We value AMZN at $327 a share. The report's $248.47 price (Oct 26, 2026) was 24% below fair value: Strong Buy.
Restated Oct 26, 2026 (method case-spread-2026-10-24); the report's original value was $327, Buy.
- Five-year return
- 16.5%
- a year, at the report's price
- Return
- +1.6%
- vs SPY
- +1.0%
- vs benchmark
- +0.3%
Close of Oct 26, 2026 to close of Oct 27, 2026
What would change the view
- AWS operating margin below 34% in Q3 or Q4 2026 while AWS growth is under 38%: depreciation is outrunning revenue; toward the pessimistic case
- 2027 capex guided above $412B with AWS growth under 35% in Q4 2026: property per dollar of revenue rising, not falling; the value reaches the price (each 0.10x every year is about -$16)
- AWS growth below 25% in any 2027 quarter with capex above $200B a year: the pessimistic case ($120)
- AWS backlog growth outside Anthropic and OpenAI stalls (under $15B a quarter for two quarters) while Anthropic moves more compute to dedicated sites: the sold-out reading behind 33% 2027 growth weakens
- 2029 capex guided at or above 2028's with no lift to AWS backlog: the 2028 property per dollar is lasting, not gliding back; keeping it for good is $268, still a Buy at this price
Catalysts
- Microsoft and Alphabet report after the close. Their 2027 capex wording, and whether the market rewards it, sets the bar Amazon's capex outlook meets the next day
- Q3 2026 results after the close (confirmed by Amazon). Ours: AWS $46.2B (+40%) at 36.5%, 2026 capex outlook about $225B, EPS $2.06 with no stake marks; the S-1 shows no priced Anthropic round in the quarter, so none is expected. Options imply about a 7.1% move. What it can settle: AWS growth and margin into 2027, and any first word on 2027 capex. What it cannot: the exit multiple
- Q4 2026 results with the first 2027 capex figure (date not announced). Our central path is $258B, our model's output and not Amazon guidance; with 2028 held at the 2027 level, each further $32B with no more revenue (0.10x property per dollar every year) is about -$16 a share, and about $412B reaches the price. Q4 may also carry the Anthropic IPO mark: $0.11 of EPS per $1 of IPO price against $58.90 on the preferred, plus $1.0-1.5 of EPS if the notes convert up to the 12% cap
- FTC antitrust trial (date held on 2026-10-09, one press source). Central legal claims $5B, pessimistic $20B
This text argues the original call. The current figures are at the top.
We keep Amazon a Buy and raise fair value to $327 from $319, against
$248.47. Nothing about the business or the forecast changed since the
fifth report this morning. One input changed: the long-run return on new
capital that the long fade uses after 2031. Our four hyperscaler reports
set it four ways; the owner approved one rule for all of them
(valuation.md, study
2026-10-26-hyperscaler-return-rule).
The rule. Each case earns, on growth after 2031, its own year-five
return on its net capital: year-five after-tax operating profit over 2026
net property of AWS and the stores ($543.4B: 2025's from the 10-K plus
2026 additions less depreciation, model.py) plus the case's 2027-31
capital spending less depreciation. Growth then needs the same capital per
revenue dollar the case reaches in 2031, about 0.68 times revenue. The
returns are 13.7%, 21.2% and 25.4% for the pessimistic, central and
optimistic cases (output.txt), floored at the 9.44% WACC and capped at
Amazon's median five-year return on the same measure, 30.9%; neither
binds. The fifth report took 11%, 15% and 18% from Amazon's after-tax
operating income over average net property in 2017-25, a whole-company
average that assumed growth after 2031 needs more capital per revenue
dollar than our own forecast holds.
What it is worth. +$8.19 a share, from $319.12 to $327.31 (output.txt). The Buy holds under every capital choice we tested (below).
