This morning we published two articles that set two numerical tests for Amazon, hours before the release landed. Both passed.
And then, in one line further down the release, Amazon explained why this report's headline is misleading.
What Was Reported
| Metric | Q2 2026 | Year ago | Change |
|---|---|---|---|
| Revenue | $200.6 billion | $167.7 billion | +20% |
| AWS | $42.2 billion | $30.9 billion | +37% |
| AWS operating income | $16.6 billion | $10.2 billion | +63% |
| North America | $116.2 billion | - | +16% |
| International | $42.2 billion | - | +15% |
| Total operating income | $27.5 billion | $19.2 billion | +43% |
| Net income | $62.6 billion | $18.2 billion | - |
| Diluted EPS | $5.75 | $1.68 | - |
And advertising grew 26% - the highest-margin revenue line outside AWS.
The Headline, and the Line That Explains It
Consensus stood at $1.82. The result: $5.75. More than three times.
But Amazon states explicitly in its own release:
"Second quarter 2026 net income includes non-operating pre-tax other income of $53.4 billion, primarily from our investments in Anthropic."
Which is why this number has to be decomposed rather than read from the headline
| Line in the accounts | Amount |
|---|---|
| Operating income | $27.5 billion |
| Other income, net (primarily Anthropic) | $53.4 billion |
| Income before taxes | $80.9 billion |
| Tax (22.5% effective rate) | $18.2 billion |
| Net income | $62.6 billion = $5.75 |
A revaluation of a stake in one private company is 1.95 times Amazon's entire operating income.
And excluding it - at the 22.5% effective tax rate the company itself recorded, and the diluted share count it reported - profit from operations is about $1.95 per share.
Meaning: a beat of roughly 13 cents against consensus, not of almost four dollars. Of the $5.75, about $3.80 is a markup.
(That calculation is our approximation from the published figures - removing other income from pre-tax profit, applying the reported effective tax rate, and dividing by the diluted share count.)
This does not mean the gain is not real. The Anthropic investment is worth what it is worth, and the increase in its value is a gain in every accounting sense. But it is not cash, it does not repeat, and it works with exactly the same force in the other direction.
And for a sense of the rate of change: last quarter that same line was $16.8 billion. This quarter - $53.4 billion. Three times, in one quarter.
The Two Tests We Set This Morning - and What Happened to Them
Test one: the cloud war, in dollars
In the capex-versus-depreciation article and the thread alongside it, we argued that growth percentages on different bases are a meaningless comparison, and computed dollar additions instead: Google Cloud added about $11.2 billion in the June quarter.
And we set a numerical threshold: to match it, AWS needed to reach about $42 billion - roughly 36% growth from a $30.9 billion base.
| The threshold we set | The result | |
|---|---|---|
| AWS revenue | ~$42.0 billion | $42.2 billion |
| Growth rate | ~36% | 36.6% |
| Dollar addition | $11.2 billion | $11.3 billion |
AWS cleared it - by about $130 million. For the first time in a while, the world's largest cloud added more dollars than Google in the same quarter.
Test two: depreciation
In the same article we showed AWS was the only one of the three clouds growing slower than Amazon's own depreciation - a spread of minus 4.8 points in the March quarter. And we set the depreciation growth rate, 27.2%, as the bar.
| March quarter | June quarter | |
|---|---|---|
| AWS growth | 28.0% | 36.6% |
| Depreciation growth (twelve months) | 27.2% | 28.4% |
| The spread | -4.8 points | +8.2 points |
The reversal happened. Cloud revenue now compounds meaningfully faster than the depreciation schedule already locked in.
But the Broader Thesis Actually Strengthened
And this is the part not to miss inside the celebration.
| Metric | March quarter | June quarter |
|---|---|---|
| Capex, quarter | $44.2 billion | $54.2 billion |
| Capex, twelve months | $151.0 billion | $173.0 billion |
| Depreciation, twelve months | $70.4 billion | $75.2 billion |
| Capex-to-depreciation ratio | 2.14 | 2.30 |
| Annual gap | $80.6 billion | $97.8 billion |
| Free cash flow, twelve months | +$1.2 billion | -$11.6 billion |
Free cash flow turned negative
Cash from operations over twelve months: $161.4 billion. Purchases of property and equipment over the same period: $173.0 billion.
The difference: minus $11.6 billion.
Amazon is spending more on infrastructure than its operations generate. A year ago free cash flow was $18.2 billion; a quarter ago, $1.2 billion; now it is negative.
And the gap between capex and depreciation - cash that has left the bank and has not yet reached the income statement - widened by $17 billion in a single quarter, to $97.8 billion a year.
Meaning: the test we set was answered positively, but the larger question got sharper. Amazon is earning faster - and spending faster still.
Guidance
| Metric | Q3 2026 | Year ago |
|---|---|---|
| Revenue | $197.0-202.0 billion | +9%-12% |
| Operating income | $22.5-26.5 billion | $17.4 billion |
And note the company's own caveat: excluding the impact of Prime Day in both years, third-quarter growth would be nearly 400 basis points higher - roughly 13%-16%. The guidance also anticipates about 80 basis points of foreign-exchange headwind.
What Jassy Said - and the Part That Touches the Whole Week
CEO Andy Jassy cited record delivery speeds for Prime members, with over 40% more items delivered same-day or overnight in the first half.
