1Executive Summary
We initiate coverage of Amazon.com (AMZN US, $213) with a BUY rating and a 12-month price target of $293, implying +37.6% upside. Our thesis rests on three mis-priced assets: (1) an AWS re-acceleration cycle powered by proprietary AI silicon that consensus under-models; (2) an advertising business generating roughly $69B of capital-light revenue that is valued inside a retail multiple; and (3) North America margin expansion driven by robotics, where execution is already running three quarters ahead of our model.
Our sum-of-the-parts assigns $1,865B to AWS (11.5× FY26E EV/Revenue), $498B to Advertising (6.0×), $703B to North America (20× operating income) and $96B to International — a gross enterprise value of $3,162B. After FY26E net debt of $24B and minority interests, equity value is approximately $3,136B, or $293 per share. Our bull case is $358 (+68% versus spot). Our bear case is $228, which is 22% below our price target but still around 7% above the current share price — the bear case is a capital misallocation scenario, not a collapse.
The single most important near-term catalyst is the Q1 2026 AWS earnings print in April 2026. If AWS sustains growth of 26% or better — our base case, against consensus of roughly 22% — a buy-side revision cycle should follow. Each percentage point of AWS growth is worth approximately $1.6B of incremental revenue, $600M of operating income and, at our 11.5× multiple, about $1.70 per share of fair value. The four-point gap between our estimate and consensus is therefore worth close to $6.9 per share if consensus converges fully; we assume partial convergence and size the April catalyst at approximately $5 per share.
2Company Overview & Segment Architecture
Amazon reported $716.9B in FY2025A net sales (+12.4% year on year) and $80.0B of operating income (+16.6%). The four business lines have materially different economic profiles, and that divergence is the foundation of the sum-of-the-parts case.
| Segment | Revenue ($B) | Growth | Op. income ($B) | Margin | Economic character |
|---|---|---|---|---|---|
| North America | 426.3 | +10.4% | 29.6 | 6.9% | Retail, third-party marketplace, Prime, embedded advertising |
| AWS | 128.7 | +19.7% | 45.6 | 35.4% | 57% of group operating income; capacity constrained |
| International | 161.9 | +13.2% | 4.7 | 2.9% | Early-stage profitability; 5–6pp FX tailwind |
| Total (reported) | 716.9 | +12.4% | 80.0 | 11.2% | — |
| Advertising* | 68.6 | +22.1% | ~56.9* | ~83%* | Not a reported segment — embedded within NA and International |
* Advertising is not a reported segment and Amazon does not disclose its operating income. The figures shown are a contribution margin estimated before allocation of the shared fulfilment, logistics and infrastructure cost that Amazon reports within North America and International. They are not comparable to segment operating income and are not additive to the $80.0B group total. See the limitation note in Section 5.
3Investment Thesis
Pillar 1 — AWS AI re-acceleration: the market under-models the capacity unlock
We model 26% AWS growth in FY2026E against consensus of approximately 22% — a revenue differential of roughly $5.2B measured on the FY2025A base. Three dynamics compound.
- Q4 2025 validated the thesis. AWS grew 24% year on year, the fastest rate in thirteen quarters, and management confirmed that demand exceeds available supply, describing the business as monetising capacity as fast as it can be installed. The backlog stands at $244B, up 40%, providing roughly eighteen months of revenue visibility with minimal credit risk.
- Custom silicon is a structural moat. Trainium2 and Trainium3 deliver an estimated 30–50% cost advantage against Nvidia H100/H200 for training workloads, and Graviton4 is broadly comparable for inference. That cost differential compounds as $200B of FY2026 capital expenditure builds scale. Combined Trainium and Graviton run-rate revenue is estimated above $10B with triple-digit growth.
- The 2026 capacity unlock is the near-term catalyst. AWS chief executive Matt Garman has stated that demand will exceed supply for the next couple of years. As new capacity comes online through 2026, the constraint on revenue recognition lifts, and growth should accelerate into the second half of 2026 rather than decelerate.
Each percentage point of AWS growth outperformance is worth approximately $1.6B of FY2026E revenue and $600M of operating income. Capitalised at our 11.5× EV/Revenue multiple, that is roughly $18.4B of enterprise value, or about $1.70 per share. The four-point gap between our 26% estimate and consensus at 22% is therefore worth close to $6.9 per share on full convergence.
Amazon shortened server useful lives in January 2025 as AI and machine-learning development cycles accelerated. Depreciation and amortisation rising from roughly 7% to 9.2% of revenue creates a meaningful headwind to earnings per share. If capacity growth outstrips demand, AWS margin could compress to 32–33% — which is precisely what our bear case assumes.