What changed from the fifth report
| Step | Pessimistic | Central | Optimistic | Weighted |
|---|---|---|---|---|
| The fifth report (2026-10-26-amzn-5) | $117.82 | $319.12 | $548.11 | $326.04 |
| Each case's long-run return holds its own 2031 capital per revenue dollar | +$2.59 | +$8.19 | +$10.26 | +$7.31 |
| This report | $120.41 | $327.31 | $558.37 | $333.35 |
The fifth report's capital rule stands: 2028 capex is held at the 2027 level, $258B, and the excess property glides back onto our AWS path by 2030. Central capex: $258B, $258B, $192B, $143B, $202B, $225B for 2027-2032. Revenue, margins, the Anthropic stake, depreciation, the Q3 2026 forecasts and the exit multiple are unchanged; the third report's Q3 forecasts stay the graded set for 2026-10-29.
The judgment range
All from capex_consistency.py (output in capex_consistency.txt); the
rating uses restate.view() at $248.47.
| Capital choice (central and optimistic) | Central capex 2027-2032 ($B) | Central | Case-weighted | Rating |
|---|---|---|---|---|
| This report: 2028 held, back on the path by 2030 | 258, 258, 192, 143, 202, 225 | $327 | $333 | Buy |
| Same, pessimistic case held too | same | $327 | $333 | Buy |
| The fourth report's floor: 2028 level kept for good | 258, 258, 248, 261, 271, 287 | $268 | $282 | Buy |
| Three-year glide: on the path only in 2031, the exit year | 258, 258, 211, 182, 155, 225 | $331 | $336 | Buy |
| Microsoft's reviewer path scaled to Amazon (not Amazon evidence) | 258, 258, 214, 214, 235, 256 | $296 | $310 | Buy |
| The third report: 2028 from the rule, cut | 258, 191, 193, 191, 202, 225 | $329 | $334 | Buy |
| This report, AWS depreciation +2 points | 267, 267, 205, 157, 216, 239 | $309 | $318 | Buy |
| This report, AWS depreciation −1.5 points | 252, 252, 182, 132, 193, 214 | $341 | $345 | Buy |
Under defensible capital and depreciation choices the central runs from $268 (the permanent floor) to $341, and every one is a Buy at $248.47. We reject the three-year glide because it puts the lowest-capex year in 2031, the year the exit capitalises.
Forecast and value
Central case (output.txt): revenue $958.9B in 2027 and $1,447.1B in 2031, as before. Operating margin 14.2% in 2028 and 17.8% in 2031 as the lighter plant after 2029 lowers depreciation. The return on new capital over 2027-2031, volume only, is 20.1%; the long-run return, holding the central's own 2031 capital per revenue dollar, is 21.2%.
Sensitivities (output.txt, each alone): AWS growth ±3 points a year −$21.5 / +$25.3; AWS cash margin 53% / 59% −$19.9 / +$19.9; AWS property per revenue ±0.15x every year −$24.3 / +$25.0; ads margin 40% −$24.2; mature return 11% −$16.2 and 18% −$4.2; stakes halved −$9.5; Amazon's own beta −$23.2; outside view at 20x / 30x −$37.5 / +$34.6.
Why the other side could be right
The case for the fourth report's number: AI plant may stay heavier per dollar of revenue than any earlier build, because accelerators wear out faster and inference demand keeps rising; then the 2028 level is the new normal and the value is $268, still a Buy. The long-run return could sit below the case's own: at the fifth report's 15% the central is $319. The case for lower still: AWS growth well below our path. The price needs AWS growth 14.3 points a year below ours, or a 14.3x exit multiple; either is the pessimistic world ($120), not a variant of ours.
Catalysts and falsifiers
- 2026-10-28, Microsoft and Alphabet results: their 2027 capex wording.
- 2026-10-29, Q3 2026 results (the third report's preview and forecasts).
- Early February 2027, Q4 results with the first 2027 capex figure; ours $258B. About $412B with no more AWS revenue than our path takes the value to the price.
