AMZN · Forward model · Bull case
The Bull case, 20 quarters out
Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.
Amazon reports THREE operating segments - North America, International and AWS - and this model carries those three exactly as disclosed, plus two lines that have no reported revenue at all: Amazon Leo, the satellite broadband constellation, and Zoox, the robotaxi. Both are held at zero across every historical quarter and start in the projection, so nothing here is a re-cut of a number Amazon published. 2025 Q1 and Q2 are the segment figures as filed. The four quarters from 2025 Q3 to 2026 Q2 are the reported segment figures rounded to $0.1B in the source, with the rounding difference against the disclosed consolidated total ($180.169B, $213.390B, $181.519B, $200.606B) carried in North America, the largest line; no quarter is an apportionment, so none is flagged estimated. Advertising services - $19.8B in 2026 Q2, growing 26% - is NOT a vertical here: Amazon discloses it as a single global product line and does not split it between North America and International, so breaking it out would require inventing a geographic split the company has never published. It sits inside the two retail segments and is the main reason their margins are widening. What IS assumed throughout is the cost split: Amazon discloses segment operating income but not segment depreciation or segment capital expenditure, so every EBITDA margin and capex intensity on this page is ours. AWS carries the overwhelming majority of the AI infrastructure spend because that is where the money is going. Net cash is carried at zero: cash and marketable securities roughly offset funded debt at the basis quarter, and the lease obligations Amazon runs its fulfilment network on are not netted here at all.
The build is the moat. AWS holds its build rate instead of fading it, Trainium takes enough of the inference workload to hold revenue per megawatt flat rather than falling, and the margin keeps widening the way it did this quarter - 650 basis points in a year while absorbing the heaviest capex in the company's history. Retail margins keep pace as advertising compounds at mid-twenties growth on a base that costs almost nothing to serve, and Leo turns the constellation into a second subscription business with AWS-like economics once it is overhead.
Latest: $530.10B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $155.67B |
| 2025Q2 | $167.70B |
| 2025Q3 | $180.17B |
| 2025Q4 | $213.39B |
| 2026Q1 | $181.52B |
| 2026Q2 | $200.61B |
| 2026Q3E | $214.74B |
| 2026Q4E | $229.02B |
| 2027Q1E | $243.75B |
| 2027Q2E | $258.50B |
| 2027Q3E | $273.42B |
| 2027Q4E | $288.55B |
| 2028Q1E | $303.91B |
| 2028Q2E | $319.50B |
| 2028Q3E | $335.34B |
| 2028Q4E | $351.45B |
| 2029Q1E | $367.83B |
| 2029Q2E | $384.50B |
| 2029Q3E | $401.48B |
| 2029Q4E | $418.78B |
| 2030Q1E | $436.41B |
| 2030Q2E | $454.39B |
| 2030Q3E | $472.73B |
| 2030Q4E | $491.45B |
| 2031Q1E | $510.57B |
| 2031Q2E | $530.10B |
What drives each segment
North America
Growth pathOnline and physical stores, third-party seller services, subscriptions and the North American half of advertising. $116.2B in 2026 Q2, up 16%, with operating income up 21% to $9.1B - revenue growing fast for a business this size and profit growing faster, which is the mix shift toward advertising and seller services showing up in the margin line rather than in the growth line.
Latest: $247.06B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $92.89B |
| 2025Q2 | $100.07B |
| 2025Q3 | $106.27B |
| 2025Q4 | $127.09B |
| 2026Q1 | $104.12B |
| 2026Q2 | $116.21B |
| 2026Q3E | $121.89B |
| 2026Q4E | $127.63B |
| 2027Q1E | $133.43B |
| 2027Q2E | $139.28B |
| 2027Q3E | $145.20B |
| 2027Q4E | $151.19B |
| 2028Q1E | $157.27B |
| 2028Q2E | $163.43B |
| 2028Q3E | $169.69B |
| 2028Q4E | $176.05B |
| 2029Q1E | $182.51B |
| 2029Q2E | $189.10B |
| 2029Q3E | $195.82B |
| 2029Q4E | $202.67B |
| 2030Q1E | $209.66B |
| 2030Q2E | $216.80B |
| 2030Q3E | $224.11B |
| 2030Q4E | $231.58B |
| 2031Q1E | $239.23B |
| 2031Q2E | $247.06B |
Assumptions & reasoning
- Carried as a growth rate rather than as customers x spend because Amazon publishes neither: the Prime member count has not been updated since the 200 million figure of 2021, and it is a global number that cannot be assigned to a segment. A customers-times-spend model here would be two invented inputs multiplied together to reproduce a number Amazon already discloses.
