Record Profit And Negative Free Cash Flow, Same Quarter

Handle the profit line carefully. Net income of $62.6 billion against $18.2 billion looks like a step change in the business. It is not. Non-operating pre-tax other income of $53.4 billion, primarily from the Anthropic stake, made up about two thirds of pre-tax income. That is paper appreciation on an equity position in a private company, not profit earned by selling goods, advertising or cloud capacity.

The correct year over year comparison is operating income of $27.5 billion against $19.2 billion, not $62.6 billion against $18.2 billion. On that basis the quarter was still excellent, with margin expanding from 11.4% to 13.7%.

The mark can reverse. Amazon reported a net loss in 2022 with positive operating income because the Rivian stake moved the other way. Any margin assumption drawn from reported net income will be wrong in both directions over time.

The cash story runs opposite to the profit story. Trailing free cash flow is an outflow of $7.6 billion, against an $18.2 billion inflow a year ago, because property and equipment purchases rose $66.1 billion year over year. Capex for 2026 has been raised to roughly $220 billion, with memory costs cited as a driver. Amazon spent $96.3 billion in the first half alone.

This is the tension the next three years resolve. Operating cash flow of $161.4 billion is enormous and growing 33%. It is currently insufficient to cover the build. The bet is that AWS revenue arrives fast enough to outrun the depreciation that all this capex will start pushing through the income statement.

AWS is doing its part so far. Growth of 37% is the fastest since 2021, up from 28% the prior quarter, and it beat expectations by six points. AWS produces about 60% of Amazon's operating income from roughly a fifth of revenue at a 39% margin. The $496 billion contracted backlog is the strongest evidence that the capex is demand-led rather than speculative.

Advertising at 26% growth on $19.8 billion is the quieter part of the story and carries margins closer to AWS than to retail.

Why 12% revenue growth. From $716.9 billion in 2025, 12% compounds to roughly $1.26 trillion by 2031. That matches the 12% Amazon delivered in 2025 and sits below the 19.6% just posted, on the view that AWS acceleration is partly offset by maturity in North America retail. The Q3 guide of 9% to 12% is depressed by the Prime Day calendar shift and should not be read as the trend.

Why a 12% margin. This is a normalised net margin, excluding investment revaluations, so it pairs with a normalised multiple. The underlying figure this quarter was about 10.6%. Expansion to 12% assumes AWS and advertising keep gaining share of the mix, partly offset by the depreciation wave from $220 billion of annual capex. That offset is real and is the reason this is not set higher.

Why 30x. Amazon's historical multiples reflected structurally depressed earnings, so they are a poor guide. Thirty times normalised earnings for a business with AWS at 39% margins and advertising compounding in the mid twenties is a premium to the market and a discount to where Amazon has often traded.

Why 9%. Mega cap, diversified, $123 billion of cash against $132 billion of notes. Nothing here justifies a high discount rate.

What today's price requires

At a market capitalisation near $2.95 trillion and a 9% discount, justifying today's price needs roughly 12% revenue growth, a 12% normalised net margin and a 30x exit multiple together. That is the base case above, which puts Amazon close to fairly valued rather than obviously mispriced in either direction.

The honest reading is that the disagreement is not about whether the assumptions are reachable. It is about whether $220 billion a year of capex produces the revenue to support them before the depreciation lands.

What kills this thesis

Free cash flow still negative on a trailing basis at the end of 2027. AWS growth decelerating below 25%. Operating margin falling below 11% as depreciation from the capex cycle flows through. Capex guidance rising materially above $220 billion without a matching acceleration in AWS backlog. Advertising growth dropping below 15%. AWS backlog growth turning negative.

Note that a large negative swing in the Anthropic mark would hit reported earnings hard without falsifying anything above. That is the point of excluding it.

Checkpoints

Q3 2026, late October: operating income against the $22.5 billion to $26.5 billion guide, AWS growth ex the Prime Day distortion, trailing free cash flow direction, and whether capex guidance moves again. Q4 and full year, early February 2027: full year free cash flow, capex against the $220 billion figure, and initial 2027 capex guidance. Through 2027: depreciation as a share of revenue, and whether AWS margin holds near 39% as the new capacity comes online.

A note on share count. This valuation divides 2031 earnings across roughly 10.55 billion shares against about 10.9 billion today, so it assumes stock compensation issuance is fully offset by repurchase. That is close to neutral and moves the per share figure by only a few percent either way, which is unusual. Unlike valuations that lean on aggressive buybacks or absorb heavy dilution, almost all of this number comes from the operating assumptions rather than the capital structure.

What this valuation leaves out. Because the margin here is normalised to exclude investment revaluations, the equity stakes themselves are not in the figure at all. Amazon holds roughly $28.7 billion in OpenAI plus an Anthropic position carried somewhere between $92.5 billion and $190 billion depending on how the convertible notes and preferred shares aggregate. At 10.9 billion shares that is somewhere between $11 and $20 per share of value sitting outside this model. Treat it as optionality rather than adding it mechanically, since the same accounting that produced a $53.4 billion gain this quarter produced a large loss on Rivian in 2022. But be aware the operating valuation understates the whole by that amount.