionality — genuinely not in any model; magnitude depends on hyperscale customer #2 inflection
Alexa+ / Kuiper / healthcare optionality — pre-scale, no analyst attribution

The revised mispricing magnitude: $100-250B of potential re-rating value on full catalyst sequence, primarily from reduced uncertainty discount and AI-silicon optionality crystallization — meaningful but not the $300-600B the H-0 thesis implied.

XI. What the durability test found

AMZN scored 19/25 on the long-term durability test — Medium-High durability, middle of the band.

The strength: Q1 (business model persistence) scored 5/5. Three-platform structure (cloud + e-commerce + advertising) is structurally durable for 10+ years. AI is a tailwind to all three.

The concerns: Q2 (moat trajectory) and Q3 (10-year capital allocation) both scored 3/5 — the load-bearing concerns.

Q2 mixed: Prime + Advertising flywheel widening, fulfillment network widening, BUT AWS share slowly eroding to Azure, 3P marketplace under regulatory pressure, Trainium counter-moat pre-inflection.

Q3 mixed: AWS foundational investment was the best capital allocation decision in modern technology history, but the 2022 fulfillment overshoot is a documented cautionary precedent, Whole Foods and MGM returns mediocre, and the 2025 AI capex cycle ($100B+) is the open IRR question. ROIC averaged ~13% over the decade — above WACC ~8.5%, but with material volatility (5% in 2015, 9% in 2022 — both barely above WACC).

Fatal flag check: 0 flags fired. ROIC above WACC on 10-year aggregate (Q3 fatal flag does not fire); disruption survival is 4/5 (Q4 fatal flag does not fire); balance sheet is A1/AA- with ~$101B liquidity (no solvency risk). But Q2/Q3 at 3/5 each are adjacent to concern thresholds — tree_v2 verdict could move these to 2/5 if the 2025 AI capex cycle generates below-WACC returns.

Comparison to MSFT (same date evidence quality):

MSFT durability: 21/25 (Q1=5, Q2=4, Q3=4, Q4=4, Q5=4)
AMZN durability: 19/25 (Q1=5, Q2=3, Q3=3, Q4=4, Q5=4)
MSFT scores higher on moat trajectory (OpenAI exclusivity + Entra ID widening) and capital allocation (Nadella record exceptional)
AMZN's three-platform structure is more diversified but less durable per individual moat

XII. Investment Scorecard (Format A — Long-term hold, 15 questions)

Per MANUAL Part K.6 long-term-hold extended checklist.

Q	Question	Tier	Wt	Verdict	Contribution
Q1	Do I understand what Amazon actually does?	Confirming	1	✅A	1.0
Q2	Will this business model still matter in 10 years?	Critical	5	✅A (three-platform structure structurally durable; AI tailwind)	5.0
Q3	Is the moat widening or eroding?	Load-bearing	3	⚠️A (Prime+Ads widening; AWS share + 3P eroding; net mixed)	1.5
Q4	Is the capital allocation grade A or B?	Load-bearing	3	⚠️A (AWS foundational = A+; Whole Foods/MGM = C; 2022 overshoot precedent for 2025 capex)	1.5
Q5	Does it survive the 3 named disruption threats?	Load-bearing	3	✅A (Azure-share / TikTok-Shop / FTC — none is 10×; three-platform diversifies)	3.0
Q6	Is reinvestment runway >5 yrs at ROIC > WACC?	Important	2	⚠️A (opportunity set exceeds FCF 1.6-1.9×; J-curve compression; IRR uncertain)	1.0
Q7	Is there upside optionality not priced?	Important	2	✅A (Trainium + Alexa+ + Kuiper + healthcare; $100-250B+ option value)	2.0
Q8	Is the balance sheet safe?	Critical	5	✅A ($101B liquidity; A1/AA-; net debt small; $38-65B† FCF base)	5.0
Q9	Is insider alignment healthy?	Important	2	⚠️A (Bezos 9.5%† but systematic sell program; Jassy <0.1%; no major insider buying)	1.0
Q10	Is valuation reasonable on 5-yr forward basis?	Load-bearing	3	⚠️A (31× forward PE; SOTP ≈ market cap at base case; bear scenario -21%)	1.5
Q11	Is the dividend secure and growing?	Important	2	✗ (no dividend; not income-suitable)	0.0
Q12	Are there any fatal flags fired?	Critical	5	✅A (0 fatal flags; ROIC>WACC 10yr aggregate; no solvency risk)	5.0
Q13	Is the position size appropriate for conviction level?	Confirming	1	✅A (Watch-only / opportunistic 1-2% starter at $270; matches 50% H-0 + 19/25 durability + capex-FCF concern)	1.0
Q14	Are sell triggers documented?	Confirming	1	✅A (RF1-RF6 documented; FF1-FF5 falsification written)	1.0
Q15	Has a second opinion been sought on the key uncertainties?	Confirming	1	✅A (R2 COMPLETE — FY2025 10-K + Q1 2026 10-Q Tier-A; AWS re-accelerated +28.4%, capex $131.8B / FCF ~$1.2B verified; Branch C SOTP-magnitude finding confirmed)	1.0

