Notebook

Amazon

AMZN Open Platforms

  • +9.2%unrealized, USD
  • 238.24average cost, as supplied
  • 260.11last · 2026-08-20

The record

Updated
2026-08-19.

Thesis

Written 2026-08-21.

Amazon owns the two hardest things to rebuild: the cloud the internet runs on, and the logistics network that moves physical goods. I intend to hold it for a long time, because neither of those is displaced by a better model or a cheaper competitor.

AWS growth, quarter by quarter

AWS revenue versus the same quarter a year earlier, per cent

0 10 20 30 40 17.5% Q2 2025 20.0% Q3 2025 24.0% Q4 2025 28.0% Q1 2026 36.7% Q2 2026 latest

Scroll the chart sideways →

What you're looking at: AWS revenue growth, each quarter against the same quarter a year before. Growth has risen in each of the last four quarters and has more than doubled across the five shown. The normal path for a business this large is a slow grind downward, because the base it is measured against keeps getting bigger - AWS turned over $42.2bn in the last quarter alone. Reaccelerating from that base is the single strongest fact I know about this company, and it is the reason the position exists. Amazon company-reported AWS segment revenue growth, Q2 2025 through Q2 2026. Q2 2025 was $30.9bn at 17.5%; Q2 2026 was $42.2bn at 36.7%, the fastest since 2021. I have not verified the quarter before this run began, so the run is described as four quarters of acceleration rather than five. Company-reported, not audited by me, not pre-registered. Source numbers: amzn-aws-growth.csv.
Show the numbers
Period ValueNote
Q2 202517.5%
Q3 202520.0%
Q4 202524.0%
Q1 202628.0%
Q2 202636.7%latest

That chart is the position. AWS grew 36.7% to $42.2bn last quarter, its fastest rate since 2021, and growth has risen in each of the last four quarters from 17.5%. Businesses this large are supposed to decelerate, because every year they are measured against a bigger base. This one is doing the opposite, which says demand for compute is outrunning what the market expected rather than fading. AWS is now at a $169bn annualised run rate, and within it the AI business and the in-house chips have each passed $25bn annualised, both more than doubling in a year.

The retail moat is physical, and physical moats age well. Fulfilment centres, delivery density, the installed base of Prime, decades of supplier relationships. Software eats a lot of things; it does not eat a warehouse network. That is the part of Amazon I am least worried about.

Where the profit comes from, and where the room is

segment operating margin, per cent, most recent quarter

AWS 39.4% $16.6bn of income North America 7.9% $9.1bn of income International 4.1% $1.7bn of income

Scroll the chart sideways →

What you're looking at: Operating margin by segment. AWS earns about five times the margin of the North American retail business and now supplies roughly 61% of all operating profit. I am not expecting retail to converge on cloud - they are different businesses and it never will. The point is what sits underneath the two short bars: retail turns over far more revenue than AWS does, so one point of margin there is a very large number, and this is the part of the company automation is aimed at. Amazon Q2 2026 reported segment results. AWS $16.6bn of operating income on $42.2bn of revenue; North America $9.1bn at 7.9%; International $1.7bn at 4.1%. The AWS margin includes derivative accounting gains - excluding them the year-on-year improvement is 520 basis points rather than 650. Company-reported, not audited by me, not pre-registered. Source numbers: amzn-margins.csv.
Show the numbers
Measure ValueNote
AWS39.4%$16.6bn of income
North America7.9%$9.1bn of income
International4.1%$1.7bn of income

The margin story is where the upside sits, and the arithmetic is forgiving. AWS already earns 39.4% and supplies around 61% of all operating profit. The two short bars are the opportunity: at Amazon's volume a single point of retail margin is an enormous number, so this case needs no heroics. North America has moved to 7.9%, and the company expects to more than double its fleet of robotic arms this year. I do not need a dramatic re-rating; I need this to keep grinding upward, which is what it has done for a decade.

The capital expenditure is the price of the next decade, and Amazon has always paid it. Guidance went to $220bn from $200bn. Over the last twelve months capital expenditure of $173bn has more than consumed $161bn of operating cash flow, turning free cash flow into an outflow of $7.6bn against an inflow of $18.2bn a year earlier. This is a company with a long record of spending its cash flow into capacity and being right - AWS itself was exactly that trade, and the chart at the top of this page is what it grew into. Part of the increase is higher memory prices, which is the same cycle I am on the other side of through Micron and SanDisk.

What I am watching

I am bullish, and these are the things I would need to see to change that.

Whether AWS growth holds. Four quarters of acceleration is the core of the thesis. A slide back toward the high teens while capital expenditure stays at this level would mean paying up for capacity into softening demand.

Whether free cash flow recovers. An outflow for a year is a build-out. An outflow through 2027 would mean the economics of the business have changed rather than its phase.

How quickly automation reaches the margin. Retail margins have improved for years and are still single digit. I expect gradual, because gradual is what the record shows - but if two more years pass with no movement, the savings are being competed away in price rather than kept.

The quality of reported profit. Q2 included a $53bn gain related to Anthropic, and the AWS margin itself includes derivative accounting gains - stripping those out, the year-on-year improvement is 520 basis points rather than 650. Both are real value and neither is operating performance, and I want to track the underlying line rather than the headline.

What would make me wrong

  • AWS decelerating for two consecutive quarters while capex guidance rises again.
  • Free cash flow still negative at the end of 2027.
  • North America retail margin flat or lower two years out.
  • A regulatory or labour-cost change that alters the economics of the fulfilment network.

Latest movements

  • 2026-07-30 — Opened at $238.24. Untouched since.