Initiate at Sell with a $231 target as $818bn of capex drains cash through FY28E
We initiate on Amazon at Sell with a $231.47 12-month target, for a −10.2% total return. Amazon runs the largest online store and the leading cloud platform, Amazon Web Services (AWS). Investors price earnings per share (EPS) while overlooking capital expenditure (capex) above depreciation. Quarterly AWS revenue growth, measured year on year before currency moves, drives the stock.
- Three-year cash capex of $818bn ($632bn in AWS) keeps free cash flow negative through FY28E. Excess capex cuts discounted cash flow (DCF) value by $62.21 a share against peer price-to-earnings (P/E).
- Server depreciation trims AWS operating margin by 5.2pp, from 38.1% to 33.0% by FY29E ($11.72 a share per 2pp). Memory limits slow FY28E AWS growth to 38%, cutting value by $1.92 a share against the 40% guide.
- Justifying $257.89 needs 7.8% terminal growth or a 9.6% discount rate. Artificial intelligence (AI) stakes in Anthropic and OpenAI add $27.29 a share after tax, leaving operations dear.
What the market is missing. FY27E diluted EPS of $10.25 sits 4.1% below $10.68 consensus on AWS depreciation and new-debt interest. Free cash flow (FCF) less stock-based compensation (SBC) hits −$88.1bn in FY27E.
Catalysts. AWS re:Invent (Dec 2026, est.) tests Trainium3; 4Q26 results (Feb 2027, est.) set the FY27 capex guide; Anthropic's lock-up ends in May 2027 (est.).
Valuation and target. Free cash flow to the firm (FCFF) DCF gives $201.58 base ($71.46 bear 20%, $201.58 base 50%, $309.00 bull 30%). Rolling $207.78 probability-weighted (PW) value at 11.4% cost of equity sets our target.
Risks. Conviction is low. Our $231.47 target sits only $0.63 below the $232.10 Hold line. Any one of our five bull catalysts, a 0.5pp lower weighted average cost of capital (WACC), or a 0.3% fall in the share price moves the call to Hold. A Buy needs the price below $210.43.
| Dec YE, $bn | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Revenue | 638.0 | 716.9 | 827.1 | 963.9 | 1,125.2 |
| Growth y/y | 11.0% | 12.4% | 15.4% | 16.5% | 16.7% |
| EBITDA | 121.4 | 145.7 | 194.5 | 257.7 | 326.5 |
| Margin | 19.0% | 20.3% | 23.5% | 26.7% | 29.0% |
| Net income | 59.2 | 77.7 | 137.7 | 113.1 | 140.4 |
| EPS ($) | 5.53 | 7.17 | 12.59 | 10.25 | 12.72 |
| Growth y/y | 90.7% | 29.7% | 75.6% | −18.6% | 24.2% |
| Free cash flow | 32.9 | 11.2 | −46.5 | −63.0 | −24.7 |
| At $257.89, 17 Nov | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| P/E (price/earnings) | 46.6x | 36.0x | 20.5x | 25.2x | 20.3x |
| EV/EBITDA | 21.8x | 18.1x | 13.6x | 10.3x | 8.1x |
| P/B (price/book) | 9.7x | 6.8x | 4.4x | 3.6x | 3.0x |
| FCF yield | 1.2% | 0.4% | −1.6% | −2.2% | −0.9% |
| Dividend yield | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
| ROE (return on equity) | 24.3% | 22.3% | 26.1% | 15.8% | 16.2% |
| Net debt / EBITDA | −0.4x | −0.4x | 0.3x | 0.5x | 0.4x |
| P/E at our target | 41.9x | 32.3x | 18.4x | 22.6x | 18.2x |
Source: Company filings, Yahoo Finance, Penmark Research model. EBITDA = earnings before interest, tax, depreciation and amortisation; EV = enterprise value ex AI stakes ($2,641.7bn). FY26E EPS includes $68.3bn 9M26 marks ($7.59 ex marks).
Base-case DCF is $202 a share, and the downside is twice the upside
Source: Yahoo Finance; Penmark Research model. Scenarios rolled forward 12m at 11.4% cost of equity.
Present value $309.00 (+19.8%). AWS grows 47%/45% in FY27E/28E on Trainium3; AWS EBITDA margin 58.5%; Anthropic 1.20x IPO; WACC 10.95%.
Present value $201.58 (−21.8%). AWS grows 42%/38% in FY27E/28E; AWS EBITDA margin 57.0%; Anthropic 0.90x IPO (stakes worth $27.29 a share); WACC 10.95%.
Present value $71.46 (−72.3%). AI overbuild slows AWS to 32%/24%; AWS EBITDA margin 54.0%; North America 10.5%; Anthropic 0.55x IPO; WACC 10.95%.
| Income stmt, $bn | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Net sales | 638.0 | 716.9 | 827.1 | 963.9 | 1,125.2 |
| Total D&A | 52.8 | 65.8 | 84.0 | 112.8 | 144.8 |
| EBITDA | 121.4 | 145.7 | 194.5 | 257.7 | 326.5 |
| Operating EBIT | 68.6 | 80.0 | 110.5 | 144.9 | 181.7 |
| Net interest & marks | 0.0 | 17.3 | 67.1 | −3.6 | −6.3 |
| Income tax (book) | 9.3 | 19.1 | 40.0 | 28.3 | 35.1 |
| Net income, ex marks | 61.2 | 65.4 | 83.0 | 113.1 | 140.4 |
| Net income, reported | 59.2 | 77.7 | 137.7 | 113.1 | 140.4 |
| Cash flow, $bn | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| EBITDA | 121.4 | 145.7 | 194.5 | 257.7 | 326.5 |
| Stock-based comp | 22.0 | 19.5 | 21.5 | 25.1 | 29.3 |
| WC, tax & content | −27.5 | −25.7 | −42.5 | −56.2 | −72.4 |
| Operating cash flow | 115.9 | 139.5 | 173.5 | 226.6 | 283.4 |
| Cash capex | −83.0 | −128.3 | −220.0 | −289.6 | −308.1 |
| Free cash flow (FCF) | 32.9 | 11.2 | −46.5 | −63.0 | −24.7 |
| FCF less SBC | 10.9 | −8.3 | −68.0 | −88.1 | −54.0 |
| Net debt issued | −5.8 | 3.8 | 111.6 | 38.6 | 24.7 |
| Balance sheet, $bn | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Cash & securities | 101.2 | 123.0 | 124.4 | 100.0 | 100.0 |
| Property & equip., net | 252.7 | 357.0 | 518.6 | 725.3 | 923.5 |
| Operating lease ROU | 76.1 | 86.1 | 101.7 | 121.0 | 143.9 |
| Strategic investments | 32.7 | 62.7 | 290.3 | 290.3 | 290.3 |
| Total assets (net WC) | 535.2 | 712.0 | 1,118.2 | 1,319.8 | 1,540.9 |
| Carrying debt | 57.6 | 68.4 | 180.0 | 218.6 | 243.3 |
| Net cash / (debt) | 43.6 | 54.6 | −55.6 | −118.6 | −143.3 |
| Shareholders' equity | 286.0 | 411.1 | 645.8 | 783.9 | 953.5 |
| Per share and KPIs | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Diluted shares (m) | 10,721 | 10,827 | 10,935 | 11,034 | 11,034 |
| EPS, ex marks ($) | 5.71 | 6.04 | 7.59 | 10.25 | 12.72 |
| EPS, reported ($) | 5.53 | 7.17 | 12.59 | 10.25 | 12.72 |
| Dividend per share ($) | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| FCF per share ($) | 3.07 | 1.03 | −4.25 | −5.71 | −2.24 |
| Book value / share ($) | 26.67 | 37.97 | 59.05 | 71.04 | 86.42 |
| AWS sales growth y/y | 18.5% | 19.7% | 37.0% | 42.0% | 38.0% |
| AWS EBIT margin | 37.0% | 35.4% | 38.1% | 35.4% | 33.8% |
Source: Company filings, Penmark Research model. Condensed balance sheet nets operating current assets against current liabilities (GAAP total assets $624.9bn FY24A, $818.0bn FY25A); carrying debt = long-term debt plus current portion ($57.6bn FY24A, $68.4bn FY25A). D&A = depreciation & amortisation; KPI = key performance indicator; PP&E = property & equipment; ROU = right-of-use; WC = working capital.
EBITDA margin reaches 29% by FY28E, yet FY27E EPS sits 4% below consensus
| Growth, margins (%) | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Growth y/y | |||||
| Sales | 11.0% | 12.4% | 15.4% | 16.5% | 16.7% |
| EBITDA | 42.0% | 20.1% | 33.5% | 32.5% | 26.7% |
| EBIT | 86.1% | 16.6% | 38.2% | 31.1% | 25.4% |
| EPS, reported | 90.7% | 29.7% | 75.6% | −18.6% | 24.2% |
| EPS, ex marks | 101.8% | 5.8% | 25.7% | 34.9% | 24.2% |
| Margins and capital intensity | |||||
| AWS EBITDA margin | 49.4% | 52.1% | 57.5% | 57.0% | 56.0% |
| EBITDA margin | 19.0% | 20.3% | 23.5% | 26.7% | 29.0% |
| EBIT margin | 10.8% | 11.2% | 13.4% | 15.0% | 16.1% |
| FCF margin | 5.2% | 1.6% | −5.6% | −6.5% | −2.2% |
| EBITDA / interest | 50.5x | 64.1x | 33.0x | 31.1x | 32.5x |
| Cash capex / sales | 13.0% | 17.9% | 26.6% | 30.0% | 27.4% |
| Ratios and valuation | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| Returns and leverage | |||||
| ROE (return on eq.) | 24.3% | 22.3% | 26.1% | 15.8% | 16.2% |
| ROA (return on ast.) | 10.3% | 10.8% | 15.0% | 9.3% | 9.8% |
| Operating ROIC | n.d. | 21.2% | 20.1% | 17.8% | 16.5% |
| Net debt / equity | −15.2% | −13.3% | 8.6% | 15.1% | 15.0% |
| Net debt / EBITDA | −0.4x | −0.4x | 0.3x | 0.5x | 0.4x |
| Valuation at $257.89 | |||||
| P/E | 46.6x | 36.0x | 20.5x | 25.2x | 20.3x |
| EV/EBITDA | 21.8x | 18.1x | 13.6x | 10.3x | 8.1x |
| P/B | 9.7x | 6.8x | 4.4x | 3.6x | 3.0x |
| FCF yield | 1.2% | 0.4% | −1.6% | −2.2% | −0.9% |
| Dividend yield | 0.0% | 0.0% | 0.0% | 0.0% | 0.0% |
| P/E at our target | 41.9x | 32.3x | 18.4x | 22.6x | 18.2x |
Source: Company filings, Penmark Research model. Operating return on invested capital (ROIC) = after-tax EBIT / average (PP&E + operating lease ROU − net operating WC liability); FY24A is n.d. on the average-capital basis. ROA uses average GAAP assets in FY24A–FY25A and condensed model assets in FY26E–FY28E.
| $bn / $/share | 4Q26E | FY26E | FY27E | FY28E |
|---|---|---|---|---|
| Net sales ($bn) | ||||
| Penmark estimate | 243.1 | 827.1 | 963.9 | 1,125.2 |
| Consensus / guide mid | 243.5 | 829.0 | 959.4 | n.a. |
| Prior-year period | 213.4 | 716.9 | 827.1 | 963.9 |
| Diff. vs consensus | −0.2% | −0.2% | +0.5% | n.a. |
| Growth y/y (Penmark) | +13.9% | +15.4% | +16.5% | +16.7% |
| Diluted EPS ($) | ||||
| Penmark estimate | 2.17 | 12.59 | 10.25 | 12.72 |
| Consensus | n.a. | 12.80 | 10.68 | n.a. |
| Prior-year period | 1.95 | 7.17 | 12.59 | 10.25 |
| Diff. vs consensus | n.a. | −1.6% | −4.1% | n.a. |
| Growth y/y (Penmark) | +11.5% | +75.6% | −18.6% | +24.2% |
Source: 24/7 Wall St. / Yahoo Finance consensus (61 analysts); 4Q26E sales row uses company $240bn–$247bn guide midpoint; Penmark Research. Standalone 4Q26E EPS is not in our consensus source (FY26E less 9M26A = $2.17).