- 3.6% a quarter is roughly the 16% year-over-year this segment just posted, and it decays to 1.75% - about 7% a year - by the end of the window. Holding mid-teens growth on a $450B-a-year revenue base for five years is the single easiest way to make this model say anything you like, so it is not held.
- The projection is smooth and Amazon is not: Q4 is seasonally about 20% above Q3 and Q1 falls back below it. Individual projected quarters are wrong in a way that cancels over a year, so read the annual totals rather than any one bar.
- The EBITDA margin of 12% is the disclosed 7.8% operating margin plus an assumed share of fulfilment and transportation depreciation. It glides to 15% because advertising and third-party seller fees grow faster than the retail revenue they attach to, which is the same mix shift that took operating income up 21% on 16% revenue growth.
- Capex intensity of 7% covers fulfilment centres, the transportation network and the robotics programme. It is a small share of the roughly $200B Amazon has guided for 2026 - almost all of that is AWS, and it is charged there.
International
Growth pathThe same retail, seller, advertising and subscription businesses everywhere outside North America. $42.2B in 2026 Q2, up 15%, with $1.7B of operating income - a 4% margin against North America's 7.8%. The bet is that the newer countries follow the same path the mature ones did, which is the reason this line's terminal margin is so much higher than its current one.
Latest: $85.81B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $33.51B |
| 2025Q2 | $36.76B |
| 2025Q3 | $40.90B |
| 2025Q4 | $50.70B |
| 2026Q1 | $39.80B |
| 2026Q2 | $42.20B |
| 2026Q3E | $44.18B |
| 2026Q4E | $46.17B |
| 2027Q1E | $48.17B |
| 2027Q2E | $50.18B |
| 2027Q3E | $52.20B |
| 2027Q4E | $54.24B |
| 2028Q1E | $56.29B |
| 2028Q2E | $58.37B |
| 2028Q3E | $60.47B |
| 2028Q4E | $62.59B |
| 2029Q1E | $64.75B |
| 2029Q2E | $66.93B |
| 2029Q3E | $69.15B |
| 2029Q4E | $71.40B |
| 2030Q1E | $73.69B |
| 2030Q2E | $76.02B |
| 2030Q3E | $78.40B |
| 2030Q4E | $80.82B |
| 2031Q1E | $83.29B |
| 2031Q2E | $85.81B |
Assumptions & reasoning
- A growth rate for the same reason as North America: no disclosed customer count, no disclosed unit volume, and a segment that spans Germany and Japan at one end and India and Brazil at the other. One invented average customer would be less informative than the growth rate itself.
- 3.4% a quarter against the 15% year-over-year just posted, decaying to 1.5%. International grows slightly faster than North America off a base a third the size and is exposed to currency in a way the domestic segment is not - a strong dollar takes two or three points off this line and nothing off the other.
- The margin assumption is the whole argument for this vertical. 7% EBITDA today gliding to 11% is the claim that the newer countries repeat what the established ones did once density arrived, and it is worth roughly a fifth of what North America's margin is worth. If India and Brazil never reach it, this line is a low-single-digit-margin business forever.
- Capex intensity of 6% is below North America's: much of the international network is newer but smaller, and the segment leans harder on third-party logistics in markets where Amazon has not built out.
AWS
Capacity × utilisation × priceCompute, storage, databases, model APIs and Amazon's own silicon. $42.2B in 2026 Q2, up 37% - the fastest in eighteen quarters - on a $169B annualised run rate, with operating margin up 650 basis points to 39.4%. It is 21% of Amazon's revenue and 60% of its operating income, and it is the only segment whose revenue is genuinely supply-gated: what AWS can sell this quarter is what it managed to energise.
Latest: $191.04B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $29.27B |
| 2025Q2 | $30.87B |
| 2025Q3 | $33.00B |
| 2025Q4 | $35.60B |
| 2026Q1 | $37.60B |
| 2026Q2 | $42.20B |
| 2026Q3E | $48.67B |
| 2026Q4E | $55.22B |
| 2027Q1E | $61.86B |
| 2027Q2E | $68.60B |
| 2027Q3E | $75.43B |
| 2027Q4E | $82.36B |
| 2028Q1E | $89.39B |
| 2028Q2E | $96.52B |
| 2028Q3E | $103.76B |
| 2028Q4E | $111.11B |
| 2029Q1E | $118.57B |
| 2029Q2E | $126.14B |
| 2029Q3E | $133.83B |
| 2029Q4E | $141.64B |
| 2030Q1E | $149.56B |
| 2030Q2E | $157.61B |
| 2030Q3E | $165.78B |
| 2030Q4E | $174.07B |
| 2031Q1E | $182.49B |
| 2031Q2E | $191.04B |
Assumptions & reasoning
- Modelled as megawatts energised x utilisation x revenue per megawatt rather than as a growth rate, because management has said repeatedly that demand exceeds what is switched on. When supply is the constraint, a growth rate is the answer rather than the question: it hides the build behind the number the build produces.
