AMZN — Amazon
AWS is worth one thing. The stores are worth another. The capex bill belongs to both.
Amazon is really two huge operating engines under one ticker: AWS, and the retail/logistics/advertising machine. Then there is the gigantic reinvestment bill sitting across both. I stopped pretending one blended P/E tells me much and value the pieces separately now.
A blended Amazon P/E has always bothered me because it takes a bunch of completely different businesses, mushes them into one earnings number and then acts like one multiple explains the whole thing.
Why I care about this one
I own Amazon because there are several excellent businesses hiding inside the same ticker.
AWS is the obvious one. Advertising is increasingly ridiculous. The retail/logistics network keeps improving its economics as delivery gets faster and inventory gets placed closer to customers. Prime ties the whole thing together.
Historically, management has been unusually good at taking periods that look financially ugly and turning them into much larger businesses later.
That earns a lot of trust from me.
It still does not earn one blended multiple and a prayer.
Q2 sales reached $200.6B, up 20%, while operating income rose 43% to $27.5B. (Q2 results)
AWS grew 37% to $42.2B and produced $16.6B of operating income in the quarter. On a trailing basis, AWS generated about $54.7B of operating income.
North America and International together produced roughly $39.0B of trailing operating income. That bucket also contains the advertising economics Amazon reports inside the geographic segments, which is why I do not value ads separately and accidentally count them twice.
Advertising itself grew 26% to $19.8B in Q2.
The businesses are working. Unfortunately, so is the infrastructure bill.
Q2 cash capex was $53.1B. Trailing operating cash flow was $161.4B, but free cash flow was negative $7.6B. Cash and marketable securities were about $123B at June 30, against roughly $132B of long-term debt face value. (Q2 10-Q)
So no, this is not a cash-flow-light story right now.
Q2 2026 / TTM
Primary sources: Q2 2026 results ↗ · Q2 2026 10-Q ↗
I killed the old synthetic $8.54 EPS × P/E model. Amazon now uses a two-engine SOTP starting from about $54.7B of TTM AWS operating income and $39.0B from North America + International, with separate growth and EV/EBIT assumptions plus net cash/debt and diluted shares. Bear / Base / Bull are roughly $198 / $487 / $677. Buy bands now come from 12% / 15% / 18% required returns to the base SOTP. Capital Allocation moves from 8.25 to 8.0/10 because the infrastructure build also came with a much larger debt load.
My position
What I actually bought and sold.
| Date | Type | Account | Dollars | Shares | Price | Return now* |
|---|---|---|---|---|---|---|
| Mar 16, 2026 | Buy | Taxable | $1,300 | 6.21 | $209 | +19.2% |
*Return now uses the latest available quote for buys and reinvestments. Sales and fees show “—”.
Valuation & scenario math
I start with the trailing operating-income anchors:
- AWS: roughly $54.7B
- Stores + advertising: roughly $39.0B
Then each scenario gives those engines different growth rates and different terminal EV/EBIT multiples.
In the base case, AWS operating income compounds 20% annually and gets 28x at the endpoint. Stores + ads compounds 12% and gets 22x. After the balance-sheet and share-count bridge, I land around $487 per share. The live page owns the return from today’s quote to that endpoint.
The exact bridge is below. The part I like is that the model now tells me where I can be wrong. Maybe AWS deserves less. Maybe retail margins mature lower. Maybe the capex never gives me the balance-sheet recovery I am assuming.
That is much more useful than arguing about one Amazon P/E.
12% / 15% / 18% required return to the base SOTP
Starter zone · 9.5% below $276 starter
| Band | Buy price | Base return to Sep 2031 |
|---|---|---|
| Starter | $276 | +12.1% |
| Normal | $242 | +15.1% |
| Aggressive | $213 | +18.2% |
The dollar bands come from the active base-case SOTP and the five-year return hurdle. There is no magic Amazon P/E I am trying to wait for.