Source: Company filings (A); 24/7 Wall St. consensus; Penmark Research estimates (E)
At 25x FY27E EPS, AMZN trades at a 9% premium to peers despite negative free cash flow
| AMZN FY27E / peer LTM* | AMZN | Peers | Diff. |
|---|---|---|---|
| Valuation | |||
| Forward P/E | 25.2x | 23.1x | +9.0% |
| P/E vs hyperscalers | 25.2x | 19.6x | +28.4% |
| EV/EBITDA | 10.3x | 16.3x | −37.2% |
| FCF yield | −2.2% | +1.8% | −4.0pp |
| Dividend yield | 0.0% | 0.7% | −0.7pp |
| Growth and returns | |||
| Sales growth | 16.5% | 19.0% | −2.5pp |
| Operating EPS growth | 34.9% | 22.8% | +12.1pp |
| EBITDA margin | 26.7% | 40.8% | −14.1pp |
| ROE | 15.8% | 33.2% | −17.4pp |
| Net debt / EBITDA | 0.5x | 0.3x | +0.2x |
| Cash capex / sales | 30.0% | 28.7% | +1.3pp |
Source: Company filings (MSFT, GOOGL, ORCL, META, WMT, COST), S&P Global / Yahoo Finance, Penmark Research model. *AMZN = FY27E at $257.89; peers = forward P/E and latest annual or last twelve months (LTM) actuals.
- Dec 2026AWS re:Invent (est.): Trainium3 benchmarks & custom silicon cost disclosures.
- Feb 20274Q26 print (est.): FY27 capex guide (we model $289.6bn vs $220bn FY26).
- Apr 20271Q27 print (est.): AWS EBIT margin step-down (35.4% FY27E) & Leo capex.
- May 2027Anthropic 180-day IPO lock-up expiry (est.) tests our 0.90x offer mark.
Source: Company calendar, Anthropic S-1/A, Penmark Research (dates est.).
- AWS D&A & AI mix. Margin −4pp below 57% base. Prob. 30%; −$23.44/sh.
- Higher discount rate. 10Y US Treasury (UST) +50bp lifts WACC. Prob. 25%; −$15.88/sh.
- AI-lab concentration. FY28E AWS slows to 26%. Prob. 25%; −$11.50/sh.
- Lower WACC / Trainium upside. 10Y UST −50bp or +3pp AWS margin moves call. Prob. 30%; +$18.32/sh.
Source: Signals quarterly trailing twelve months (TTM) valuation metrics, 4Q22–3Q26. Standard deviation (SD) measures typical historical spread around the 17.5x mean; EV/EBITDA looks low because it ignores capex.
Source: S&P Global / Finbox peer set; Yahoo Finance for global comparables (*BABA, MELI, SHOP outside six-peer model set), Nov 2026. R-squared (R²) measures linear fit across the mixed retail/cloud set.
Cloud sales momentum lifts the stock, while $818bn of three-year capex depresses cash value
Amazon's share price moves with AWS top-line acceleration, yet cash conversion runs the other way. Four exhibits trace our Sell call from a 0.77 AWS-to-price correlation through three years of negative free cash flow. AWS operating margin also falls 516 basis points (bp), anchoring our $201.58 base DCF.
Source: Company filings; Signals KPIs; Yahoo Finance, 4Q23–3Q26.
Source: Company filings (A); Penmark Research estimates (E).
Source: Company filings (A); Penmark Research estimates (E).
Source: Penmark Research model.
Why the four charts point to Sell
Investors pay up whenever AWS growth accelerates, yet FY26E–FY28E cash capex totals $817.7bn ($632.2bn for AWS). Cash capex reaches $289.6bn in FY27E and $308.1bn in FY28E, outrunning operating cash flow until FY29E. Because peer P/E multiples price earnings after depreciation, not cash capex, they miss $62.21 a share of near-term cash drain.
Four debates explain $34 of the $56 gap between the share price and our base DCF
We sit below market expectations on all four debates: discount rate, AWS margin durability, AI stake valuation and FY28E cloud pace. The remaining $21.84 of the $56.31 price-to-DCF gap reflects other model factors. Our 3.5% terminal growth rate sits below the model's terminal-growth cap of about 4%.
| Debate | Market view | Penmark view | Evidence | What changes our view | Resolves |
|---|---|---|---|---|---|
| Capex return hurdle | WACC ~10.45% (price implies 9.61% WACC) ($219.90 a share) | 10.95% WACC; $818bn 3-yr capex drains cash ($201.58 a share) | TTM FCF −$42.4bn in 3Q26; operating ROIC eases from 21.2% in FY25A to 15.9% in FY29E (above 10.95% WACC); 10Y UST at 5.10%. | 10Y UST falls 50–100bp or FY27 capex guides below $250bn. | Feb 2027* |
| Cloud margin durability | AWS EBITDA margin ~59% ($213.30 a share) | 57.0% FY27E, 54.5% FY30E as D&A hits 22.2% of AWS sales | AWS 3Q26 EBIT margin slipped 70bp q/q to 38.7%; server lives cut to five years in 2025 (+$1.4bn FY25 D&A). | Trainium3 external adoption holds AWS EBIT margin above 37% in 1H27. | Apr 2027* |
| AI stake valuation | Anthropic at 1.00x–1.20x IPO ($204.09 a share at 1.00x) | Anthropic 0.90x IPO after 21% tax; OpenAI at $50bn cost ($27.29 a share) | $238.7bn 3Q26 carrying value plus 1.448x IPO mark faces 180-day lock-up supply. | Anthropic holds above $100 a share past its 180-day IPO lock-up expiry. | May 2027* |
| Cloud pace | AWS grows ≥40% in FY27E–FY28E (mgmt guide; $203.50 a share) | 42.0% FY27E, slowing to 38.0% FY28E ($201.58 a share) | $578bn backlog covers 2.31x FY27E sales, but high-bandwidth memory (HBM) & AI-lab concentration cap FY28E. | Quarterly AWS backlog additions stay above $80bn through 1H27. | Feb 2027* |
Source: Company filings, 3Q26 call, 24/7 Wall St. consensus, Penmark Research model. *2027 resolution dates est.
Source: Penmark Research model sensitivities.
Source: Company calendar; Anthropic filings; Penmark Research.
How the debates map to our thesis
Each debate anchors one thesis chapter later in this report. Capex return hurdle drives our cash-conversion chapter ($18.32 a share on WACC plus $62.21 a share against P/E). Cloud margin durability and Cloud pace anchor our AWS margin chapter ($11.72 and $1.92 a share). AI stake valuation anchors our reverse-DCF chapter ($2.51 a share).
Amazon brings in $717bn a year, yet its capex build-out will consume all operating cash through FY28E
Amazon runs the world's largest online retail marketplace alongside the leading public cloud computing platform. Thirty-two years after founding an asset-light bookseller, management is spending $818bn over three years on AI servers and warehouses, turning a cash engine into a net borrower.
The scene
On 29 October 2026, Amazon reported $28.1bn of quarterly operating income at a record 13.9% margin, and the stock dropped 16% intraday (closing down 4.2%). Chief Executive Officer Andy Jassy had just told listeners that FY27 capital expenditure would run meaningfully above FY26's $220bn pace while trailing twelve-month free cash flow sat at minus $42.4bn. That session captured Amazon's paradox: record accounting profit alongside heavy cash drain.
What it actually does
Shoppers see brown boxes arriving the same day; engineers see rented computing power on demand. Behind both sits one logistics and data-centre utility serving households, independent merchants, advertisers and enterprises across North America, International stores and AWS. Independent sellers supply 61% of worldwide paid units sold, turning Amazon's warehouses into a toll road rather than a traditional inventory retailer.
How the money comes in
Five streams brought in $716.9bn of FY25A sales, led by $269.3bn of online stores and $172.2bn of third-party seller fees. AWS added $128.7bn of cloud revenue and advertising brought in $68.6bn. Subscriptions ($49.6bn), physical stores ($22.6bn) and other sales ($5.9bn) contributed the remaining $78.1bn. Of each FY25A sales dollar, 11.2 cents survived as operating profit ($80.0bn), yet only 1.6 cents became free cash flow ($11.2bn) after $128.3bn of capex.
Where the story turns
Amazon's first pivot came in March 2006 with AWS Simple Storage Service, seeding a cloud business that earned a 35.4% operating margin in FY25A. Its second pivot arrived in 2025–2026: AWS backlog hit $578bn in 3Q26, driving $75.9bn of quarterly capex and lifting carrying debt to $180.0bn.
Why it works, and why it might not
Prime speed locks in shoppers, seller volume draws high-margin ads, and enterprise data anchors AWS. Yet if short-lived AI chips depreciate faster than prices hold, both Capex return hurdle and Cloud margin durability turn against shareholders.
Exhibit 24: One retail and platform flywheel, start to finish, built from FY25A–3Q26 filings and Penmark FY27E cash model.
AWS generates 18% of FY25A sales and 57% of group operating profit
Source: Company filings, AWS launch archive, Penmark Research.
1 · Customers
2 · Offerings
3 · Revenue ($717bn)
4 · Infrastructure
5 · Cost ($637bn)
6 · What is left
Source: FY25 Form 10-K, 3Q26 10-Q & KPIs, Penmark Research model. Fulfillment by Amazon (FBA); consumer packaged goods (CPG); cost of goods sold (COGS); general & administrative (G&A); operating cash flow (OCF).
Two economic engines under one balance sheet
Retail supplies scale and working-capital float; AWS and advertising supply high-margin operating profit. In Stores, customers pay at checkout while Amazon pays suppliers weeks later, creating $94.6bn of net negative operating working capital in FY25A.
Pricing and payment mechanics
Cloud bills monthly on consumption or multi-year reserved contracts; marketplace charges 8–15% referral fees plus fulfilment and cost-per-click ads. Neither segment requires customer receivables financing, yet AWS requires heavy upfront server and power capex 12–24 months before usage ramps.
Of each FY25A sales dollar, 50 cents is gross profit, 11 cents EBIT and under 2 cents FCF
Source: FY25 Form 10-K (subscriptions $49.6bn, physical stores $22.6bn, other $5.9bn).
Mix shift toward AWS and advertising lifted gross margin to 50.3% in FY25A, while fulfilment ($109.1bn) and technology and infrastructure ($108.5bn, up 23% year on year) led operating expenses. Fulfilment scales largely with unit volume, whereas server depreciation inside technology and infrastructure rises ahead of revenue as AI clusters come online.
Source: FY25 10-K; 3Q26 10-Q & KPIs; Penmark Research estimates (FY27E). Remaining performance obligations (RPO); shaded nodes mark confirmed stock drivers.