- Falsifiers: AWS margin below 34% in Q3 or Q4 2026 with growth under 38%; 2027 capex above $412B with Q4 AWS growth under 35%; AWS growth below 25% in any 2027 quarter with capex above $200B a year; backlog growth outside the two labs under $15B a quarter for two quarters; a 2029 capex figure at or above 2028's with no lift to AWS backlog, which points to the fourth report's floor, $268 under this report's rule.
- Anthropic tripwires (added by the amendment below). None moves the
central by itself; each raises the weight on the pessimistic case, where
the labs miss and AWS capacity sits idle.
- V1. Amazon books a quarterly loss on its Anthropic stake. Not
possible in Q3 (no priced round; note value changes go to other
comprehensive income); at the IPO, a price below about $56.27, the
preferred's carrying value a share (inference,
funding.txt). - V2. Anthropic's IPO prices, or trades for 30 days, below $58.90, the Series H price a common share. Each $1 below is $1.64B pre-tax on the preferred, $0.12 of EPS.
- V3. Broadcom's notes to Anthropic step from 9% to 13% paid in kind (they do when Anthropic misses revenue conditions; S-1 p. 167), or Amazon's $15.0B facility stops being drawn because compute-capacity milestones are missed (S-1 p. 164-167; Amazon 10-Q Note 2).
- V4. Vendor money rises above 25% of the new compute Anthropic
signs. About 20% today: $82B of vendor facilities against about
$417B of minimum payments signed or amended January 1 to June 1, 2026
(S-1 p. 164-169, F-64; our arithmetic in
../2026-10-26-googl-5/funding.txt).
- V1. Amazon books a quarterly loss on its Anthropic stake. Not
possible in Q3 (no priced round; note value changes go to other
comprehensive income); at the IPO, a price below about $56.27, the
preferred's carrying value a share (inference,
Limits
- Depreciation and exit are not yet one rule across the three companies. Amazon's AWS depreciation (13.5% rising to 16%) and asset lives stay its own, as do Alphabet's (14.2% steady, 25-year long life) and Microsoft's (18.3% steady, 20-year long life). Microsoft's exit capitalises free cash flow; Amazon's and Alphabet's capitalise operating profit. A shared treatment for both is follow-up work after this week's results.
- The 2030 capex dip ($143B) comes from a two-year glide that absorbs the 2028 excess in one step. A three-year glide with the trough in 2031 gives about the same value ($331).
- The long-run return holds 2031's capital per revenue dollar for good; that is a judgment, the case's own level rather than a rising or a falling one.
The S-1 reading, the stake bridge, the earnings-with-marks table, the
reaction hypotheses and the Q3 preview are the third report's
(../2026-10-26-amzn-3/report.md) and are not repeated here.
Funding the build, and the Anthropic marks (amzn-5 amendment, unchanged)
Added to amzn-5 the same day, before the 2026-10-29 print; it changes no
fair value, rating or forecast and stands as written. The Amazon
arithmetic is in funding.py (output funding.txt); the four-company
table is ../2026-10-26-googl-5/funding.txt.
Amazon spends more than its operating cash flow and borrows the difference. Fact (10-Q cash flow statement): Q2 2026 operating cash flow $45.4B, purchases of property and equipment $54.2B, 119% of it (120% with $0.4B of finance-lease principal; 117% net of $1.1B of sale proceeds and incentives, the basis of Amazon's free cash flow). Trailing twelve-month free cash flow was −$7.6B (10-Q). Q2 funding: $13.6B of long-term debt ($67.0B in the first half), no equity. Q3 funding (8-Ks): $25.0B of dollar notes closed July 9 and £4.25B of sterling notes closed September 14; a $17.5B delayed-draw term loan was open to September 30 (10-Q Note 5), its draw not yet disclosed.