Pillar 2 — Advertising: a $500B asset buried in the retail P&L
At $68.6B of FY2025A revenue growing 22%, Amazon Advertising is the third-largest digital advertising platform globally. Because it is embedded in the retail profit-and-loss statement, it is valued at a retail multiple of roughly 1× EV/Revenue rather than the 6–8× that comparable media assets command. We believe this creates a structural undervaluation of approximately $415–581B.
- The margin profile is exceptional. We estimate a contribution margin of around 83%. Advertising is close to a pure revenue-recognition model once Amazon's first-party data infrastructure — funded by the retail business — is in place, and each incremental advertising dollar requires near-zero marginal capital.
- The data moat is structural rather than replicable. Amazon's closed-loop attribution, knowing which impressions drove which purchases in real time, is structurally superior to Google's search-intent signal and Meta's social signal. Better advertiser returns on ad spend sustain both pricing power and volume growth.
- Prime Video multiplies the opportunity. The average ad-supported audience is 315 million viewers across sixteen countries, up from 200 million in April 2024. Management confirmed that Prime Video contributed meaningfully to Q4 2025 advertising revenue. We size this as a $5–10B incremental annual opportunity that consensus has barely begun to model.
- Retail Ad Service extends the platform externally. Amazon is licensing its demand-side platform and attribution infrastructure to other retailers. A single major signing would confirm platform status and could trigger a re-rating from retail towards media comparables.
- Standalone valuation at 6.0× EV/Revenue is $498B. A re-rating to 7.0× — still below Meta's 7.5× — would add approximately $7.70 per share. We hold 6.0× as the base case to stay conservative.
Pillar 3 — North America margin expansion: execution ahead of model
We model North America operating margin expanding from 6.9% in FY2025A to 7.5% in FY2026E and 8.5% in FY2027E. Q4 2025 already printed 9.0% — three quarters ahead of our schedule — while consensus still models below 8%. That gap creates a persistent positive surprise cycle.
- Robotics is the primary structural driver. Amazon deployed its one-millionth robot in July 2025 and is targeting automation of roughly 600,000 fulfilment centre roles. Internal documents suggest this could reduce cost-to-serve by around 30 cents per item by 2027. Every 100 basis points of North America margin is worth approximately $4.7B of operating income and, at our 20× multiple, about $8.75 per share.
- Advertising mix shift is a passive tailwind. As advertising revenue grows faster than first-party retail, the blended North America margin improves with no operational change required.
- Same-day delivery economics compound. Volumes doubled year on year in 2025. Route density improvements are driving unit economics towards profitability, and Amazon is targeting more than 2,300 communities for same-day perishable delivery.
- Amazon Leo is transient noise. Roughly $1B of incremental year-on-year cost falls in Q1 2026. We treat this as a deferred automation investment rather than a structural headwind.
4Financial Model Summary
The model is built bottom-up by segment. All FY2023–FY2025 actuals are verified against Amazon's published earnings filings.
| Metric | FY23A | FY24A | FY25A | FY26E | FY27E | FY28E | CAGR* |
|---|---|---|---|---|---|---|---|
| AWS revenue | 90.8 | 107.6 | 128.7 | 162.2 | 201.1 | 245.3 | 24% |
| AWS op. margin | 26.9% | 37.4% | 35.4% | 35.5% | 36.5% | 37.5% | — |
| Advertising revenue | 46.9 | 56.2 | 68.6 | 83.0 | 101.3 | 121.5 | 21% |
| North America revenue | 352.8 | 386.1 | 426.3 | 468.9 | 515.8 | 562.2 | 10% |
| NA op. margin | 5.2% | 6.5% | 6.9% | 7.5% | 8.5% | 9.2% | — |
| International revenue | 131.2 | 143.0 | 161.9 | 182.9 | 204.9 | 229.5 | 12% |
| Total revenue | 574.8 | 638.0 | 716.9 | 814.0 | 921.8 | 1,037.1 | 13% |
| Total operating income | 36.9 | 68.6 | 80.0 | 99.1 | 126.5 | 156.3 | 25% |
| Consolidated margin | 6.4% | 10.8% | 11.2% | 12.2% | 13.7% | 15.1% | — |
| EBITDA | 71.9 | 112.1 | 145.7 | 176.4 | 215.9 | 257.9 | — |
| Net income | 30.4 | 59.2 | 77.7 | 85.7 | 108.7 | 134.3 | — |
| Diluted EPS ($) | 2.90 | 5.53 | 7.17 | 7.98 | 10.12 | 12.50 | 20% |
| Consensus EPS ($) | — | — | — | 8.16 | 9.90 | 12.78 | — |
| vs. consensus | — | — | — | −2.2% | +2.2% | −2.2% | — |
| Capital expenditure | — | 83.0 | 131.8 | ~200E | ~180E | ~150E | — |
| Operating cash flow | — | 115.9 | 139.5 | ~155E | ~195E | ~250E | — |
* Compound annual growth rate, FY2025A–FY2028E. FY2025A diluted EPS of $7.17 and operating income of $80.0B verified against the Q4 2025 press release. Depreciation and amortisation is restated to 9.2% of revenue ($65.7B) from 7.1%, reflecting the January 2025 shortening of server useful lives; FY2026E D&A is $77.3B on the same basis. Free cash flow turns negative in FY2026E on $200B of capital expenditure, recovering to approximately $100B by FY2028E as capital intensity normalises. Diluted EPS is calculated on period-weighted average diluted shares, which differ from the period-end counts in Appendix A.