Source: FY25 Form 10-K geographic note.
Source: FY25 Form 10-K segment note.
Global cloud infrastructure reached a $671bn annualised pace in 3Q26, led by AWS at 26%
Source: Company filings; Synergy Research & eMarketer external estimates; Penmark Research. Infrastructure/platform-as-a-service (IaaS/PaaS); retail media network (RMN).
Source: Synergy Research Group external estimate, 3Q26 ($167.7bn quarterly spend).
Source: eMarketer external estimates, 2026E (Target Roundel 1.5%).
| Market pool | 2026E size | Market y/y | AMZN share | Top three share | Cycle & margin dynamic |
|---|---|---|---|---|---|
| Global cloud IaaS / PaaS | $671bn ann. | +55% (3Q26) | 26.0% (AWS) | 61.0% | Capacity-constrained on power & HBM memory; heavy AI capex |
| Global enterprise IT (total) | ~$4.5tn | +8–10% | ~3.9% | n.d. | ~85% still on-premises per management; multi-decade shift |
| US e-commerce GMV | ~$1.3tn | +7–9% | 39.5% | ~55% | ~80% of total retail still in physical stores; speed widens moat |
| US retail media advertising | ~$70bn | +17–20% | 79.4% | 89.9% | Closed-loop purchase attribution earns >50% incremental margin |
Source: Synergy Research (cloud), eMarketer (US e-commerce & retail media external estimates), management commentary on 3Q26 call, Penmark Research. Gross merchandise volume (GMV).
Demand drivers and cycle position
Enterprise cloud migration and generative AI training pushed global cloud spend up 55% year on year to $167.7bn in 3Q26. Management estimates 85% of global IT spend remains on-premises and 80% of retail stays in physical stores. Near term, cloud sits in a capital-heavy build phase where data-centre power and high-bandwidth memory (HBM) govern growth.
Who earns the margin, and key threats
Merchant AI chip suppliers capture 55–75% gross margins at the top of the stack, squeezing cloud cash conversion. Amazon's custom Graviton and Trainium chips aim to claw back silicon rent by late 2027. In retail, AI shopping agents could bypass Amazon's search page (a bear risk worth $8.91 a share), while Federal Trade Commission (FTC) litigation tests seller fees.
Amazon leads on four of seven moat pillars, yet AI silicon and multi-cloud narrow its cost edge
Source: Company filings; Penmark Research model; Yahoo Finance TTM statistics, Nov 2026; bubble size = TTM revenue ($797.5bn AMZN). *SHOP ($14.2bn) outside six-peer model set.
| Source | Strength | Evidence |
|---|---|---|
| Scale | ●●●● | $716.9bn FY25 sales; regional US hubs lower cost per unit |
| Network | ●●●● | 61% 3P unit share feeds 79.4% US retail-media ad share |
| Switching | ●●●● | $578bn AWS backlog (multi-year RPO) & data gravity |
| Cost | ●●●○ | Regionalised parcel network; Nvidia GPU dependency dilutes cloud |
| Brand | ●●●● | $49.6bn subscription base makes Prime default checkout |
| IP & silicon | ●●●○ | Graviton CPUs proven; Trainium3 external ramp late 2027 |
| Licences | ●●○○ | FCC orbital spectrum for Amazon Leo; active FTC litigation |
Source: Company filings, 3Q26 call, Penmark Research assessment. Graphics/central processing units (GPU/CPU); Federal Communications Commission (FCC).
| Segment & customer buying criterion | Weight | Amazon | MSFT / WMT | GOOGL / COST | ORCL / SHOP |
|---|---|---|---|---|---|
| Cloud: core breadth, uptime & database ecosystem | 30% | 1st (AWS) | 2nd (Azure) | 3rd (GCP) | 4th (OCI) |
| Cloud: frontier AI models, custom silicon & SaaS | 25% | 2nd (Bedrock) | 1st (OpenAI/M365) | 2nd (Gemini/TPU) | 4th (GPU cluster) |
| Commerce: SKU selection, Prime speed & parcel cost | 25% | 1st (39.5% US) | 2nd (WMT stores) | 3rd (COST clubs) | 2nd (SHOP D2C) |
| Retail media: closed-loop SKU ad conversion | 20% | 1st (79.4% US) | 2nd (WMT 9.0%) | 4th | 3rd (Shop Ads) |
Source: Synergy Research, eMarketer external estimates, company filings, Penmark Research rankings. Software-as-a-service (SaaS); tensor processing unit (TPU); stock-keeping unit (SKU); direct-to-consumer (D2C).
Why the moat endures, and where it leaks cash
Amazon's fulfilment density and AWS's $578bn remaining performance obligation (RPO) drive high customer retention. Worldwide paid units grew 12% year on year in 3Q26 while third-party sellers held a 61% unit share, funding $19.6bn of quarterly ad revenue. Yet generative AI lets frontier labs pit AWS, Azure, Google Cloud and Oracle against one another on GPU cluster pricing.
Competitors and substitutes by segment
Amazon faces distinct mega-cap rivals in each profit pool, so no single peer multiple fits the group. In cloud, Microsoft Azure (19% share) leverages Office seats and OpenAI, Google Cloud (16%) deploys custom TPUs, and Oracle builds multi-gigawatt clusters. In US commerce, Walmart (11.5% online share) uses over 4,600 stores for grocery delivery, while Temu, Shein and Shopify target apparel.
Insiders own 9% of Amazon, and 83% of five-year cash went to capex
| Executive | Tenure | How they are paid (FY25 proxy) | What they own (at $257.89) |
|---|---|---|---|
| Andrew R. JassyPresident & CEO | 5.4 yrs in roleCEO Jul 2021; joined 1997 (29 yrs). Prior: HBS. | $2.07m SCT, FY25 $365k salary + $1.70m security/401k; $0 bonus; time-vested RSUs (share-price linked; no FCF/ROIC gate). |
$867.3m <0.1%2.31m held + 1.05m unvested RSUs; 2,376x salary. Routine RSU tax/10b5-1 sales. |
| Brian T. OlsavskySVP & CFO | 11.5 yrs in roleCFO Jun 2015; joined Apr 2002 (24 yrs). Prior: Fisher. | $0.37m SCT, FY25 $365k salary + $7k 401k; $0 bonus; $0 FY25 grant; time-vested RSUs (share-price linked; no FCF/ROIC gate). |
$81.7m <0.1%0.11m held + 0.21m unvested RSUs; 224x salary. Aug 2026 10b5-1 sale (6.2k sh). |
| Matthew S. GarmanCEO, AWS | 2.5 yrs in roleAWS CEO Jun 2024; joined 2006 (20 yrs). AWS veteran. | $0.62m SCT, FY25 $365k salary + $252k security/401k; $0 bonus; time-vested RSUs (share-price linked; no FCF/ROIC gate). |
$74.6m <0.1%0.02m held + 0.27m unvested RSUs; 204x salary. Aug 2026 10b5-1 sales (14.5k sh). |
| Douglas J. HerringtonCEO, Worldwide Stores | 4.3 yrs in roleStores CEO Jul 2022; joined 2005 (21 yrs). Prior: Webvan. | $0.43m SCT, FY25 $365k salary + $60k security/401k; $0 bonus; time-vested RSUs (share-price linked; no FCF/ROIC gate). |
$211.2m <0.1%0.48m held (479,292 sh) + 0.34m RSUs; 579x salary. 10b5-1 trades 2 and 16 Nov (Form 4s filed 4 and 17 Nov). |
| David A. ZapolskyChief Global Affairs/Legal | 12.5 yrs GC/SVPRole expanded 2023; joined 1999 (27 yrs). Prior: Dorsey. | $0.37m SCT, FY25 $365k salary + $7k 401k; $0 bonus; time-vested RSUs (share-price linked; no FCF/ROIC gate). |
$64.1m <0.1%0.04m held + 0.21m unvested RSUs; 176x salary. Aug 2026 10b5-1 sale (9.3k sh). |
| Jeffrey P. BezosFounder & Exec. Chair | 5.4 yrs ChairFounded Jul 1994 (32 yrs); CEO 1994–Jul 2021. | $1.68m SCT, FY25 $81.8k salary + $1.60m security; never took equity grants or cash bonuses. |
$245.11bn 8.8%950.43m beneficial (878.55m direct after Nov gifts); ~3.0mx salary. |
Source: 2026 DEF 14A (FY25 Summary Compensation Table [SCT], beneficial shares at 24 Feb 2026, unvested RSUs at 31 Dec 2025); direct shares updated via Form 4 to 17 Nov 2026 (Herrington 10b5-1 trades 2 and 16 Nov [Form 4s filed 4 and 17 Nov]: 473,681 sh after 2 Nov sale; 479,292 sh after 15 Nov vest and 16 Nov sales), valued at $257.89. Named executive officers (NEO).
Source: 2026 DEF 14A unvested RSUs at 31 Dec 2025, valued at $257.89.
Pay structure and governance
Base salaries cap at $365,000 with zero cash bonuses, while time-vested restricted stock units (RSUs) are share-price linked with no ROIC or FCF-per-share gate. Single-class shares give one vote per share (9 of 11 directors independent), though $2.7bn of Blue Origin launch contracts for Amazon Leo pose related-party optics with Executive Chair Bezos.
Source: FY21–FY25 Form 10-K cash flow statements (OCF $433.4bn + net debt/other $36.8bn = $470.2bn; gross PP&E $392.2bn vs $388.7bn net of FY25 proceeds in Ex 43). Named executive officers (NEO).
EBIT beat the top of guidance in five of seven quarters, while operating ROIC eases to 16%
Source: 4Q24–3Q26 earnings releases (3Q25 missed midpoint on FTC legal charge; 4Q25 cleared midpoint).
Source: Company filings (A); Penmark Research model operating ROIC (E) (after-tax EBIT at 20% / average operating capital).
| Capital allocation & dilution ($m) | FY21A | FY22A | FY23A | FY24A | FY25A | 5Y sum / chg |
|---|---|---|---|---|---|---|
| Operating cash flow | 46,327 | 46,752 | 84,946 | 115,877 | 139,514 | 433,416 |
| Cash capex (PP&E, model basis)* | (61,053) | (63,645) | (52,729) | (82,999) | (128,320) | (388,746) |
| Strategic stakes & M&A, net | (1,985) | (8,316) | (5,839) | (7,082) | (10,192) | (33,414) |
| Share buybacks (dividends: $0) | 0 | (6,000) | 0 | 0 | 0 | (6,000) |
| Stock-based compensation (SBC) | 12,757 | 19,621 | 24,023 | 22,011 | 19,467 | 97,879 |
| Diluted shares (m, split-adj.) | 10,296 | 10,189 | 10,492 | 10,721 | 10,827 | +5.2% (+531m) |
| Operating ROIC (%) | n.d. | n.d. | n.d. | n.d. | 21.2% | 15.9% in FY29E |
Source: FY21–FY25 Form 10-K; Penmark Research model. *FY25A net of $3,499m PP&E proceeds ($131,819m gross; 5Y gross $392,245m in Ex 40). Operating ROIC uses average operating capital (FY21A–FY24A opening operating WC not in model).
Return on each use of cash
Amazon's capital record splits between operational discipline, lucrative AI venture bets and cyclical over-builds. Regionalising US fulfilment lifted North America operating margin from minus 0.9% in FY22 to 6.9% in FY25A, while early Anthropic notes generated over $200bn of paper and IPO gains. Conversely, Amazon paid zero dividends, paused buybacks after $6.0bn in FY22, and let SBC expand diluted shares by 5.2% over FY21–FY25.