Q3 and the fund's bear reading (two or more of Microsoft, Alphabet, Amazon and Meta with Q3 capex above operating cash flow, with debt or stock funding the gap). Meta is already there: Q3 capex 103.5% of operating cash flow, 106.5% with finance leases (10-Q, 2026-10-22; met from cash raised by May's notes, with $35B of stock and mandatory convertible preferred filed October 26). Alphabet, reporting a day before Amazon, is likely to be the second (its preview, section 8). Amazon is likely to be a third: about $220B of 2026 cash capex (Q2 call) less $96.3B in the first half leaves about $56-62B for Q3, against Q3 operating cash flow we infer at $42-52B (Q3 2025 $35.5B; Q1 and Q2 2026 up 53% and 40% on the year). July's notes already supply the debt. The deciding number is Q3 "Net cash provided by (used in) operating activities" against "Purchases of property and equipment" in the release's cash flow statement. Our model already has Amazon building $258B in 2027, so a tripped reading confirms the path we value rather than adding to it (inference); what would count against us is Q3 capex above $62B with AWS growth below our 40%.
The Anthropic mark: none in Q3; the IPO sets the next, likely in Q4.
| Item | Amount | Quarter | Source |
|---|---|---|---|
| Series G-2: 19.3M preferred, 0.193B common shares at $25.91 | step-up to $58.90: $6.37B pre-tax | Q2, already booked inside the $50.5B of upward adjustments ("primarily" Anthropic preferred) | S-1 F-47, F-64; 10-Q Note 1 |
| Series H: $5.0B in May at $589.0095 a preferred share | 0.085B common shares | Q2 | S-1 F-43; 10-Q Note 2 and MD&A ("$10.0 billion" of Anthropic preferred in Q2) |
| No priced round in Q3 | none in EPS; note value changes go to other comprehensive income | Q3 | S-1 read; 10-Q Note 2 |
| IPO price against $58.90, preferred | $1.64B pre-tax per $1; EPS $0.12 per $1 | Q4, if it prices this year | funding.txt |
| Notes converting at the IPO under Amazon's 12% cap | $9.5-21.2B of other comprehensive income moved into other income, EPS $0.67-1.49 | Q4 | funding.txt |
The third report counted Amazon's preferred as 1.559B common-equivalent shares, the series the S-1 itemises; the $5.0B of Series H Amazon bought in May adds 0.085B, so 1.644B. At $58.90 that is $96.8B against $92.5B carried, about $56.27 a share carried (inference: a discount Amazon applies for the non-voting class or other rights; it may or may not close at the IPO). No value changes: the model counts the stakes at their Q2 carrying values after tax. Counting the Series H shares against the cap also lowers the note shares that can convert at the IPO, hence the wider $9.5-21.2B range (the third report: $14-21B).
The notes: $6.2B of principal that Anthropic carries at $108.6B (S-1 F-6, $108,566M), about $58.99 a share on 1.84B shares; Amazon carries them at $97.9B fair value, about $53.19 a share, with a $92.0B gain held in other comprehensive income (10-Q Note 2).
EPS (GAAP, diluted; the third report): Q3 $2.06 without marks, and the same with Anthropic. Q4, if the IPO prices then: at $47.12 (20% below $58.90) −$1.37 on the preferred, −$0.69 to +$0.13 with the notes' conversion; at $58.90 nil on the preferred, +$0.67 to +$1.49 with the notes; at $70.68 (20% above) +$1.37, +$2.04 to +$2.86 with the notes. A Q4 gain is earnings noise, not new value.
Backlog. Anthropic's Amazon minimum is at least $110.0B, May 2026 to April 2036, with a true-up (S-1 p. 172), about 22% of the $496B AWS backlog at June 30. Signed in April with no later amendment disclosed, it was in the June backlog already: no Anthropic step in Q3's (the S-1 read, finding 2).