K.3.5 Weighted-score derivation

Tier	Questions	Raw weighted contribution
Critical (5×)	Q2✅(5.0) + Q8✅(5.0) + Q12✅(5.0)	15.0
Load-bearing (3×)	Q3⚠️(1.5) + Q4⚠️(1.5) + Q5✅(3.0) + Q10⚠️(1.5)	7.5
Important (2×)	Q6⚠️(1.0) + Q7✅(2.0) + Q9⚠️(1.0) + Q11✗(0.0)	4.0
Confirming (1×)	Q1✅(1.0) + Q13✅(1.0) + Q14✅(1.0) + Q15✅(1.0)	4.0
TOTAL		30.5 / 39 = 78% (raw; Tier-C haircut retired — evidence now Tier-A)

Note: Q11 (dividend) scores 0 because Amazon does not pay a dividend. For a non-income-focused mega-cap growth investment this is not a defect but it does compress the K.3.5 score relative to dividend-paying peers (MSFT scored 2.0 on Q11). If Q11 is excluded as not-applicable, the adjusted score is 30.0 / 37 = 81%. The 77% raw score is the conservative reading; 81% adjusted is the apples-to-MSFT-comparable read.

77% raw / 81% adjusted (Moderate-to-high-conviction band) per K.3.5 banding (≥85% high, ≥65% moderate, ≥45% wait/skip).

K.3.1 fatal-flag override check: No fatal flags fired. Q3/Q4 at 3/5 weighted (1.5 each on contribution) are below ideal but not approaching the 1/5 override threshold. No override applies.

Evidence tier: R2-VERIFIED (Tier-A on FY2025/Q1-2026 financials; Tier-B on price). The Tier-C scaffold haircut is retired (Q15 ⚠️→✅). The 78% raw / 82% adjusted (ex-Q11 dividend) reflects a genuinely durable three-platform business whose moderate conviction is set by the Branch-C SOTP-magnitude falsification + the now-acute capex/FCF question (FCF ~$1.2B), not by evidence quality.

Scorecard summary

Weighted score: 78% raw / 82% adjusted (Moderate-conviction band, Tier-A)

Verdict: Watch-only / opportunistic 1-2% starter at $270 — AWS re-accelerating (+28.4%) and Street PT $312.79 (+15.6%) improve the case vs the scaffold's Watch-only; scale to 3% on FCF recovery + Trainium customer #2 + AWS sustaining 28%. Do not chase above ~$310 (Street PT) without FCF recovery evidence — the $131.8B capex / ~$1.2B FCF collapse is the binding risk.

Why not "Buy" (scale to 3-5%): H-0 is 50% (the SOTP-magnitude thesis stays partially falsified), durability is 19/25 (lowest of the mega-caps), and the capex-IRR question is now acute (FCF collapsed to ~$1.2B; +$53.8B new debt). The 1.61× asymmetry + AWS acceleration justify an opportunistic starter, not conviction sizing.

Why not "Avoid": The three-platform model is genuinely durable (Q1=5/5), AWS RE-ACCELERATED to +28.4% (the share-erosion concern is softening), Advertising is bigger than modeled ($68.6B), the balance sheet absorbs the capex (A1/AA-), and Street is Strong Buy with a $313 PT (+15.6%). AMZN is investable — just not at conviction sizing while FCF is suppressed.

Why MSFT > AMZN at the same date: MSFT (21/25 durability, 90% XII score, 2.84× asymmetry, RPO +99% backlog visibility) has cleaner cloud-AI exposure and a doubling backlog that AMZN's opaque AWS RPO doesn't match. Capital flows to MSFT first; AMZN is the secondary opportunistic position.