What it means for value per share
Frequent quarterly EBIT beats explain why Street sentiment stands at 59 Buys and zero Sells, yet cash flow diverges from operating profit. Model operating ROIC stays above our 10.95% WACC, easing from 21.2% in FY25A to 15.9% in FY29E. Even so, $817.7bn of FY26E–FY28E cash capex keeps free cash flow negative through FY28E, deferring buybacks until FY31E.
Capex return hurdles and AWS margin durability govern AMZN far more than headline EPS
Amazon does not trade on near-term earnings beats: it trades on AWS revenue velocity on the way up and capital-spending shocks on the way down. Across four empirical tests (eight earnings-day reactions, 1/3/5-year return decomposition, five-year weekly factor regressions, and 35 analyst question blocks across eight calls), three operating drivers and one balance-sheet stake pass at least two tests.
| Confirmed driver | What moves the stock | Tests confirmed | Market vs model value | Named debate | Next number to watch |
|---|---|---|---|---|---|
| Capex return hurdle | Annual cash capex guide & TTM FCF burn vs OCF | 3 of 4 (events, return split, call Q&A) | 29 Oct fell −4.2% (−$10.81); excess capex costs −$62.21/sh in DCF | Capex return hurdle | FY27 capex guide (est. early Feb 2027): $289.6bnE vs ~$220bn FY26 guide |
| Cloud margin durability | AWS operating & EBITDA margin as AI D&A scales | 2 of 4 (return split, call Q&A) | Market extrapolates 38.7% 3Q26 EBIT mgn; 2pp shift = $11.72/sh (−4.5%) | Cloud margin durability | 4Q26 AWS EBIT margin (est. early Feb 2027): 37.0%E vs 38.7% in 3Q26 |
| Cloud pace | AWS y/y sales growth & $578bn backlog additions | 3 of 4 (events, 0.77 price corr., call Q&A) | 2Q26 jumped +14.6pp (+$34); +2pp FY28E AWS growth = +$1.92/sh (+0.7%) | Cloud pace | 4Q26 AWS y/y growth (est. early Feb 2027): 42.0%E vs 39.3% in 3Q26 |
| AI stake valuation | Anthropic public mark & GAAP non-operating EPS | 2 of 4 (return split, balance-sheet mark) | Inflates FY26E EPS to $12.59; 1.0x vs 0.9x IPO = +$2.51/sh (+1.0%) | AI stake valuation | Anthropic post-listing price vs $83 offer (Nov 2026–est. May 2027) |
Source: Signals transcripts and filings; Yahoo Finance price history to 17 Nov 2026; Penmark Research valuation model.
Source: Yahoo Finance T+1 close-to-close vs S&P 500 (3Q26: 29 Oct −16% intraday / −4.2% close; 30 Oct −1.4% vs S&P −1.8% = +0.4pp).
Source: Yahoo Finance returns to 17 Nov 2026 (1Y SPY +11.2% vs AMZN +10.7%; S&P +13.4% to 13 Nov).
Event-day asymmetry: AWS rallies against capex sell-offs
Five of the last eight earnings prints produced negative next-day (T+1) excess returns despite five top-of-guide EBIT beats. When AWS accelerated to 37% in 2Q26, AMZN jumped 14.6pp ahead of the index. When management raised FY26 capex in 4Q25 or flagged higher FY27 spend on 29 October 2026 (down 4.2% at the close), capex fears overshadowed record margins.
Return decomposition: earnings grew as ex-mark P/E eased
Over five years, AMZN rose 45.3% with zero dividends while trailing P/E excluding AI marks eased from about 55x in late 2021 to about 34x in 3Q26. Headline GAAP trailing P/E looks lower near 21x only because about $70bn of pre-tax AI investment marks inflate earnings. Trailing operating profit roughly quadrupled over three years, lifting AMZN 77.6%, ahead of the S&P 500's 62.6%, though surging capex tempered multiple gains.
Idiosyncratic factors drive 38% of the stock's variance, and 71% of analyst questions probe AWS demand and capex
Source: Yahoo Finance weekly returns, Nov 2021–Nov 2026 (n=261, R²=62.4%).
Source: Signals transcripts, 4Q24–3Q26 (35 non-management analyst question blocks across eight earnings calls).
Capex return hurdle: FCF burn anchors the multiple
Market reaction: On capex step-ups, the stock underperformed the index by 7.5pp in 4Q25 and fell $10.81 a share (4.2%) on 29 October 2026 after management guided FY27 capex meaningfully above $220bn. Model against market: Excess capex over depreciation costs $62.21 a share (24.1% of price), $51.40 a share larger than the 29 October drop. Debate and watch: Links to Capex return hurdle; watch FY27 capex guide (est. early Feb 2027: Penmark $289.6bn against about $220bn in FY26).
Cloud margin durability: D&A lags cash spend
Market reaction: Analysts devoted 5.7% of questions to AWS margins and 17.1% to capex returns, while management noted margins will fluctuate as AI depreciation ramps. Model against market: Investors extrapolate 2Q26–3Q26 AWS operating margins (38.7%–39.4%), yet a 2.0pp margin drop cuts DCF value by $11.72 a share (4.5%). Debate and watch: Links to Cloud margin durability; watch 4Q26 AWS EBIT margin (est. early Feb 2027: Penmark 37.0%E against 38.7% in 3Q26).
Cloud pace: priced in at 38–42%, punitive on slowdown
Market reaction: AWS growth led call questions (54.3%, confident answers on $578bn backlog) and sparked a 14.6pp rally ($34 a share) in 2Q26. Model against market: Because AWS capex scales with sales, 2pp faster FY28E AWS growth (40% versus 38%) adds just $1.92 a share (0.7%). Debate and watch: Links to Cloud pace; watch 4Q26 AWS year-on-year sales growth (est. early Feb 2027: Penmark 42.0%E against 39.3% in 3Q26).
AI stake valuation: headline EPS optics against $27 a share
Market reaction: Non-operating Anthropic gains lifted 9M26 GAAP net income to $113.6bn and FY26E EPS to $12.59, setting up FY27E's optical drop to $10.25. Model against market: Marking Anthropic at 1.00x IPO ($83) rather than our 0.90x haircut adds $2.51 a share (1.0%), or $7.53 a share at 1.20x. Debate and watch: Links to AI stake valuation; watch Anthropic trading against the $83 offer through lock-up expiry (est. May 2027).
| Candidate tested | Empirical test evidence | Why dropped from primary drivers |
|---|---|---|
| Weekly 10Y UST moves | 5Y weekly regression beta −0.013, incremental R² 0.39% (0.14% univariate) | Fails weekly variance test; structural WACC level matters in DCF, not weekly rate noise |
| Worldwide paid-unit growth | Stuck in a tight 10%–12% band across 2024–3Q26 (12% in 3Q26); 0% of Q&A | Low variance and zero event-day dispersion; subsumed into Stores operating leverage |
| Advertising revenue beat | Slowed from +26% y/y in 2Q26 to +11% in 3Q26 ($19.6bn); 0 lead Q&A blocks | Evaluated by investors inside North America retail margin rather than as a standalone driver |
| FX translation impact | −60bp guided 4Q26 sales drag; <6% combined macro/tariff/FX Q&A share | Analysed on a constant-currency disclosure basis; minimal standalone DCF weight |
Source: Yahoo Finance factor regressions (Nov 2021–Nov 2026); Signals KPI series and transcripts.
Capex above depreciation absorbs $62 a share of equity value through FY29E
The claim: Amazon spends $908.5bn on cash capex in FY27E–FY29E, exceeding depreciation by $580.8bn. Linked to our Capex return hurdle driver, this cash drain opens a gap of $62.21 a share between forward P/E ($263.80) and base DCF ($201.58).
What the market believes
Fifty-nine of 61 covering analysts rate AMZN Buy ($326.29 mean target), applying a 24x–31x P/E multiple to FY27E consensus EPS of $10.68. Accounting earnings spread server costs over five years and shells over 25–40 years, whereas the cash goes out upfront at an 11.4% cost of equity.
Three independent sources of evidence
Reported filings, call guidance and debt issuance confirm the cash drain. First, 3Q26 cash capex surged 122% year on year to $75.9bn ($172.2bn in 9M26), pulling trailing free cash flow to minus $42.4bn. Second, 3Q26 call guidance pointed to about $220bn of FY26 capex and meaningfully more in FY27, mostly for AWS (78% in our model). Third, carrying debt rose from $68.4bn at end-2025 to $180.0bn in 3Q26 ($88.8bn to $204.2bn including finance leases and financing obligations).
Source: Company filings (A); Penmark Research model (E).
| US$m (except ratios) | FY24A | FY25A | FY26E | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|---|---|
| AWS cash capex | n.d. | n.d. | 165,000 | 225,344 | 241,869 | 243,251 | 242,809 |
| Stores & corporate cash capex | n.d. | n.d. | 55,000 | 64,218 | 66,276 | 67,541 | 68,157 |
| Total cash capex | 82,999 | 128,320 | 220,000 | 289,561 | 308,145 | 310,792 | 310,966 |
| Capex / net sales | 13.0% | 17.9% | 26.6% | 30.0% | 27.4% | 24.2% | 21.5% |
| P&E depreciation (in EBIT) | 32,067 | 41,860 | 58,378 | 82,928 | 109,919 | 134,863 | 156,865 |
| Excess capex over P&E D&A | 50,932 | 86,460 | 161,622 | 206,633 | 198,226 | 175,929 | 154,101 |
| Operating cash flow (OCF) | 115,877 | 139,514 | 173,528 | 226,573 | 283,426 | 340,682 | 398,340 |
| Free cash flow (OCF less capex) | 32,878 | 11,194 | −46,472 | −62,988 | −24,719 | 29,891 | 87,374 |
| FCF less stock-based comp. | 10,867 | −8,273 | −67,978 | −88,050 | −53,975 | −3,559 | 49,717 |
| Reinvestment rate (% of NOPAT) | n.d. | n.d. | n.m. | 165.1% | 130.3% | 98.5% | 75.0% |
Source: FY25 Form 10-K; Penmark Research model (FY24A gross PP&E; FY25A–FY30E net of proceeds). Net operating profit after tax (NOPAT).
Free cash flow less SBC stays negative until FY30E, requiring $175bn of net debt issuance
Source: Company filings (A); Penmark Research model (E).
Source: Penmark Research Valuation and Scenarios schedules.
What it is worth per share
Peer forward P/E (23.08x) on FY27E EPS plus $27.29 a share of AI stakes gives $263.80 a share, 30.9% above our $201.58 base DCF, a gap of $62.21 a share. That gap reflects FY27E–FY29E reinvestment absorbing 99%–165% of net operating profit after tax (NOPAT). Defending the $100.0bn cash floor requires $174.9bn of net debt issuance through FY28E and postpones buybacks until FY31E. Counting stock awards ($25.1bn in FY27E) as a real cost deepens FY27E free cash flow less SBC to minus $88.1bn.
Strongest counter-evidence against our Sell
Bulls cite two real strengths: AWS's $578bn 3Q26 backlog covers 2.31x FY27E cloud sales, and data-centre shells earn returns into the 2030s. Our DCF already credits a 15.0% terminal return on new invested capital (RONIC) and $381.6bn of FY40E cash flow. At a 10.95% WACC, however, cash inflows from FY30E lose 32%–76% of face value when discounted back. Even so, our target sits just below the Hold threshold, leaving low conviction as detailed in the valuation chapter.