5Valuation — Sum-of-the-Parts
We value Amazon on a sum-of-the-parts basis anchored to FY2026E. A blended consolidated multiple understates value because it averages a cloud and AI compounder, a capital-light media business and a physical logistics operation at a single rate — a structurally inappropriate approach for a group of this composition.
| Segment | FY26E metric | Multiple | Implied EV ($B) | % of EV | Peer benchmark |
|---|---|---|---|---|---|
| AWS | $162.2B revenue | 11.5× | 1,865 | 59.0% | MSFT ~13×, GCP ~9×; leadership discount applied |
| Advertising | $83.0B revenue | 6.0× | 498 | 15.7% | META ~7.5×, GOOGL ~6.0×; non-disclosure discount |
| North America | $35.2B op. income | 20× | 703 | 22.2% | WMT ~18×, COST ~25× EBITDA; growth premium |
| International | $6.4B op. income | 15× | 96 | 3.0% | EM retail; early-stage profitability discount |
| Gross enterprise value | — | — | 3,162 | 100% | — |
| Bridge | $B | Basis |
|---|---|---|
| Gross enterprise value | 3,162 | Table 3 |
| Less: FY26E net debt | (24) | FY25A net cash $21.2B ($86.8B cash less $65.6B long-term debt); FY26E outflow $45B (OCF $155B less capex $200B) |
| Less: minority interests | (2) | Estimated |
| Equity value | 3,136 | ÷ 10.73B diluted shares = $293 per share (+37.6%) |
Per-share value is calculated on 10.73B FY2025A diluted shares, giving $292.26, rounded to $293. On the FY2026E share count of 10.78B the implied value is $291. We use the FY2025A count for consistency with the current market capitalisation reference.
Advertising is valued here on revenue while North America is valued on operating income that still contains the advertising contribution. The two components are therefore not fully independent, and some profit is captured twice. We set the North America multiple at 20× — below the level its growth profile alone would justify — partly to absorb this overlap, but we do not claim the adjustment is precisely calibrated, and Amazon's non-disclosure of advertising segment economics makes precision impossible from outside. A reader who believes the overlap is larger than we have assumed should discount the North America component rather than the advertising one, since the advertising revenue figure is disclosed while the North America profit split is not.
| AWS ↓ / Ads → | 5.0× | 6.0× (base) | 7.0× | 8.0× |
|---|---|---|---|---|
| 10.0× | $263 | $271 | $278 | $286 |
| 11.0× | $278 | $286 | $293 | $301 |
| 11.5× (base) | $285 | $293 | $301 | $308 |
| 12.0× | $293 | $301 | $308 | $316 |
| 13.0× | $308 | $316 | $323 | $331 |
Cells are rounded to the nearest dollar and derived from the base case at ±$15.00 per 1.0× of AWS multiple and ±$7.70 per 1.0× of advertising multiple. The base-case cell is stated at the headline price target of $293; the unrounded figure is $292.26.
