The quarter's last open question about demand got its answer: AWS grew at its fastest since 2021, with $496 billion of orders signed but not yet delivered. A $20 billion rise in the spending plan arrived in the same breath. Investors rewarded it.
| Line | Expected | Delivered |
|---|---|---|
| Revenue | own guidance $194–199B | $200.6B (+20% on a year earlier) |
| Operating profit | own guidance $20–24B | $27.5B |
| AWS growth on a year earlier | ≈31% (analysts); 28% last quarter | 37%; $169B a year at current rate +37% vs +40% required: closing in, still below |
| AWS orders signed, not yet delivered | — | $496B under contract, not yet delivered |
| Net profit / earnings per share | ≈$1.82 per share (analysts) | $62.6B / $5.75; incl. $53.4B gain outside operations, mostly its Anthropic stake |
| 2026 capital spending | $200B own guidance | raised to $220B |
| Q3 revenue, own guidance | — | $197–202B |
Capital spending (capex) is money spent on things that last for years: chips, buildings, power. The higher spending plan counts against the capex clock, not against the quarter's profits: more spending is a bigger bet, not a worse quarter. Actual figures are from Amazon's filing with the US securities regulator (the SEC 8-K), captured on the night of the results by the data pipeline this site reads from.
The test the front page set for Amazon was blunt: growth at AWS, Amazon's cloud business, had to hold or speed up, because a slide back towards 20-something percent would undercut the whole story about demand. It sped up: 37% on a year earlier, the fastest since 2021, against the roughly 31% analysts expected, and at the current quarterly rate $169 billion a year. By the front page's arithmetic that is still below the roughly 40% growth a year, compounded, that the revenue floor requires. The revenue floor is the revenue the AI spending must earn every year to pay for itself (the front page's §2). But AWS is the second-largest cloud business reporting this quarter, and it is closing in on that line from below at a very large size. And the order book changes the question. $496 billion of work under contract but not yet delivered means much of the growth of the next few years is already sold. The open question shifts from will demand appear to how fast can signed orders be turned into revenue. That is a question of capacity, which is exactly how Amazon justified the higher spending.
Amazon's own guidance for 2026 capital spending went from $200 billion to $220 billion. That is the third raise of the quarter, after Alphabet's and Meta's, leaving Microsoft alone in holding its number. The five biggest AI builders named by the Bank for International Settlements (the BIS, a bank owned by the world's central banks) have now announced roughly $840 billion of spending for 2026 between them: Alphabet, Amazon, Meta, Microsoft and Oracle. Note the pattern this quarter has settled into: every piece of evidence of demand arrives attached to a bigger bill. That is the front page's argument behaving exactly as its model says. The revenue floor is not a fixed target the sector moves towards. It moves up with every quarter of renewed commitment. The front page's gap chart, which compares revenue earned with revenue required, has had its required line raised to match.
The quarter's most quotable number, $62.6 billion of net profit, is mostly not profit from running the business. $53.4 billion of it is a gain from outside operations, mainly the rise in the recorded value of Amazon's stake in Anthropic. That is an accounting gain, not cash. Hold the whole loop in view: Amazon invests in an AI lab, sells that lab the computing power it trains its models on, and books the rise in the lab's valuation as its own record profit. Every step is legal and disclosed. But it is the clearest example yet, in a single set of accounts, of what the BIS calls circular financing. That is firms in the boom investing in, buying from and selling to each other, so that one firm's spending is another's revenue. It is why this site's gap chart counts revenue, never profit, against the floor.
Two numbers to watch from here. First, how fast signed orders turn into revenue. If the order book keeps growing while AWS growth flattens, orders are piling up faster than Amazon can build the capacity to serve them. That is good for the builders, but it pushes the proof further into the future while the chips already bought keep wearing out. Second, the growth rate itself. 37% must become 40% or more, and stay there for years, not quarters, for the sums on the front page to add up. Results like tonight's every quarter until 2028 would win the bet. Results like tonight's, followed by a fade, are exactly the shape of the best year of every previous investment mania.
Apple reported the same evening (full analysis): the control group, the big tech firm with no AI build-out of its own, and punished by investors all the same. SpaceX followed on 4 August: a sixth builder outside the BIS's five, whose AI customers include Anthropic, the same lab whose rising valuation produced the paper gain on this page. The July US jobs report lands on 7 August. It is the other side of the ledger, and the test the front page sets: are results beating forecasts while the jobs figures stay flat? Nvidia, which sells the chips and so sees the spending from the supplier's side, reports on 26 August. Earlier this quarter: Alphabet, Microsoft, Meta.