Dated signposts to confirm or disprove the capex hurdle claim
Confirm Sell: Formal FY27 cash capex guide above $275bn (Penmark $289.6bn) without AWS growth accelerating past 44%. Disprove: Capex guided below $250bn while AWS holds >40%.
Confirm Sell: Trailing-12-month FCF sinks past −$55bn toward −$63.0bn and carrying debt exceeds $210bn. Disprove: Working-capital inflows or capex pacing keep TTM FCF above −$30bn.
Confirm Sell: Capex/sales stays above 27% in FY28E and FCF less SBC remains negative through FY29E (−$3.6bn). Disprove: Positive FCF less SBC achieved two years early in FY28E.
Rising AI server depreciation trims AWS operating margin by 5pp from FY26E to FY29E
The claim: AWS spends $632.2bn on capex in FY26E–FY28E, lifting depreciation to 22.2% of AWS sales by FY28E and cutting AWS operating margin from 38.1% to 33.0% by FY29E. Linked to Cloud margin durability and Cloud pace, memory supply also caps FY28E AWS growth at 38.0%.
Source: FY25 10-K; 3Q26 10-Q; Penmark Research Segments schedule.
Source: Penmark Research Sensitivities schedule ($/share at price date).
Market view against three sources of evidence
Investors extrapolate 2Q26–3Q26 AWS operating margins (about 39%) and management's north-of-40% FY28 growth target. First, cutting server lives to five years in January 2025 added $1.4bn of FY25 depreciation (a $0.10 drag on EPS), and 3Q26 AWS margin eased 70bp sequentially to 38.7%. Second, 3Q26 call remarks confirmed early AI workloads carry lower gross margins while external Trainium sales wait until late 2027. Third, memory constraints and AI-lab concentration cap our FY28E AWS growth at 38.0% despite $578bn of backlog.
What it is worth and dated signposts
Lifting AWS EBITDA margin by 2pp adds $11.72 a share, whereas 40% versus 38% FY28E growth adds only $1.92 a share. Even 50% FY27E AWS growth plus 4pp higher margin reaches $233.36, still 9.5% below $257.89, while 4pp lower margin costs $23.44 a share. Signposts: AWS re:Invent (est. Dec 2026: Trainium3 unit costs), 4Q26 AWS EBIT margin (37.0%E against 38.7% in 3Q26), and 2H27 memory supply.
| AWS segment (US$m / %) | FY24A | FY25A | FY26E | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|---|---|
| AWS net sales | 107,556 | 128,725 | 176,325 | 250,382 | 345,527 | 442,274 | 539,575 |
| AWS revenue growth (y/y) | 18.5% | 19.7% | 37.0% | 42.0% | 38.0% | 28.0% | 22.0% |
| AWS EBITDA (P&E D&A basis) | 53,154 | 67,056 | 101,318 | 142,718 | 193,495 | 243,251 | 294,068 |
| AWS EBITDA margin | 49.4% | 52.1% | 57.5% | 57.0% | 56.0% | 55.0% | 54.5% |
| AWS P&E depreciation | 13,320 | 21,450 | 34,069 | 54,207 | 76,632 | 97,373 | 115,580 |
| AWS P&E D&A / AWS sales | 12.4% | 16.7% | 19.3% | 21.6% | 22.2% | 22.0% | 21.4% |
| AWS operating income | 39,834 | 45,606 | 67,248 | 88,510 | 116,863 | 145,878 | 178,488 |
| AWS operating margin | 37.0% | 35.4% | 38.1% | 35.4% | 33.8% | 33.0% | 33.1% |
| AWS closing net PP&E | n.d. | 190,055 | 320,986 | 492,122 | 657,359 | 803,237 | 930,465 |
Source: FY25 Form 10-K; Penmark Research Segments schedule (FY25A closing AWS PP&E equals FY26E opening AWS PP&E).
Justifying $258 requires about 8% terminal growth, while AI marks flatter FY26E GAAP EPS
The claim: Even after crediting $301.2bn ($27.29 a share) after tax for Anthropic and OpenAI, justifying $257.89 requires 7.82% terminal growth or a 64.9% FY30E AWS EBITDA margin. Linked to AI stake valuation, 9M26 AI gains also inflate FY26E GAAP EPS to $12.59 before FY27E resets to $10.25.
Source: Penmark Research Sensitivities reverse-DCF solves.
Source: 3Q26 10-Q Note 2; Anthropic IPO pricing (18 Nov 2026); Penmark model (30 Sep 2026 $/sh).
Market view and three sources of evidence
A reverse DCF solves backward from $257.89 to reveal what investors implicitly assume. At a 10.95% WACC, $257.89 demands 7.82% perpetuity growth past FY40E, roughly double our terminal-growth cap of about 4% (and long-run US nominal GDP). Alternatively, it requires a 10.4pp perpetual AWS EBITDA margin jump to 64.9%. Meanwhile, 3Q26 Note 2 shows $238.7bn of Anthropic holdings and $50.0bn of OpenAI at cost. The $68.3bn of pre-tax investment gains lift FY26E GAAP EPS to $12.59 against $7.59 operating EPS, creating FY27E's optical GAAP drop to $10.25.
What it is worth and dated signposts
We include strategic stakes once in the equity bridge so AI marks are neither missed nor double-counted. Anthropic's 18 November 2026 IPO ($83 offer, 23.2% inferred economic interest; 14.7% of total capital stock in the S-1/A) lifts gross value to $345.7bn. Applying a 0.90x haircut and 21% tax over cost leaves $249.6bn, or $301.2bn ($27.29 a share) including OpenAI and other stakes. Marking Anthropic at 1.00x adds $2.51 a share, or $7.53 a share at 1.20x. Signposts: Anthropic trading against $83 through est. May 2027 lock-up expiry, and 1Q27–2Q27 ex-mark EPS.
| Strategic investment bridge (US$m / US$/sh) | 3Q26 book | Bear (0.55x) | Base (0.90x) | IPO offer (1.0x) | Bull (1.20x) |
|---|---|---|---|---|---|
| Anthropic carrying value (preferred $92.5bn + notes $146.2bn) | 238,700 | 238,700 | 238,700 | 238,700 | 238,700 |
| Anthropic gross value after 1.448x IPO uplift & scenario mark | 238,700 | 190,125 | 311,114 | 345,682 | 414,818 |
| Less: 21% corporate tax on gain above cost basis | n.a. | −36,146 | −61,554 | −68,813 | −83,332 |
| Anthropic stake after tax | n.a. | 153,979 | 249,560 | 276,869 | 331,486 |
| OpenAI Series C preferred (at cost) + other private stakes ($1,595m) | 51,595 | 51,595 | 51,595 | 51,595 | 51,595 |
| Total strategic investments after tax (US$m) | 290,295 | 205,574 | 301,155 | 328,464 | 383,081 |
| Strategic investments per diluted share (US$) | 26.31 | 18.63 | 27.29 | 29.77 | 34.72 |
| Full-model DCF value per share at scenario AI mark (base ops) | n.a. | 192.80 | 201.58 | 204.09 | 209.11 |
Source: 3Q26 Form 10-Q Note 2; Penmark Research Valuation and Sensitivities schedules (Exhibit 58 includes full scenario roll-up).
EBIT tripled to $80bn over FY21–FY25, yet SBC, useful-life shifts and AI marks flatter GAAP
Organic mix lifted Amazon's operating margin from 5.3% to 11.2% over FY21A–FY25A: AWS cloud, seller services (61% of global units) and advertising ($68.6bn in FY25A). Mergers and acquisitions (M&A) played almost no role. Cash conversion diverged once AI capex surged, leaving five-year free cash flow less SBC at minus $53.2bn.
| Track record (US$m / %) | FY21A | FY22A | FY23A | FY24A | FY25A | 9M26A |
|---|---|---|---|---|---|---|
| Net sales | 469,822 | 513,983 | 574,785 | 637,959 | 716,924 | 584,069 |
| Net sales growth (y/y) | 21.7% | 9.4% | 11.8% | 11.0% | 12.4% | 16.0% |
| AWS share of net sales | 13.2% | 15.6% | 15.8% | 16.9% | 18.0% | 21.5% |
| Operating income (GAAP EBIT) | 24,879 | 12,248 | 36,852 | 68,593 | 79,975 | 79,389 |
| Operating margin | 5.3% | 2.4% | 6.4% | 10.8% | 11.2% | 13.6% |
| Stock-based compensation (in EBIT) | 12,757 | 19,621 | 24,023 | 22,011 | 19,467 | 15,476 |
| Other income/(exp.) incl. AI/Rivian marks | 14,633 | −16,806 | 938 | −2,250 | 15,229 | 68,325 |
| GAAP net income | 33,364 | −2,722 | 30,425 | 59,248 | 77,670 | 113,587 |
| Operating cash flow (OCF) | 46,327 | 46,752 | 84,946 | 115,877 | 139,514 | 113,841 |
| Cash capex (FY21–24 gross; FY25–9M26 net) | −61,053 | −63,645 | −52,729 | −82,999 | −128,320 | −172,188 |
| Free cash flow (OCF less cash capex) | −14,726 | −16,893 | 32,217 | 32,878 | 11,194 | −58,347 |
| Free cash flow less SBC | −27,483 | −36,514 | 8,194 | 10,867 | −8,273 | −73,823 |
Source: FY21–FY25 Form 10-K; 3Q26 Form 10-Q; Penmark Research model.
Source: Company filings (A); Penmark Research model (E).
| Item | Reported impact & Penmark treatment |
|---|---|
| SBC add-back | $97.9bn expensed in FY21–FY25 EBIT but added back in OCF; kept as cash cost in FCFF + 223m awards in 11,034m shares. |
| Server lives | Extended 4→5→6 yrs in 2022–2024, then cut to five years in Jan 2025 (+$1.4bn actual FY25 10-K D&A, −$1.0bn net income, −$0.10/sh EPS); AI racks lift D&A to FY29E. |
| Equity marks | Rivian (~−$12.7bn within FY22's −$16.8bn other exp.) and Anthropic (+$68.3bn 9M26) swing GAAP net income; zeroed from 4Q26E, valued once in bridge. |
| Leo & leases | $24.2bn 3Q26 finance leases/obligations counted in debt; Amazon Leo satellite spend capitalises from 1Q27 (lifts NA EBIT mgn toward 9.3%). |
Source: FY25 10-K; 3Q26 10-Q; Penmark Research model.
Growth decomposition: unit volume, marketplace take-rate and cloud mix
Foreign exchange trimmed less than 1pp annually across FY23A–FY25A, and acquisitions added under 0.5pp. Instead, North America and International grew 9%–13% via 10%–12% paid-unit volume and advertising mix, while AWS compounded 19.9% annually from $62.2bn to $128.7bn.