- 7,000 MW energised is derived from Amazon's own disclosure that it added more than 3.8 GW of capacity in a trailing year and intends to double total capacity by 2027. It is not a figure Amazon publishes as a stock, and revenue per megawatt follows from it: $42.2B over 7,000 MW at 88% utilised is about $6.85M a megawatt a quarter, or $27M a year.
- The build rate is set to fade 2% a quarter rather than compound. Holding 950 MW a quarter flat for five years puts AWS above 26 GW and roughly $500B of annual revenue by 2031, which is two thirds of what the whole company earns today - the model would be reporting the build rate as a result rather than testing it. The fade is the honest version, and it is the input most worth dragging.
- Revenue per megawatt drifts down 2% a quarter toward -1%. Compute prices fall, and inference in particular falls fast; what offsets it is mix, as managed AI services and Trainium capacity sell for more per watt than raw EC2 did. Holding price flat would add roughly a fifth to the terminal revenue on no evidence at all.
- The EBITDA margin of 65% is the disclosed 39.4% operating margin plus an assumed share of depreciation, which for a segment absorbing most of a $200B capital programme is very large and getting larger. The margin and the capex intensity are two views of the same money, so moving one without the other double-counts the build.
- Capex intensity of 85% of segment revenue is what roughly $200B of guided 2026 spending looks like once North America's fulfilment share is removed. It glides to 35%, which is the assumption the entire fair value rests on: it is the claim that this build has an end, and that free cash flow returns once it does.
Amazon Leo
Subscribers × ARPUThe satellite broadband constellation formerly called Project Kuiper: an FCC licence for 3,236 satellites, more than $10B committed, consumer terminals and an enterprise and aviation book sold against Starlink. Amazon has never disclosed a dollar of revenue from it, so it earns nothing here until 2027 and starts from zero in every historical quarter.
Latest: $5.40B (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $0.00 |
| 2025Q2 | $0.00 |
| 2025Q3 | $0.00 |
| 2025Q4 | $0.00 |
| 2026Q1 | $0.00 |
| 2026Q2 | $0.00 |
| 2026Q3E | $0.00 |
| 2026Q4E | $0.00 |
| 2027Q1E | $298M |
| 2027Q2E | $409M |
| 2027Q3E | $541M |
| 2027Q4E | $694M |
| 2028Q1E | $869M |
| 2028Q2E | $1.06B |
| 2028Q3E | $1.28B |
| 2028Q4E | $1.52B |
| 2029Q1E | $1.79B |
| 2029Q2E | $2.08B |
| 2029Q3E | $2.39B |
| 2029Q4E | $2.74B |
| 2030Q1E | $3.10B |
| 2030Q2E | $3.50B |
| 2030Q3E | $3.93B |
| 2030Q4E | $4.38B |
| 2031Q1E | $4.87B |
| 2031Q2E | $5.40B |
Assumptions & reasoning
- This vertical has NO reported history. Whatever Leo bills today sits inside North America's 'Other' revenue and is too small to see; rather than apportion a number out of a segment Amazon reports in full, the line is held at zero across all six quarters and opens in 2027 Q1.
- Net-adds mode rather than attach mode, because the constraint is satellites and terminals rather than willingness to pay: what limits subscribers is how much capacity is overhead and how many dishes can be built, which is exactly what a net-add ceiling describes.
- 900,000 net adds a quarter at the ceiling is a Starlink-shaped ramp arriving several years later against an incumbent that already holds the customers. It reaches roughly 11 million subscribers by 2031 - a real business at about $10B a year on the exit run rate, and still under 1% of Amazon.
- $70 a month is priced below Starlink's US residential tier, because Leo arrives second and the markets where it has the strongest claim - India, Africa, Latin America - will not pay a US price. The $150M of non-subscriber revenue is the enterprise, government and aviation book, which is where the announced contracts have actually been.
- The constellation itself is not charged to this vertical's capex intensity: it is a fixed-size programme on its own schedule, so it sits in corporate programmes at $20B over sixteen quarters and is spent whether or not a single subscriber signs up. The 120% intensity here covers ground stations, terminals and replenishment only.