But the remark that connects directly to the chain we covered all week is a different one: Jassy points to multi-year, multi-gigawatt commitments from Anthropic and OpenAI, and to growing adoption of Trainium chips by start-ups.
Gigawatts. That is a unit for a power station, not a data center - and it is exactly what Quanta and the utilities are dealing with from the other side.
The Bull Thesis
Whoever reads it positively will point first to AWS: 37% growth with operating income jumping 63% to $16.6 billion. That is not just revenue growth - it is a margin expanding alongside it.
Beyond that: the breadth. North America +16%, International +15%, advertising +26%, and total operating income +43%. And next quarter's guide implies operating income of up to $26.5 billion against $17.4 billion a year ago.
And the positioning: gigawatt commitments from Anthropic and OpenAI are contracts, not intentions.
The Bear Thesis
Whoever reads it critically will note first that nearly two-thirds of reported profit is not from the business. $53.4 billion from revaluing a private holding, against $27.5 billion of operating income. Anyone reading the headline alone gets a completely wrong picture.
Second, free cash flow is negative - minus $11.6 billion over twelve months, from $18.2 billion positive a year ago.
Third, the gap between capex and depreciation widened to $97.8 billion a year. That is a future cost already paid in cash and not yet recognized.
And fourth, guided growth for next quarter is 9%-12% - below this quarter's 20%, even if the explanation (Prime Day) is legitimate.
The debate in one line
The bulls see AWS growing 37% with operating income +63%, advertising +26%, total operating income +43%, and gigawatt commitments from the largest AI customers. The bears see reported profit of which nearly two-thirds is a revaluation of an untraded holding, free cash flow that turned negative, and a $97.8 billion annual gap between cash out and cost recognized. Both sides are reading the same report.
My Angle
A personal opinion of Ilan Abramov - not advice, not a recommendation
What catches me in this report is that two parts of it tell opposite stories, and only one will make the headlines.
The first part is AWS, and it is genuinely excellent. 37% growth with operating income up 63% is a hard combination to find: revenue grew, and the margin grew with it. That says the capacity being built is filling, and that pricing is holding.
And in the article we published this morning we set two numerical bars for this before the figure landed - 36% for dollar-add parity with Google, and 27.2% against the depreciation rate. Both were answered: 36.6%, and 36.6% against 28.4%. I note that not as a victory but because it is what a good framework is supposed to do - produce a number you can check, rather than an assessment nobody can refute.
And the second part is the Anthropic line, and it troubles me more than anything else in the report.
$53.4 billion. 1.95 times Amazon's entire operating income. Last quarter it was $16.8 billion. One private company, with no market-set price, produced a gain in Amazon's accounts double that of its entire operating business.
I am not arguing the value is not real. I am saying this is a line with no cash behind it, set in a funding round rather than on an exchange, and it will work with exactly the same force when the direction reverses. Anyone buying the stock on "earnings of $5.75" is buying something other than what they think.
And what I hold as the important point: free cash flow turned negative. A year ago it was $18.2 billion positive, today minus $11.6 billion. Amazon is spending more on infrastructure than its business generates - and the difference is funded from the balance sheet.
And what I will watch: the capex-to-depreciation ratio. It rose from 2.14 to 2.30. As long as it climbs, the bill is still waiting and has not begun to be paid. AWS crossing above its depreciation rate is genuinely good news - but it is a race against a schedule that keeps getting longer.
Summary
Amazon closed the cloud triangle with AWS growing 37% to $42.2 billion and operating income jumping 63% to $16.6 billion. Total revenue $200.6 billion (+20%), advertising +26%, and total operating income +43%.
And both tests we set this morning passed: AWS added $11.3 billion against Google's $11.2 billion, and grew 8.2 points above its own depreciation rate - after minus 4.8 in the prior quarter.
But the headline misleads: of the $5.75 per share, about $3.80 is a revaluation of the Anthropic investment - $53.4 billion, 1.95 times operating income. Excluding it, EPS is about $1.95, a 13-cent beat.
And in parallel: capex jumped to $173.0 billion over twelve months, the capex-to-depreciation ratio rose to 2.30, and free cash flow turned negative - minus $11.6 billion.
The question for the investor is not whether AWS works - it works, and the numbers are unambiguous. The question is how much of reported profit is business and how much is markup, and when the $97.8 billion bill starts coming due.
Sources: Amazon.com's official results release for the second quarter of 2026 (July 30, 2026), as filed with the U.S. Securities and Exchange Commission on Form 8-K, including revenue and the segment breakdown, AWS revenue and its operating income, total operating income, the non-operating income and its attribution to Anthropic, tax, net and diluted earnings, purchases of property and equipment, depreciation, cash from operations and third-quarter guidance; CEO Andy Jassy's remarks from the release; prior-quarter figures from the results release of April 29, 2026 and year-ago figures from the release of July 31, 2025; Google Cloud figures from Alphabet's second-quarter 2026 release. The calculation of earnings excluding non-operating income is our approximation based on the published figures. Data accurate as of the time of writing. The chart is shown in real time via TradingView. Nothing herein constitutes a forecast, recommendation or advice - see the full disclaimer at the bottom of the page.