Group sales compound 15% a year to $1.4tn by FY30E as AWS reaches 37% of revenue
Source: Penmark Research Segments schedule (FY28E driver nodes; blue-shaded nodes mark confirmed stock drivers).
| Segment build (US$m / %) | FY24A | FY25A | FY26E | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|---|---|
| AWS net sales | 107,556 | 128,725 | 176,325 | 250,382 | 345,527 | 442,274 | 539,575 |
| AWS growth / EBIT margin | 18.5%/37.0% | 19.7%/35.4% | 37.0%/38.1% | 42.0%/35.4% | 38.0%/33.8% | 28.0%/33.0% | 22.0%/33.1% |
| North America net sales | 387,497 | 426,305 | 471,692 | 516,503 | 562,989 | 608,028 | 653,630 |
| NA growth / EBIT margin | 9.8%/6.4% | 10.0%/6.9% | 10.6%/7.9% | 9.5%/9.3% | 9.0%/9.7% | 8.0%/10.1% | 7.5%/10.4% |
| International net sales | 142,906 | 161,894 | 179,115 | 197,027 | 216,729 | 236,235 | 255,134 |
| Intl growth / EBIT margin | 8.9%/2.7% | 13.3%/2.9% | 10.6%/3.4% | 10.0%/4.3% | 10.0%/4.7% | 9.0%/5.2% | 8.0%/5.6% |
| Group net sales | 637,959 | 716,924 | 827,133 | 963,912 | 1,125,245 | 1,286,537 | 1,448,338 |
| Group operating income (EBIT) | 68,593 | 79,975 | 110,514 | 144,885 | 181,726 | 219,637 | 260,793 |
| Group EBIT mgn / AWS share | 10.8%/58% | 11.2%/57% | 13.4%/61% | 15.0%/61% | 16.1%/64% | 17.1%/66% | 18.0%/68% |
Source: FY25 Form 10-K; Penmark Research Segments schedule.
Source: Company filings (A); Penmark Research estimates (E).
Revenue path and fixed-against-variable costs
Our revenue build is operationally constructive. AWS accelerates to 42.0% in FY27E ($250.4bn) as $225.3bn of new AI racks light up. Growth then moderates to 38.0% in FY28E and 22.0% by FY30E ($539.6bn) on scale and memory supply.
Fixed-cost leverage lifts Stores while fixed depreciation weighs on AWS. North America EBITDA margin expands from 10.6% in FY25A to 15.2% in FY30E via robotics, advertising mix and 1Q27 Leo satellite capitalisation. AWS EBITDA margin eases 3.0pp from FY26E to 54.5% in FY30E, holding AWS EBIT margin near 33%.
FY27E sales sit in line with the Street but EPS is 4% below, as debt peaks at $243bn
| Line & period | Penmark | Guide / Street | Gap | Analytical reason for gap |
|---|---|---|---|---|
| 4Q26E net sales | $243,064m | $240–247bn guide | −0.2% vs mid | AWS +42.0% ($50.5bn), NA +9.0% ($138.5bn), Intl +6.5%. |
| 4Q26E operating inc. | $31,125m | $28.5–33.5bn guide | +0.4% vs mid | AWS 37.0% EBIT margin ($18.7bn) and an 8.0% North America margin on holiday ad mix. |
| FY26E / FY27E sales | $827.1 / $963.9bn | $829.0 / $959.4bn | −0.2% / +0.5% | 42.0% FY27E AWS growth lifts sales 0.5% above Street. |
| FY26E / FY27E EPS | $12.59 / $10.25 | $12.80 / $10.68 | −1.6% / −4.1% | AWS server D&A ($54.2bn) and interest on $218.6bn of debt. |
| FY27E / FY28E AWS g | 42.0% / 38.0% | >40% / >40% guide | +2pp / −2pp | 2.31x backlog backs FY27E; HBM caps FY28E at 38.0%. |
| FY26E / FY27E capex | $220.0 / $289.6bn | ~$220bn / “>$220bn” | In line / +32% | Group FY27E capex of $289.6bn sits inside the $256.5–395.0bn Street range; AWS $225.3bn. |
Source: 3Q26 release and call guidance; Yahoo Finance / 24/7 Wall St. consensus; Penmark Research model.
Source: Penmark Research model.
Cash floor, working capital and balance sheet
We hold a $100.0bn cash floor across the forecast. Negative FCF through FY28E pushes carrying debt from $68.4bn in FY25A to $243.3bn in FY28E, lifting FY29E interest to $11.2bn.
Working capital absorbs about $9bn–$10bn a year as AWS receivables outgrow retail payables. Positive FCF from FY29E repays debt to $100.0bn by FY31E and unlocks $121.7bn of FY31E buybacks.
| Confirmed driver shift | FY27E EBIT / FCF | DCF ($/sh) | vs $201.58 | Mechanism in model |
|---|---|---|---|---|
| AWS FY27E capex +10pp | −$2.5bn / −$25.0bn | $199.52 | −$2.06 (−1.0%) | Lifts FY27E AWS capex past $250bn with flat sales. |
| AWS EBITDA mgn +3 / −4pp | +$7.5 / −$10.0bn EBIT | $219.16 / $178.14 | +$17.58 / −$23.44 | Trainium & owned shells vs AI compute & D&A drag. |
| AWS g: 47% FY27 / 26% FY28 | +$4.4 / −$11.0bn EBIT | $205.96 / $190.09 | +$4.38 / −$11.50 | Capex scales with AWS sales, muting growth upside. |
| NA EBITDA mgn +1.3 / −1.5pp | +$6.7 / −$7.7bn EBIT | $209.30 / $192.67 | +$7.72 / −$8.91 | Ad mix & Leo capex shift vs AI shopping ad risk. |
Source: Penmark Research model.
Our DCF values Amazon at $202 a share, 24% below our $264 P/E cross-check
We value Amazon primarily by a 14-year free-cash-flow-to-firm (FCFF) DCF through FY40E so the AI capex surge and 2030s harvest are both captured. Weighted average cost of capital (WACC), the blended annual return debt and equity holders require, is 10.95%. That blends an 11.40% cost of equity (5.10% Treasury yield, 5.0% equity risk premium [ERP], 1.26 beta) with 4.62% after-tax debt. Terminal growth, the steady perpetual growth rate after FY40E, is 3.50% at 15.0% RONIC. We reject EV/EBITDA (it ignores capex), FCF yield (FCF is negative) and a segment sum-of-the-parts (no pure-play AWS peer).
| FCFF DCF build (US$m) | 4Q26E | FY27E | FY28E | FY29E | FY30E | FY35E | FY40E |
|---|---|---|---|---|---|---|---|
| Operating income (EBIT, after SBC) | 31,125 | 144,885 | 181,726 | 219,637 | 260,793 | 464,242 | 575,553 |
| Cash taxes on EBIT | −3,548 | −14,489 | −21,807 | −30,749 | −41,727 | −92,849 | −115,111 |
| Net operating profit after tax (NOPAT) | 27,577 | 130,397 | 159,919 | 188,888 | 219,066 | 371,394 | 460,442 |
| Plus: P&E depreciation | 16,533 | 82,928 | 109,919 | 134,863 | 156,865 | 238,485 | 301,154 |
| Less: Cash capex & working capital | −25,613 | −298,178 | −318,309 | −320,953 | −321,160 | −357,272 | −379,983 |
| Free cash flow to firm (FCFF) | 18,497 | −84,853 | −48,471 | 2,798 | 54,771 | 252,607 | 381,613 |
| Mid-period discount factor (10.95% WACC) | 0.987x | 0.925x | 0.834x | 0.751x | 0.677x | 0.403x | 0.240x |
| Present value of FCFF | 18,257 | −78,477 | −40,406 | 2,102 | 37,093 | 101,770 | 91,461 |
Source: Penmark Research Valuation schedule (sum of 4Q26E–FY40E PV = $872,526m; FY40E TV = $4,906,682m, PV of TV = $1,116,461m; EV = $1,988,987m).
| Equity bridge & cross-check | US$m | $/sh |
|---|---|---|
| Enterprise value (30 Sep 2026) | 1,988,987 | 180.26 |
| Cash & marketable securities | 114,226 | 10.35 |
| Carrying debt (−$179,952m) & leases/oblig. | −204,173 | −18.50 |
| Income-tax contingencies | −7,300 | −0.66 |
| Strategic stakes after 21% tax on gain | 301,155 | 27.29 |
| Equity value at 30 Sep 2026 (11,034m sh) | 2,192,895 | 198.74 |
| Rolled forward to 17 Nov 2026 (x1.0143) | 2,224,250 | 201.58 |
| Cross-check: FY27E EPS $10.25 x 23.08x peer P/E | 2,609,573 | 236.50 |
| P/E cross-check + $27.29/sh AI stakes | 2,910,728 | 263.80 |
| FY27E P/E: at $257.89 price / at $231.47 target | 25.2x | 22.6x |
Source: Penmark Research Valuation and Peers schedules (operating leases sit in EBIT, not debt).
Source: Penmark Research model (hyperscaler bar = 16.8x–25.3x P/E + $27.29/sh AI stakes); Street targets via TickerNerd / 24/7 Wall St..
Why our P/E cross-check sits 31% (+$62 a share) above DCF
Peer forward P/E (six-peer median 23.08x: hyperscalers 19.6x, mass retail 42.5x) prices earnings after depreciation rather than upfront cash capex. Amazon spends $580.8bn above depreciation over FY27E–FY29E at an 11.40% cost of equity, and our model attributes most of the $62.21 a share gap to that near-term capex intensity.
Scenario-weighted value rolls forward to our $231 target
Source: Penmark Research Scenarios schedule ($/share at price date).
Source: Penmark Research Sensitivities schedule.
| Scenario & target build | Bear (20%) | Base (50%) | Bull (30%) | Prob.-weighted |
|---|---|---|---|---|
| AWS revenue growth (FY27E / FY28E) | 32.0% / 24.0% | 42.0% / 38.0% | 47.0% / 45.0% | 41.5% / 37.3% |
| AWS EBITDA margin (FY27E) | 54.0% | 57.0% | 58.5% | 56.9% |
| North America EBITDA margin (FY27E) | 10.5% | 13.5% | 14.8% | 13.3% |
| Anthropic mark vs IPO offer ($83) | 0.55x | 0.90x | 1.20x | 0.92x |
| DCF value per share at price date (17 Nov 2026) | $71.46 | $201.58 | $309.00 | $207.78 |
| Return to present value vs $257.89 close | −72.3% | −21.8% | +19.8% | −19.4% |
| 12-month target (value x 1.1140 cost of equity) | $79.60 | $224.56 | $344.23 | $231.47 |
| 12-month total return (0% dividend) / implied call | −69.1% (Sell) | −12.9% (Sell) | +33.5% (Buy) | −10.2% (Sell) |
Source: Penmark Research Summary, Thesis, Valuation and Scenarios schedules (all three scenarios use 10.95% WACC and 11.40% cost of equity).
Reconciling thesis range, fan chart and Page 1 target
Our revised present-value thesis range is $187.00–$228.60 a share, containing our $207.78 probability-weighted (PW) value. Compounding $207.78 forward 12 months at the 11.40% cost of equity gives our $231.47 target (a 10.24% negative return). Page 1's call box ($344.23 bull, $79.60 bear) and Exhibit 3's fan show 12-month rolled values, whereas DCF tables report present values ($71.46, $201.58 and $309.00).
Narrow Hold margin and limits of scenario weights
Conviction is low because our $231.47 target sits just $0.63 below the $232.10 Hold threshold. Any one of five bull catalysts, any of our four debates at the market view, or a 0.27% dip to $257.19 moves the call to Hold, while Buy requires a price below $210.43. Lowering WACC by 0.5pp lifts the PW target to about $251 (Hold), or $273.16 with a 1.0pp cut, and two-year beta 1.06 gives $269.77 (Hold). Scenario weights (20%/50%/30%) are analyst judgements, not statistical frequencies.