- The margin starts at -180% and glides to 45%. A constellation is a fixed-cost asset with a marginal cost near zero once it is overhead, which is why the terminal margin is closer to AWS's than to retail's - and why a case where the subscribers never arrive is far worse than the revenue line alone suggests.
Zoox
Capacity × utilisation × priceThe purpose-built robotaxi Amazon has been running in Las Vegas and San Francisco, with no steering wheel and no disclosed revenue. It is the smallest line in the model and the one with the widest range of outcomes: a fleet business whose revenue is set by how many vehicles are on the road and how many hours a day each of them earns.
Latest: $777M (2031Q2E)
| Period | Value |
|---|---|
| 2025Q1 | $0.00 |
| 2025Q2 | $0.00 |
| 2025Q3 | $0.00 |
| 2025Q4 | $0.00 |
| 2026Q1 | $0.00 |
| 2026Q2 | $0.00 |
| 2026Q3E | $0.00 |
| 2026Q4E | $0.00 |
| 2027Q1E | $0.00 |
| 2027Q2E | $40M |
| 2027Q3E | $55M |
| 2027Q4E | $72M |
| 2028Q1E | $93M |
| 2028Q2E | $116M |
| 2028Q3E | $143M |
| 2028Q4E | $173M |
| 2029Q1E | $207M |
| 2029Q2E | $246M |
| 2029Q3E | $290M |
| 2029Q4E | $339M |
| 2030Q1E | $393M |
| 2030Q2E | $454M |
| 2030Q3E | $523M |
| 2030Q4E | $599M |
| 2031Q1E | $683M |
| 2031Q2E | $777M |
Assumptions & reasoning
- No reported history and no carved-out base: Amazon has never disclosed Zoox revenue, ridership or fleet size, and the early rides have been free. The line is held at zero across every historical quarter and takes three quarters to open, so the first revenue counted is 2027 Q2.
- Modelled as vehicles x utilisation x revenue per vehicle rather than as a growth rate, because a robotaxi fleet has no revenue to grow from - the only question worth asking is how fast vehicles reach the road and how hard each one works once there.
- $40,000 per vehicle per quarter is roughly $440 of gross bookings a day, which is a busy vehicle in a dense city and a generous one anywhere else. Utilisation of 40% gliding to 65% is the share of the fleet actually earning: depot time, charging, cleaning and weather take the rest.
- The fleet reaches roughly 15,000 vehicles by 2031 in the base case, which is a city-by-city rollout rather than a national one, and produces well under $2B a year. Zoox is an option on Amazon, not a driver of it, and at this scale it moves the fair value by less than a dollar a share.
- The margin starts at -220% and glides to 30%: vehicles, depots, remote operations and safety staff are all real costs against almost no revenue at the start, and the terminal margin is deliberately below AWS's because a fleet wears out and has to be replaced.
Where each case comes from
Jassy case — primary sources
The primary sources this case is built from — filings, calls and posts. Where they are claims by an interested party rather than disclosures, every number in the Jassy column is what happens if they are taken at face value.
What Amazon says the spending is for
From cash flow to fair value
| Present value of free cash flow, 20 quarters | $692.23B |
| Terminal-year revenue | $2.00T |
| Terminal-year EBITDA | $761.82B |
| Exit multiple, on revenue | 4.5x |
| Terminal value | $9.02T |
| Discounted at 8.5% a year, terminal value becomes | $6.00T |
| Enterprise value | $6.69T |
| Net cash | $0 |
| Equity value | $6.69T |
| Diluted shares | 10.89B |
| Fair value per share | $614.69 |
| Against the current price of $260.11 | +136% |
3.5x the terminal year's revenue against roughly 3.7x trailing today. The mix is better by then - AWS is a larger share of a bigger company - but the growth rate is not, so paying more than today's multiple for a slower Amazon needs an argument this model does not make. It is still the largest single input in the file.
Read the other way round: at $260.11 the market is paying 1.6x terminal-year revenue, holding every other assumption on this page fixed. That is the number to argue about.
Capex outside the verticals
Each vertical's capex is a share of its own revenue, which is the right shape for capacity that scales with sales. These are not that: fixed-size programmes with their own schedule that spend whether or not any segment grows. They are added on top, and they are why free cash flow dips in the years below.
Leo constellation build
2026 Q3 → 2030 Q2The 3,236-satellite constellation and its launch contracts. Amazon has said the programme is worth more than $10B; $20B over four years is what a full constellation plus early replenishment costs, and it is spent on the launch schedule rather than on subscriber growth, which is why it is not inside Leo's capex intensity.