6Catalysts & Re-Rating Events
| Timing | Catalyst | Impact | Prob. | Investment logic |
|---|---|---|---|---|
| Apr 2026 | Q1 2026 AWS growth ≥26% | +~$5 | High | Full convergence of consensus to our 26% estimate is worth roughly $6.9 per share at our 11.5× multiple; we assume partial convergence. Azure showed the same revision pattern in 2023–24. |
| H1 2026 | Standalone Prime Video ad revenue disclosure | +~$4 | Med | Connected-TV multiple applied to a $5B-plus incremental stream; the advertising component re-rates towards media peers. |
| 2026 | Major external retailer signs Retail Ad Service | +~$4 | Med | Platform status confirmed; DSP network effects compound; addressable market re-priced from Amazon's own retail to total retail media. |
| Q3–Q4 2026 | NA operating margin ≥8% for two consecutive quarters | +~$7 | High | Sell-side estimates revised; robotics thesis validated; margin expansion capitalised. |
| 2026–27 | Fed cuts of 50–100bps cumulative | Multiple | Med | Long-duration asset re-rating; discount rate compression benefits FY2027–28 cash flows most acutely. |
Anatomy of the AWS revision cycle
When a hyperscaler prints an unexpected growth acceleration, consensus typically takes two to three quarters to reprice forward estimates fully. Microsoft Azure demonstrated this pattern in both 2023 and 2024: first-quarter beats were only partially modelled by consensus, producing sustained six to nine month windows of positive estimate revision. We expect AMZN to follow the same pattern from Q1 2026 onward if AWS meets our 26% growth assumption.
7Key Risks & Bear Case
Our bear case is $228 — 22% below our price target, though still roughly 7% above the current share price of $213. This is deliberately not a collapse scenario. It is a capital misallocation scenario, in which the capital expenditure programme proceeds but the demand to fill it arrives more slowly than management expects. It assumes AWS growth of 21% (consensus as at Q3 2025), AWS margin compressing to 32.5% and North America margin holding flat at 7.5%, producing diluted EPS of approximately $6.50 which, at roughly 35× earnings, gives $228.
| Risk | Severity | Prob. | Description | Monitor / downgrade trigger |
|---|---|---|---|---|
| Capex without demand | High | Med | $200B of FY2026 capex creates a D&A burden of approximately $77B, or 9.5% of revenue. If AWS growth stays below 22%, overcapacity compresses margin to 32–33% and free cash flow turns deeply negative. Price target falls to $255–265. | Q1 2026 AWS growth rate; revisions to management capex guidance |
| AWS margin compression | Med | Med | Shorter server useful lives combined with high capex means rising D&A. If AWS margin falls below 33%, the AWS component of our SOTP drops by $250–300B, or $23–28 per share. | Quarterly AWS operating margin disclosure |
| AI commoditisation | Med | Low | Open-source models erode AI pricing power. Google Cloud is growing faster in percentage terms from a smaller base. If AWS loses AI workload share, growth re-rates towards an 18% run-rate. | AWS versus GCP and Azure share data; pricing moves |
| FX reversal | Low | Med | The dollar fell roughly 10% in 2025 and 5–6pp of International's 13% growth is translation. Dollar recovery would decelerate reported International growth to 7–8%. | DXY trend; International organic ex-FX disclosure |
| Regulatory / antitrust | Low | Low | FTC marketplace investigation and EU DMA compliance costs. Forced unbundling of Prime or of advertising data practices would be material. | FTC and EU enforcement docket |
| Consumer boycott | Low | Low | The February 2025 boycott showed minimal impact, with footfall unchanged per Placer.ai. Prime switching costs are high and sustained structural churn would require twelve months or more. | Prime subscriber count; NA growth in H1 2026 |
| Scenario | AWS growth FY26E | AWS margin FY26E | NA margin FY26E | Implied PT | vs. PT | vs. spot |
|---|---|---|---|---|---|---|
| Bear | 21% | 32.5% | 7.5% | $228 | −22% | +7% |
| Base (our estimate) | 26% | 35.5% | 7.5% | $293 | — | +37.6% |
| Bull | 28% | 37.0% | 8.5% | $358 | +22% | +68% |
8Comparable Company Analysis
Amazon's consolidated valuation understates the quality of its constituent businesses because retail dilutes the blended multiple. This is the central argument for using a sum-of-the-parts methodology rather than a single group multiple.