Six dated events over 12–18 months test FY27 capex discipline, AWS margins and Anthropic's valuation
Because our target sits just below the Hold line, upcoming capital-spending and margin disclosures carry immediate rating-change weight. Dates beyond management's announced schedule are estimated from historical reporting cadence.
| Date | Catalyst event | Penmark expectation vs guidance / Street | Driver & rating |
|---|---|---|---|
| Est. Dec 2026 | AWS re:Invent conference (Las Vegas) | Trainium3 customer benchmarks & Bedrock agent pricing; management guides external Trainium revenue to begin in late 2027. | Cloud margin durability (neutral) |
| 1Q27 | Amazon Leo satellite capitalisation begins | Launch & production costs move from P&L to capex in 1Q27 (lifts NA EBIT margin toward 9.3% FY27E vs 7.8% 9M26, zero FCF benefit). | Capex return hurdle (in base) |
| Est. 4 Feb 2027 | 4Q26 results & formal FY27 capex guide | 4Q26E sales $243.1bn (vs $243.5bn guide midpoint) & EBIT $31.1bn ($28.5–33.5bn guide); FY27E cash capex $289.6bn ($256.5–395.0bn Street range vs >$220bn guide); 4Q26E AWS EBIT margin 37.0% vs 38.7% in 3Q26. | Capex hurdle & margin (confirm Sell) |
| Est. late Apr / May 2027 | 1Q27 print & Anthropic 180-day lock-up expiry | 1Q27 GAAP vs ex-mark EPS volatility surfaces; Anthropic secondary supply tests our 0.90x haircut vs $83 IPO offer. | AI stake valuation (±$2.5–$7.5/sh) |
| Est. Jul–Oct 2027 | 2Q27 & 3Q27 peak FCF burn & debt prints | Trailing-12-month FCF approaches our −$63.0bn FY27E trough and carrying debt passes $210bn under the $100bn cash floor. | Capex return hurdle (confirm Sell) |
| Late 2027–1H 2028 | External Trainium sales & FY28 AWS pace | FY28E AWS slows to 38.0% vs management's >40% goal on memory/AI-lab concentration; external silicon tests +3pp margin bull case. | Cloud pace & margin (could move call) |
Source: Company guidance and filings; Penmark Research Thesis and Scenarios schedules (Street FY27E capex range includes Citi $288bn to JPM $395bn; dates estimated).
| Stock driver | Next number to watch | Thesis confirmed (keep Sell) if | Thesis broken (review for Hold/Buy) if |
|---|---|---|---|
| Capex return hurdle | FY27 cash capex guide (est. 4 Feb 2027): $289.6bnE | FY27 capex guided >$275bn and FY27E TTM FCF tracks worse than −$50bn. | FY27 capex guided <$250bn while AWS backlog conversion holds >40% growth. |
| Cloud margin durability | 4Q26 AWS operating margin (est. 4 Feb 2027): 37.0%E | AWS EBIT margin steps down toward 35%–36% in FY27E as AI D&A exceeds 21% of sales. | Custom silicon holds AWS EBITDA margin ≥58.5% and EBIT margin ≥38.5% through 2027. |
| Cloud pace | 4Q26 AWS y/y sales growth (est. 4 Feb 2027): 42.0%E | Sequential AWS dollar additions plateau or FY28E pace slips below 40%. | AWS sustains ≥45% y/y growth across FY27E–FY28E without capex/sales exceeding 30%. |
| AI stake valuation | Anthropic share price vs $83 offer (Nov 2026–May 2027) | Anthropic trades ≤$83 post-lock-up, capping after-tax AI stakes near $27–$30/sh. | Sustained >1.20x premium (>$100/sh) alongside OpenAI up-round (+$7.53/sh+). |
Source: Penmark Research monitoring framework derived from Segments, Valuation and Sensitivities schedules.
What would change our rating from Sell
Our target sits just below the Hold threshold, as detailed on the valuation page. Three single events would move the call to Hold: a sustained 50bp fall in the 10-year Treasury yield ($18.32 a share), a 3pp higher AWS margin from Trainium ($17.58 a share), or FY27 capex below $250bn alongside 45% AWS growth. A Buy rating would require the share price below $210.43 with fundamentals intact.
Lower Treasury yields and Trainium (about +$18/sh each) lead upside risks to our Sell
Source: Penmark Research Thesis and Sensitivities schedules ($/share vs $201.58 base DCF).
| Risk / catalyst | Prob. | $/sh impact | Mechanism and what would end our thesis |
|---|---|---|---|
| 10Y UST yield eases −50bp | 30% | +$18.32 | Upside risk to Sell: Trims discount rate 50bp ($219.90 base DCF), lifting our target into Hold on its own. |
| Trainium & owned shells | 25% | +$17.58 | Upside risk to Sell: Custom silicon lifts FY27E–FY40E AWS EBITDA margin +3pp, offsetting AI D&A. |
| Retail ad & grocery margin | 30% | +$7.72 | Upside risk to Sell: Robotics, ads & 1Q27 Leo capex shift (+1.3pp NA mgn); 16 Nov Leo launch failure highlights execution risk. |
| Anthropic re-rates to 1.2x | 40% | +$7.53 | Upside risk to Sell: Sustained secondary trading 20% above $83 IPO offer ($34.72/sh AI stakes). |
| AWS FY27E growth hits 47% | 35% | +$4.38 | Upside risk to Sell: Faster backlog burn (+5pp FY27E); combined with margin upside gives $309 bull. |
| AI mix & D&A cut AWS −4pp | 30% | −$23.44 | Downside catalyst: GPU pricing competition and 5-year D&A push AWS margin toward FY22 lows. |
| AI-lab concentration / HBM | 25% | −$11.50 | Downside catalyst: Memory bottlenecks or AI-lab funding cuts slow FY28E AWS growth to 26%. |
| Agentic shopping bypass | 20% | −$8.91 | Downside catalyst: External AI shopping agents bypass Sponsored Products, cutting NA margin −1.5pp. |
Source: Penmark Research Thesis, Scenarios and Sensitivities schedules.
Regulatory and legal risks
The US Federal Trade Commission (FTC) and 17-state antitrust suit (trial expected late 2026 or 2027) targets Prime bundling and seller pricing rules. Unbundling Fulfillment by Amazon from Buy Box placement would pressure Amazon's 61% worldwide third-party unit mix and ad yields. European Union Digital Markets Act rules separately cap self-preferencing in International.
Power-grid and Leo launch risks
Grid hookup delays and satellite setbacks move DCF value directly. AMZN fell 3.3% to $257.75 on 16 November 2026 after an Amazon Leo launch failure, threatening our North America margin bull case on 1Q27 cost capitalisation. Meanwhile, $2.7bn of Blue Origin launch contracts with Executive Chair Jeff Bezos carry related-party scrutiny.
Crowded consensus (59 Buy, 0 Sell) and under 1% short interest leave sharp downside on capex surprises
Institutional 13F filers hold 65.6% of Amazon's common shares alongside Founder Jeff Bezos's 8.8% beneficial stake, while short interest is under 1% of float. Near-unanimous bullish Street positioning explains why 3Q26 capex guidance sparked an immediate 16% intraday sell-off.
| Holder | Shares (m) | % out | q/q (m) |
|---|---|---|---|
| Jeffrey P. Bezos (Founder / Chair) | 950.43 | 8.84% | Proxy |
| The Vanguard Group | 861.13 | 7.96% | +2.35 |
| BlackRock, Inc. | 747.88 | 6.92% | −0.52 |
| State Street Corp. | 400.49 | 3.70% | +3.30 |
| FMR LLC (Fidelity) | 322.64 | 2.98% | −43.02 |
| Geode Capital Management | 236.35 | 2.19% | +3.14 |
| Morgan Stanley | 186.80 | 1.73% | +7.43 |
| JPMorgan Chase & Co. | 179.42 | 1.66% | +11.44 |
| Total 3Q26 13F institutional filers | 7,095.53 | 65.62% | −1.22pp |
Source: 2026 DEF 14A proxy (Bezos as of 24 Feb 2026); 3Q26 vs 2Q26 SEC Form 13F filings (30 Sep 2026; 10,813.1m cover shares).
| Metric | Reading / status |
|---|---|
| 3-mth avg. daily vol. (ADTV, 17 Nov) | $10,091.1m (~39.1m sh/day) |
| 52-week close range | $198.79–$284.02 ($257.89 last) |
| Common / diluted shares | 10,811m basic / 11,034m diluted |
| Short interest (30 Oct 2026) | 94.36m sh (0.96% of float; ~2.4d) |
| Major index memberships | S&P 500, Nasdaq-100, DJIA, Russell 1000 |
| Voting structure | Single class ($0.01 par; 1 vote/sh) |
| Latest insider Form 4 | Herrington 10b5-1 trades 2 and 16 Nov (Form 4s filed 4 and 17 Nov; 479.3k sh) |
| Strategic equity lock-up | Anthropic ~180d post-IPO (est. May '27) |
| Share repurchase outlook | $0 modelled FY26E–FY30E (cash floor) |
Source: SEC filings; Yahoo Finance; MarketBeat / ChartMill short-interest data; Penmark Research.
Source: TickerNerd / 24/7 Wall St., 14–18 Nov 2026; Penmark Research.
Source: 24/7 Wall St. consensus snapshot, 18 Nov 2026 (61 analysts: 16 Strong Buy, 43 Buy, 2 Hold, 0 Sell).
What positioning means for risk-reward: 96.7% of covering analysts rate the stock Buy with a $326.29 mean target, and active funds remain heavily invested despite Fidelity trimming 43.0m shares in 3Q26. With zero buyback support modelled before FY31E, negative cash-flow surprises meet thin marginal buyer support. Meanwhile, 0.96% short interest creates negligible short-squeeze risk against our Sell.