The projected path
| Quarter | North America | International | AWS | Amazon Leo | Zoox | Revenue | YoY | EBITDA | Capex | FCF | R40 | PV of FCF |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 2026 Q3E | $121.89B | $44.18B | $48.67B | $0 | $0 | $214.74B | +19% | $58.46B | $50.71B | $6.51B | +22 | $6.38B |
| 2026 Q4E | $127.63B | $46.17B | $55.22B | $0 | $0 | $229.02B | +7% | $64.99B | $53.33B | $9.79B | +12 | $9.40B |
| 2027 Q1E | $133.43B | $48.17B | $61.86B | $298M | $0 | $243.75B | +34% | $71.38B | $55.99B | $12.92B | +40 | $12.16B |
| 2027 Q2E | $139.28B | $50.18B | $68.60B | $409M | $40M | $258.50B | +29% | $78.06B | $58.32B | $16.58B | +35 | $15.28B |
| 2027 Q3E | $145.20B | $52.20B | $75.43B | $541M | $55M | $273.42B | +27% | $84.88B | $60.52B | $20.47B | +35 | $18.49B |
| 2027 Q4E | $151.19B | $54.24B | $82.36B | $694M | $72M | $288.55B | +26% | $91.83B | $62.64B | $24.52B | +34 | $21.70B |
| 2028 Q1E | $157.27B | $56.29B | $89.39B | $869M | $93M | $303.91B | +25% | $98.91B | $64.73B | $28.71B | +34 | $24.89B |
| 2028 Q2E | $163.43B | $58.37B | $96.52B | $1.06B | $116M | $319.50B | +24% | $106.12B | $66.83B | $33.01B | +34 | $28.04B |
| 2028 Q3E | $169.69B | $60.47B | $103.76B | $1.28B | $143M | $335.34B | +23% | $113.46B | $68.94B | $37.40B | +34 | $31.12B |
| 2028 Q4E | $176.05B | $62.59B | $111.11B | $1.52B | $173M | $351.45B | +22% | $120.94B | $71.10B | $41.86B | +34 | $34.14B |
| 2029 Q1E | $182.51B | $64.75B | $118.57B | $1.79B | $207M | $367.83B | +21% | $128.55B | $73.32B | $46.40B | +34 | $37.07B |
| 2029 Q2E | $189.10B | $66.93B | $126.14B | $2.08B | $246M | $384.50B | +20% | $136.31B | $75.61B | $50.99B | +34 | $39.92B |
| 2029 Q3E | $195.82B | $69.15B | $133.83B | $2.39B | $290M | $401.48B | +20% | $144.21B | $77.99B | $55.63B | +34 | $42.67B |
| 2029 Q4E | $202.67B | $71.40B | $141.64B | $2.74B | $339M | $418.78B | +19% | $152.25B | $80.45B | $60.31B | +34 | $45.33B |
| 2030 Q1E | $209.66B | $73.69B | $149.56B | $3.10B | $393M | $436.41B | +19% | $160.45B | $83.01B | $65.04B | +34 | $47.90B |
| 2030 Q2E | $216.80B | $76.02B | $157.61B | $3.50B | $454M | $454.39B | +18% | $168.79B | $85.68B | $69.81B | +34 | $50.38B |
| 2030 Q3E | $224.11B | $78.40B | $165.78B | $3.93B | $523M | $472.73B | +18% | $177.30B | $87.20B | $75.68B | +34 | $53.50B |
| 2030 Q4E | $231.58B | $80.82B | $174.07B | $4.38B | $599M | $491.45B | +17% | $185.96B | $90.09B | $80.53B | +34 | $55.79B |
| 2031 Q1E | $239.23B | $83.29B | $182.49B | $4.87B | $683M | $510.57B | +17% | $194.78B | $93.09B | $85.43B | +34 | $57.98B |
| 2031 Q2E | $247.06B | $85.81B | $191.04B | $5.40B | $777M | $530.10B | +17% | $203.78B | $96.20B | $90.37B | +34 | $60.10B |
Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.
Model revisions
Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.
| Date | Changed | Fair value then | Note |
|---|---|---|---|
| 2026-08-20 | all | $335.46 | Initial model, built off the 2026 Q2 print: $200.606B of net sales, $27.5B of operating income, North America $116.2B, International $42.2B, AWS $42.2B growing 37% at a 39.4% operating margin, and roughly $200B of guided 2026 capital spending against trailing free cash flow of -$7.6B. Five verticals: the three reported segments plus Amazon Leo and Zoox, neither of which has any disclosed revenue, opening in 2027 Q1 and 2027 Q2. AWS is modelled as 7,000 MW energised at $6.85M per megawatt a quarter, derived from the disclosed 3.8 GW trailing-year build. |