| Company | NTM rev. growth | EV/NTM rev. | NTM P/E | Op. margin | Relevance to the AMZN SOTP |
|---|---|---|---|---|---|
| Microsoft (MSFT) | +13% | 12.8× | 31× | 45% | Azure peer — anchor for the AWS multiple |
| Alphabet (GOOGL) | +11% | 6.0× | 22× | 32% | Advertising and cloud dual comparable |
| Meta Platforms (META) | +17% | 7.5× | 23× | 42% | Digital advertising comparable — anchor for the ads multiple |
| Walmart (WMT) | +4% | 0.9× | 28× | 5% | North America retail comparable |
| Costco (COST) | +7% | 1.6× | 50× | 4% | Membership retail comparable |
| AMZN — consolidated* | +14% | 2.9× | 26.7× | 12.2%E | Blended; understates the AWS and advertising assets |
* Consolidated EV/Revenue is calculated on a market capitalisation of $2,285B plus FY2026E net debt of $24B, divided by FY2026E revenue of $814.0B. NTM P/E is the $213 share price divided by our FY2026E EPS of $7.98. Next-twelve-month figures are based on our estimates rather than consensus; peer data as at March 2026.
AAppendix A — Key Assumptions & Verification
| Assumption | FY23A | FY24A | FY25A | FY26E | FY27E | Rationale |
|---|---|---|---|---|---|---|
| AWS revenue growth | 13.0% | 18.5% | 19.7% | 26.0% | 24.0% | Q4 2025 actual +24%; backlog $244B; capacity unlock |
| AWS op. margin | 26.9% | 37.4% | 35.4% | 35.5% | 36.5% | FY25A verified; slight compression then expansion |
| Advertising growth | 27.0% | 19.8% | 22.1% | 21.0% | 22.0% | Prime Video ramp; retail media TAM expansion |
| Advertising contribution margin (est.) | 80% | 82% | 83% | 83% | 84% | Not disclosed; estimated from DSP and media peers, pre-allocation |
| NA revenue growth | 12% | 10% | 10% | 10% | 10% | Consistent with the three-year trend |
| NA op. margin | 5.2% | 6.5% | 6.9% | 7.5% | 8.5% | Q4 2025 actual 9.0%; robotics plus advertising mix |
| International op. margin | −2.1% | 2.7% | 2.9% | 3.5% | 4.0% | FX tailwind; Prime attach in India and Brazil |
| D&A as % of revenue | 6.1% | 6.8% | 9.2% | 9.5% | 9.7% | Server life shortening (January 2025) plus $200B capex |
| Diluted shares (B) | 10.55 | 10.62 | 10.73 | 10.78 | 10.82 | Minimal dilution; limited buyback programme |
Data verification
All FY2023A–FY2025A actuals are verified against Amazon's Q4 2025 earnings press release of 5 February 2026 and the FY2025 Form 10-K. Key restatements against our earlier working model: FY2025A net cash restated to $21.2B ($86.8B cash less $65.6B long-term debt); FY2026E net debt restated to $24B, since capital expenditure of $200B exceeds operating cash flow of roughly $155B; and depreciation and amortisation restated to 9.2% of revenue following the January 2025 shortening of server useful lives. All consensus figures are sourced from FactSet and Bloomberg as at March 2026.
Methodological notes
- AWS multiple. We apply 11.5× to FY2026E revenue, a discount of roughly 12% to Microsoft at approximately 13×, reflecting Amazon's greater exposure to commodity infrastructure-as-a-service against Microsoft's higher-margin software mix. The premium to Google Cloud at approximately 9× is justified by AWS's superior margin, above 35% against 17–20%, and longer-tenured enterprise relationships.
- Advertising multiple. We apply 6.0× to FY2026E revenue, in line with Alphabet's current trading despite a superior margin profile. The discount to Meta at 7.5× reflects the non-disclosure of segment financials and lower brand equity as a media business. Each 1.0× of re-rating is worth approximately $83B of enterprise value, or $7.70 per share.
- North America. We apply 20× FY2026E operating income rather than EBITDA to avoid distortion from the capital expenditure cycle, sitting between Walmart at 18× and Costco at 25×, reflecting superior growth but lower asset quality. The multiple is also set to absorb part of the advertising overlap described in Section 5.
- Net debt bridge. FY2025A: $86.8B of cash less $65.6B of long-term debt gives $21.2B of net cash, verified against the Form 10-K. FY2026E is estimated at $24B of net debt, starting from $21.2B of net cash with operating cash flow of $155B against capital expenditure of $200B, a net outflow of $45B. The $76B swing from the $52B of net cash assumed in our original working model to $24B of net debt is material: it reduces equity value by $76B, or approximately $7.08 per share.
BAppendix B — Revision Note
This note was first published in March 2026. It was revised in July 2026 for internal consistency and presentation only. No estimate, assumption, rating, price target or source datum has been changed, and no information published after the original March 2026 date has been incorporated. The analysis stands or falls on what was knowable at first publication.