Revision history
| Report | View | Pessimistic | Central | Optimistic | Central change | Price | Blinding |
|---|---|---|---|---|---|---|---|
| 2026-10-15-amzn | hold | 98 | 255 | 418 | first | 254.06 | partly unblinded |
| 2026-10-22-amzn-2 | buy | 150 | 321 | 443 | +66 | 250.38 | unblinded |
| 2026-10-26-amzn-3 | buy | 118 | 320 | 550 | −1 | 249.87 | unblinded |
| 2026-10-26-amzn-4 | buy | 118 | 266 | 468 | −54 | 249.87 | unblinded |
| 2026-10-26-amzn-5 | buy | 118 | 319 | 548 | +53 | 248.47 | unblinded |
| 2026-10-26-amzn-6 | buy | 120 | 327 | 558 | +8 | 248.47 | unblinded |
Sources
- The third report's sources (
../2026-10-26-amzn-3/report.md), unchanged: Amazon's 10-Q and 10-K filings and calls, Anthropic's S-1, Reuters (2026-10-23), market data market data. - The long-run-return rule:
valuation.md("Mature return: the case's own capital held") and the study 2026-10-26-hyperscaler-return-rule. - The shared capital rule: amzn-5,
../2026-10-26-msft-4/and../2026-10-26-googl-5/, set 2026-10-26 after two independent reviews (an Opus reviewer and a Codex reviewer) of the fourth reports. - Market quote, 2026-10-26 16:16 UTC.
- The amendment: Amazon's 10-Q for Q2 2026 and 8-Ks of 2026-06-10,
06-12, 07-09 and 09-14 (SEC EDGAR, pinned in
data-manifest.json); Anthropic's S-1 (accession 0001628280-26-067168); the four-company sources listed in../2026-10-26-googl-5/funding.py.
This work was done with Claude, a model made by Anthropic, whose S-1 the third report reads: a conflict of interest.
Details
- Report date
- Oct 26, 2026
- Price source
- Last regular-session trade $248.47 at 2026-10-26 16:16:15 UTC; official close 2026-10-23 $248.00 (sip-list-exchange-close), Oct 26, 2026, 9:16 AM PDT
- Bear / bull
- $120 / $558
- Exit price
- $505 (on Dec 31, 2031)
- Confidence
- 28%
- Price against our scenarios
- The price is justified in over 50% of our scenarios4,000 simulated outcomes, each drawing the uncertain drivers from ranges we chose: our judgement, not a measured chance
- Worlds
- 3 stories about the businessEach outcome first picks one of these stories by its weight, then draws that story's own driver ranges; the answer above is their mix.
- World: base (50%)
- the base drivers. The price is justified in over 50% of its scenarios; middle outcome $327, 10th to 90th $176 to $632Weight: Our judgment, matching the central case's 50% weight.
- World: bear_lean (25%)
- The AI labs miss their plans in 2027; AWS capacity built for them sits partly idle. The price is justified in 20-50% of its scenarios; middle outcome $230, 10th to 90th $123 to $447Weight: Our judgment, matching the pessimistic case's 25% weight.
- World: bull_lean (25%)
- Enterprise inference arrives on top of the labs and the halls fill faster. The price is justified in over 50% of its scenarios; middle outcome $415, 10th to 90th $222 to $802Weight: Our judgment, matching the optimistic case's 25% weight.
- What moves the answer
- the exit price (market outcome) 75%; AWS revenue growth, points a year above or below the central path (33%, 27%, 19%, 15%, 12%, 10% in 2027-32) 13%; AWS net property per dollar of next year's AWS revenue, change every year from the central path (1.64x at end-2027 to 1.33x by 2032) 12%; AWS operating margin before depreciation, from 2027 10%; Share of the after-tax Anthropic and OpenAI stake value lost (negative: worth more than carried) 8%How much of the spread in value each driver explains on its own; drivers that move together share credit, so the shares need not add to 100%.
- Leaning cases
- 2 other views we ranThe same simulation with some ranges moved to another view, to show how much the answer depends on our ranges.