Two-engine SOTP using trailing operating income. AWS gets its own growth and EV/EBIT assumptions. Stores/Ads gets its own. Then I bridge through net cash/debt and diluted shares. A single consolidated P/E is easier. It is also hiding the exact thing I care about.
| AWS TTM operating income | $54.7B |
| AWS annual growth | 20% |
| AWS operating income by Sep 25, 2031 | ~$136.1B |
| AWS terminal EV/EBIT | 28x |
| Stores + Ads TTM operating income | $39.0B |
| Stores + Ads annual growth | 12% |
| Stores + Ads operating income by Sep 25, 2031 | ~$68.8B |
| Stores + Ads terminal EV/EBIT | 22x |
| Enterprise value | ~$5.32T |
| Endpoint net cash | $30B |
| Diluted shares | 11.0B |
| Implied base price | ~$487/share |
AWS and Stores/Ads earn their own growth rates and multiples. Then I bridge enterprise value through the balance sheet and share count. No blended Amazon P/E hiding the work.
| Case | What has to be true | Target range | Annualized return from today |
|---|---|---|---|
| Bear | AWS EBIT 10% growth · 18x EV/EBIT; Stores/Ads EBIT 4% growth · 14x. AWS remains good but capital intensity stays high, retail/ads operating income barely compounds, Amazon remains net-debt and dilution continues. | $184–$212 center ~$198 | -4.6% |
| Base | AWS EBIT 20% growth · 28x EV/EBIT; Stores/Ads EBIT 12% growth · 22x. AWS compounds strongly, Stores/Ads keep expanding operating income and the current infrastructure wave eventually restores modest net cash. | $453–$521 center ~$487 | +14.4% |
| Bull | AWS EBIT 25% growth · 32x EV/EBIT; Stores/Ads EBIT 15% growth · 25x. AI demand keeps AWS compounding at an exceptional rate, retail/ads deliver major operating leverage and capex converts into a much stronger balance sheet. | $629–$724 center ~$677 | +22.3% |
Returns use the current quote and the time remaining to Sep 25, 2031. The scenario targets stay fixed until the underwriting changes.
Capital allocation
The reinvestment history is elite. The current bill is still enormous.
AWS, Prime, advertising and logistics are about as good an argument for productive reinvestment as you can get. I still cut the balance-sheet score because long-term debt rose sharply while cash capex exploded, and direct capital return/share-count shrinkage remain pretty limited.
Stage-relative, not some objective rating-agency grade. The number is there to force the judgment into the open. See the rubric →
What management is actually doing with the cash
Amazon has earned a ridiculous amount of reinvestment trust. AWS, Prime and the logistics network are why. That still does not mean every new infrastructure dollar gets a blank check. Q2 cash capex was $53.1B, long-term debt face value was about $133B and trailing free cash flow was negative. Great history buys management patience. The current build still has to earn a return.
What I’m doing with the next dollar
The position still has plenty of room inside the 3%–5% target.
At 0.5%, I have room to buy. I am in no rush to use it.
The buy framework works backward from the base SOTP instead of pretending there is one correct Amazon P/E.
Roughly:
- starter = a 12% five-year base-case return
- normal = 15%
- aggressive = 18%
That currently works out around $276 / $242 / $213.
Whether Amazon is in the starter, normal or aggressive zone should come from the live buy framework above, not an old close frozen into this paragraph.
That feels about right to me. Interesting enough that the tiny position can become less tiny. Not cheap enough that I suddenly stop comparing the next dollar with Microsoft, Google, the index, or anything else I already own.
What can break this
AWS can remain a great business and the stock can still disappoint.
If AI demand stays strong but serving it requires permanently worse capital intensity, AWS deserves a lower multiple than a software-like growth rate would normally suggest.
The stores can disappoint in a different way. Faster delivery, inventory placement, automation and ads should keep improving retail economics. If operating income stalls while the network keeps getting more expensive, that piece of the SOTP gets a lot less valuable.
And I am still ignoring Q2’s $53.4B pre-tax Anthropic gain for operating valuation purposes.
Nice mark.
Still not an operating segment, lol.
The model gets better if AWS keeps compounding around 20% while the infrastructure spend eventually turns back into meaningful free cash flow and the stores keep expanding operating income.
It gets worse if capex stays enormous because enormous capex is simply what the business requires now.
If the capex stays enormous because the business simply requires it, my base case is too optimistic.
Amazon has one of the best reinvestment records in the market, and I am happy to give management credit for that.
I am just not letting ‘they have done it before’ do the work of a valuation model.
The evidence I’m watching
AWS operating income compounds around or above 20%, Stores/Ads keep expanding margins, and the infrastructure wave starts throwing off enough owner cash that the balance sheet clearly moves back toward net cash.
AWS turns out to need structurally worse capital intensity, Stores/Ads operating leverage stalls, capex stays near current levels without a matching earnings step-up, or debt/share issuance becomes a normal part of funding the build.