Linked annual statements, quarterly cadence and segment summary: FCF less SBC turns positive in FY30E
| Line ($m unless stated) | FY23A | FY24A | FY25A | FY26E | FY27E | FY28E | FY29E | FY30E |
|---|---|---|---|---|---|---|---|---|
| Income statement & per-share data | ||||||||
| Net sales | 574,785 | 637,959 | 716,924 | 827,133 | 963,912 | 1,125,245 | 1,286,537 | 1,448,338 |
| Net sales growth (%) | 11.8% | 11.0% | 12.4% | 15.4% | 16.5% | 16.7% | 14.3% | 12.6% |
| Total depreciation & amortisation | 48,663 | 52,795 | 65,756 | 84,019 | 112,810 | 144,802 | 174,745 | 201,763 |
| EBITDA | 85,515 | 121,388 | 145,731 | 194,534 | 257,695 | 326,527 | 394,383 | 462,556 |
| Operating income (EBIT) | 36,852 | 68,593 | 79,975 | 110,514 | 144,885 | 181,726 | 219,637 | 260,793 |
| EBIT margin (%) | 6.4% | 10.8% | 11.2% | 13.4% | 15.0% | 16.1% | 17.1% | 18.0% |
| Net income (GAAP) | 30,425 | 59,248 | 77,670 | 137,693 | 113,066 | 140,377 | 169,797 | 203,821 |
| Diluted EPS / ex AI marks ($) | 2.90 | 5.53 | 7.17 | 12.59/7.59 | 10.25 | 12.72 | 15.39 | 18.47 |
| Diluted shares (m) | 10,492 | 10,721 | 10,827 | 10,935 | 11,034 | 11,034 | 11,034 | 11,034 |
| Cash flow, condensed balance sheet & returns | ||||||||
| Operating cash flow (OCF) | 84,946 | 115,877 | 139,514 | 173,528 | 226,573 | 283,426 | 340,682 | 398,340 |
| Cash capex (outflow) | −52,729 | −82,999 | −128,320 | −220,000 | −289,561 | −308,145 | −310,792 | −310,966 |
| Free cash flow (OCF less capex) | 32,217 | 32,878 | 11,194 | −46,472 | −62,988 | −24,719 | 29,891 | 87,374 |
| Free cash flow less SBC | 8,194 | 10,867 | −8,273 | −67,978 | −88,050 | −53,975 | −3,559 | 49,717 |
| Cash & marketable securities | 86,780 | 101,202 | 123,029 | 124,360 | 100,000 | 100,000 | 100,000 | 100,000 |
| Property & equipment, net | 204,177 | 252,665 | 357,025 | 518,647 | 725,280 | 923,505 | 1,099,435 | 1,253,536 |
| Debt carrying value | 58,314 | 52,623 | 68,396 | 179,952 | 218,581 | 243,300 | 213,409 | 126,035 |
| Shareholders' equity | 201,875 | 285,970 | 411,065 | 645,755 | 783,883 | 953,516 | 1,156,762 | 1,398,240 |
| Operating ROIC / ROE (%) | n.d./17.5% | n.d./24.3% | 21.2%/22.3% | 20.1%/26.1% | 17.8%/15.8% | 16.5%/16.2% | 15.9%/16.1% | 16.0%/16.0% |
| Debt less cash / EBITDA (x) | −0.33x | −0.40x | −0.37x | 0.29x | 0.46x | 0.44x | 0.29x | 0.06x |
Source: 10-K filings (FY23A–FY25A); Penmark Research Statements schedule (FY26E–FY30E; ROIC on average operating capital from FY25A; FY23A–24A LT debt ex $5,017m FY24A current portion; FY23A–24A capex = gross).
| Segment & KPI detail ($m) | 4Q25A | 1Q26A | 2Q26A | 3Q26A | 4Q26E | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|---|---|---|---|
| North America net sales | 127,083 | 104,143 | 116,177 | 112,852 | 138,520 | 426,305 | 471,692 | 516,503 | 562,989 |
| International net sales | 50,724 | 39,789 | 42,197 | 43,108 | 54,021 | 161,894 | 179,115 | 197,027 | 216,729 |
| AWS net sales | 35,579 | 37,587 | 42,232 | 45,984 | 50,522 | 128,725 | 176,325 | 250,382 | 345,527 |
| AWS y/y sales growth (%) | 23.6% | 28.4% | 36.8% | 39.3% | 42.0% | 19.7% | 37.0% | 42.0% | 38.0% |
| North America EBIT | 11,472 | 8,267 | 9,123 | 8,723 | 11,082 | 29,619 | 37,195 | 47,900 | 54,646 |
| International EBIT | 1,040 | 1,424 | 1,717 | 1,580 | 1,351 | 4,750 | 6,072 | 8,475 | 10,216 |
| AWS EBIT | 12,465 | 14,161 | 16,621 | 17,773 | 18,693 | 45,606 | 67,248 | 88,510 | 116,863 |
| AWS EBIT margin (%) | 35.0% | 37.7% | 39.4% | 38.7% | 37.0% | 35.4% | 38.1% | 35.4% | 33.8% |
| Group cash capex | 38,469 | 43,200 | 53,100 | 75,900 | 47,812 | 128,320 | 220,000 | 289,561 | 308,145 |
| AWS backlog ($bn, period end) | n.d. | 364.0 | 496.0 | 578.0 | n.m. | n.d. | n.m. | 2.31x rev | n.m. |
Source: Earnings releases and Form 10-Q/10-K filings; Penmark Research Historicals and Segments schedules.
GAAP-to-operating reconciliations, glossary, ratio definitions, peer set and methodology note
| Reconciliation line ($m unless stated) | FY24A | FY25A | FY26E | FY27E | FY28E |
|---|---|---|---|---|---|
| GAAP net income | 59,248 | 77,670 | 137,693 | 113,066 | 140,377 |
| Less: pre-tax investment marks & other non-op income | 2,250 | −15,229 | −68,325 | 0 | 0 |
| Tax effect on non-operating items | −304 | 2,987 | 13,665 | 0 | 0 |
| Net income excl. investment marks / Diluted EPS ex marks ($) | 61,194 / 5.71 | 65,428 / 6.04 | 83,033 / 7.59 | 113,066 / 10.25 | 140,377 / 12.72 |
| Operating cash flow (GAAP) less cash capex = FCF | 32,878 | 11,194 | −46,472 | −62,988 | −24,719 |
| Less: stock-based compensation (SBC) expensed as cash | −22,011 | −19,467 | −21,505 | −25,062 | −29,256 |
| Free cash flow less SBC / FCFF (NOPAT + P&E D&A − capex ± WC) | 10,867 / n.m. | −8,273 / n.m. | −67,978 / +18,497* | −88,050 / −84,853 | −53,975 / −48,471 |
Source: 10-K/10-Q filings; Penmark Research Statements and Valuation schedules (*FY26E FCFF = 4Q26E stub from 30 Sep 2026 valuation date).
| Term / peer | Definition / basis |
|---|---|
| AWS RPO | Remaining performance obligations ($578bn 3Q26; contracted multi-year cloud spend). |
| 3P / FBA | Third-party marketplace sellers (61% of units) & Fulfillment by Amazon logistics. |
| Trainium / Leo | AWS custom AI training silicon; low-Earth-orbit broadband constellation (capex 1Q27). |
| FCFF | NOPAT + P&E depreciation − cash capex ± op. WC; SBC treated as cash cost. |
| ROIC / RONIC | After-tax EBIT ÷ avg. (net PP&E + ROU − net op. WC liab.); terminal RONIC 15%. |
| EV & Net debt | Mkt cap + $97.2bn net debt (incl. $7.3bn tax conting.) − AI stakes = $2,641.7bn EV. |
| Hyperscalers | MSFT (25.3x NTM), GOOGL (18.3x CY27E), ORCL (16.8x NTM), META (20.9x NTM). |
| Mass retail | WMT (41.0x NTM), COST (44.0x NTM); 6-peer median forward P/E = 23.08x. |
Source: Penmark Research Peers and Methodology schedules (NTM = next twelve months; CY27E = calendar-year 2027 estimate).
| Topic | Treatment & source cut-off |
|---|---|
| Archetype | Hybrid: AWS (capacity × utilisation × price); Stores (units × price + 3P take + ads + subs). |
| Valuation | 14.25yr FCFF DCF ($201.58 base) + 6-peer P/E cross-check ($263.80); SOTP & EV/EBITDA rejected. |
| Target & rule | 30 Sep '26 DCF rolled to 17 Nov '26 ($257.89); prob.-weighted (PW) $207.78 × 1.114 = $231.47 (Sell). |
| Guidance gap | FY28E AWS modelled at 38% vs >40% guide (HBM memory risk & slower $3.8bn q/q 3Q26 add). |
| SBC & leases | SBC expensed in FCFF; 223m awards in 11,034m shares; op. leases in EBIT, not bridge. |
| AI stakes | Anthropic ($238.7bn × 1.448x IPO × 0.9x − 21% tax = $249.6bn) + OpenAI ($50bn cost). |
| Cash & debt | $100bn cash floor from FY27E; forecast debt carries 4.6% interest (WACC pre-tax debt 5.85%). |
| Data sources | 10-K (6 Feb '26), 3Q26 10-Q (30 Oct '26), DEF 14A (24 Feb '26), prices/UST (17 Nov '26). |
Source: SEC filings; Penmark Research Methodology schedule (19 Nov 2026).
Analyst certification and important disclosures
Ratings
Ratings are based on the total return from the share price to our 12-month price target, including dividends. The price target is weighted across bull, base and bear scenarios.
| Buy | Total return to price target above +10% |
| Hold | Total return to price target from −10% to +10%, inclusive |
| Sell | Total return to price target below −10% |
We review ratings at each set of results and whenever the facts change materially. A share price can move outside its band before the rating changes, and targets may not be reached.
Analyst certification
The Penmark Team certifies that (1) the views in this report accurately reflect its views about the subject company and its securities, and (2) no part of its compensation was, is or will be directly or indirectly related to the specific ratings, price targets or views in this report.
Contact
Penmark Research
Questions about our research:
research@penmarkresearch.com
All our research and coverage:
penmarkresearch.com
Basis of preparation
This report uses information publicly available up to its publication date. Share prices are closing prices on the trading day before the publication date, unless stated otherwise. Forecasts, valuations and price targets are Penmark Research’s own, and forecast periods are marked E. Sources are shown beneath each exhibit.
Valuation and price targets
Price targets are based on the valuation methods set out in each report and are weighted across bull, base and bear scenarios. The main risks to each rating and price target are described in the report. Penmark Research updates its coverage after each set of results but is not obliged to update a report at any other time.
Independence and conflicts of interest
Penmark Research is independent. It does not provide investment banking, brokerage, market-making or asset management services, does not trade securities, and does not accept payment from any company for research coverage. Companies do not see or review Penmark Research reports before publication.
No member of the Penmark Team holds a long or short position, directly or through derivatives, in the securities of any company Penmark Research covers.
Regulatory status
Penmark Research publishes general, impersonal investment research on a regular basis. It is not registered as an investment adviser with the US Securities and Exchange Commission or any state securities authority, is not a registered broker-dealer and is not a member of FINRA. Penmark Research does not give advice tailored to any person’s circumstances.
Not investment advice or an offer
This report is for information only. It is not personalised investment advice and is not an offer or solicitation to buy or sell any security or other financial instrument. It does not take into account any reader’s objectives, financial situation or needs. Readers should make their own investment decisions and seek independent professional advice where appropriate.
Forward-looking statements and risk
Forecasts, valuations and price targets are estimates based on assumptions that may prove wrong, and they may not be achieved. Opinions are current as of the publication date and may change without notice. Investing in equities involves risk, including the loss of capital. Share prices can be volatile, and past performance is not a reliable indicator of future results.
Data and sources
Company financials, operating data and filings come from company filings, earnings releases and earnings calls. Market data and any consensus estimates come from third-party sources named beneath the relevant exhibit; consensus figures from aggregators are unofficial. Penmark Research does not use other brokers’ research. Third-party data belong to their providers and may not be extracted from this report.
Accuracy and liability
Information is taken from sources believed to be reliable, but Penmark Research makes no representation or warranty as to its accuracy, completeness or timeliness. To the fullest extent permitted by law, Penmark Research accepts no liability for any loss arising from use of this report. Descriptions of litigation and regulatory matters reflect allegations and public records, not findings, unless stated. If a published report is corrected, the corrected version carries an “Updated” date at penmarkresearch.com.
Distribution and access
Penmark Research reports are available to registered users of penmarkresearch.com. Downloads and programmatic access are governed by the Penmark Research subscription terms. This report is not directed at, or intended for, any person in Singapore or India, and access from those countries is blocked. It is not intended for distribution to, or use by, any person in any jurisdiction where that would be contrary to law or regulation or would subject Penmark Research to registration or licensing requirements there.
Copyright and permitted use
© Penmark Research. All rights reserved. Penmark Research is a trade name of Brightriver Labs. Readers may share a link to this report but may not copy, reproduce, redistribute, sell or use its content, including for text and data mining, without written consent. Contact: research@penmarkresearch.com.