- Case: bear_lean
- The AI labs miss their plans in 2027; AWS capacity built for them sits partly idle. The price is justified in 20-50% of this case's scenarios; middle outcome $230, 10th to 90th $123 to $447
- Case: bull_lean
- Enterprise inference arrives on top of the labs and the halls fill faster. The price is justified in over 50% of this case's scenarios; middle outcome $415, 10th to 90th $222 to $802
- Cost of equity (original)
- 10.0%
- Original figures
- fair value $327 (bear $120, bull $558), exit $505; Buy
- Street
- average $334, median $325, low $300, high $390; 32 targets in the last quarter; Street consensus targets, Oct 26, 2026, 12:15 PM PDT
- Restated
- Oct 26, 2026, method case-spread-2026-10-24; forecast.json valued by method.py at market inputs 2026-10-01
- Cost of equity calibration
- +0.0 points
- Calibration source
- none: the 1-point cut set 2026-10-16 to pull the batch toward the Street was removed on 2026-10-18 (Dong); the cost of equity is risk-free plus beta times the premium
- View rule
- strong buy when the price is at or below 0.80x fair value; buy when at or below 0.90x fair value; avoid when fair value is below 0.85x the price; otherwise hold
- Blind step
- Unblinded; the model was committed in 40087e5 before the record
Original call, Oct 26, 2026 (before the Oct 26, 2026 restatement)
We keep Amazon a Buy at $248.47 and raise fair value to $327 from $319, 32% above the price. Nothing about the business or the forecast changed since this morning's fifth report. One input did: the long-run return on new capital after 2031. Our four hyperscaler reports set it four ways, and the owner approved one rule for all of them: each case earns, on growth after 2031, its own year-five return on its net capital, so growth needs the capital per revenue dollar the case itself reaches in 2031. For the central that is 21.2%, against 15% from Amazon's history (+$8.19). The view is a Buy under every capital choice we tested: the permanent floor gives $268, Microsoft's reviewer path scaled to Amazon $296, AWS depreciation 2 points faster $309. A 2027 capex figure above about $412B with no more AWS revenue would end the Buy.
As of 2026-10-26 · unblinded: a same-day revision of the fifth report under one long-run-return rule the owner approved for MSFT, GOOGL, AMZN and META (assumptions.md). Model committed before this report (40087e5). Revises the 2026-10-26 fifth report ($319, Buy), which stays on the record.
- Fair value today
- $327 per diluted share, at a 10.01% cost of equity. Weighting the pessimistic, central and optimistic cases 25/50/25 (our weights) gives $333
- Price
- $248.47, last trade 2026-10-26 16:16 UTC (official close 2026-10-23 $248.00) → 24% below fair value
- Five-year exit price
- $505 on 2031-12-31: the average of a long fade of returns toward the cost of capital ($392) and an outside view at the peers' 25.2 times operating profit ($619). Fading over 10 or 20 years gives $489 or $521
- Five-year return
- 16.5% a year from $248.47 to the central exit (no dividends)
- Bear / bull today
- $120 / $558: the pessimistic and optimistic cases, each one coherent world
- Confidence
- About 28%: the share of our simulated values within 20% of $327, from odds.json's percentiles (subjective, not calibrated)
- Moat
- Scale in fulfilment and in cloud; $496B of contracted AWS revenue; an ads business on the store's traffic. No proven moat yet in renting AI capacity
- Market expects
- On our forecast, about 14.3 times 2031 operating profit for the outside view, against 25.2 times for the peers
- Price needs
- Any one of: AWS growth 14.3 points a year below our path; AWS margin before depreciation 44.1% (ours 56%); AWS property per dollar of revenue 0.48x higher every year (2027 capex about $412B, ours $258B); an ads margin of 17.5% (ours 50%)
- Odds
- P(price justified): 67%, the share of our simulated values at or above the price. Subjective, not calibrated (odds.json)
- Key driver
- The exit multiple: each 5 turns of 2031 operating profit is about $35 a share. In the forecast, AWS property per dollar of revenue (0.10x every year is −$16) and AWS margin (3 points is about $20)
- Street
- Average 12-month target $333.68 (median $325, range $300-390) (market data, read 2026-10-26 for the fourth report). An expectations reference, not validation