2-minute pitch

Why I'm watching Amazon (not buying yet) (2-minute version for a sophisticated investor):

Amazon is a three-platform conglomerate where AWS (37% margin, 17% growth, $108B revenue) drives ~58% of operating income, Amazon Advertising (~70% margin, 18% growth, $56B revenue) is roughly co-equal in OI dollar terms despite being hidden in segment reporting, and the e-commerce + logistics flywheel is the customer-acquisition engine that feeds both high-margin platforms. The headline mispricing thesis — that Advertising is hidden at a deep discount inside the blended multiple — turned out to be partially wrong on Stage 3 quantification: reverse-engineering plausible peer multiples shows the market is already pricing Advertising at a Google/Meta-equivalent multiple in aggregate, just without explicit attribution. The disclosure catalyst (Amazon separating Advertising as a reportable segment) would reallocate value across segments and reduce uncertainty discount, generating maybe +5-15% re-rating — not the +25-30% the original thesis implied.

What remains attractive: a Medium-High durability (19/25) compounder with $100-250B of genuine optionality (Trainium AI silicon, Alexa+, Kuiper) NOT in current models, AWS as a structurally durable cloud platform even if slowly losing share to Azure, and a balance sheet that can absorb the $100B+ AI capex cycle without solvency stress.

What concerns me: the 2022 fulfillment-overshoot precedent argues that big-capex cycles can over-build by 12-18 months even when management commitment is sincere. AWS share is slowly eroding to Azure (32% → potentially 28% over 5 years if Microsoft's enterprise lever continues). The FTC trial overhang in 2026 is a low-probability but high-magnitude tail risk if scope expands to Advertising.

At the real $270 the asymmetry is 1.61× (modest-favorable); the bear case is a real −20.5% downside, not a benign floor. AWS re-accelerating to +28.4% and the Street PT $312.79 (+15.6%) improve the read vs the scaffold — so I would take an opportunistic 1-2% starter here, scaling to 3% on FCF recovery + Trainium customer #2 + AWS sustaining 28%. I would NOT chase above ~$310 (Street PT) without FCF-recovery evidence — the $131.8B capex / ~$1.2B FCF collapse is the binding risk. MSFT (same date) offers cleaner cloud-AI exposure with a doubling RPO backlog; capital flows there first.

Risk types (K.4)

Primary risks in this investment:

Execution risk (Medium-High): $100B+ AI capex IRR realization; 2022 fulfillment-overshoot precedent
Competition risk (Medium): AWS share erosion to Azure; TikTok Shop / Temu at e-commerce low-price tier
Regulatory risk (Medium): FTC trial 2026 outcome scope; EU DMA compliance
Technology risk (Medium): Trainium vs. NVIDIA CUDA-ecosystem battle; AMD MI400 emergence
Valuation risk (Medium): 31× forward PE is above 5yr median; bear case implies 28× — meaningful compression
Hype risk (Low-Moderate at $270): not bubble, but AI-capex trade is crowded; FCF-collapse compounds multiple-compression risk

"When NOT to buy" anti-patterns (K.5)

Do not buy AMZN if:

You are buying because "everyone says AI infrastructure will moon" — AWS growth is structurally exposed to Azure's share gains in AI workloads; not a clean AI-infrastructure long
You are buying expecting a quick Advertising-disclosure re-rating — Stage 3 Branch C showed the re-rating magnitude is +5-15%, not +25-30%; the headline thesis is partially falsified
You are buying above $260 without T1 AND T2 catalyst confirmation — the bear case becomes the dominant scenario at full-bull pricing without evidence
You are buying expecting dividend income — Amazon does not pay a dividend; income investors should look elsewhere (MSFT, COST, F)
You are allocating >3% without bear-case pricing or dual-catalyst confirmation — durability score (19/25, middle of Medium-High band) caps the position type; the fundamentals don't justify a concentrated bet at current evidence quality
You are already overweight cloud-infrastructure via MSFT/GOOGL — adding AMZN at 3% on top of MSFT 3% creates 6% cloud-AI concentration; correlated exposure
You assume the $100B+ AI capex cycle will replicate AWS's foundational 2006-2015 ROIC pattern — Amazon's capex track record is mixed; the 2022 fulfillment overshoot is the more recent and relevant precedent

Financial figures are R2-verified Tier-A (FY2025 10-K accn 0001018724-26-000004; Q1 2026 10-Q accn 0001018724-26-000014, SEC EDGAR). Price/multiples Tier-B (2026-05-29). Forward scenario targets (†) are analytical constructs. Supersedes the 2026-05-13 GENERATE-mode scaffold; see update_2026-05-30.md. This report does not constitute financial advice. Ming Zhong makes his own investment decisions.

Built on 2026-05-13. The Stage 3 Branch C finding (SOTP magnitude smaller than H-0 claimed) is the load-bearing analytical revision in this tree; this finding alone should inform any future tree_v2 re-assessment.