Apple is the one giant that is not making the trillion-dollar bet. It posted a record quarter while spending almost nothing on AI hardware. Its shares fell anyway, because of its guidance for the next quarter, not because of what it spent.
| Line | Expected | Delivered |
|---|---|---|
| Revenue | ≈$108.9B | $109.4B (up 16% on a year earlier) |
| Earnings per share | ≈$1.89 | $2.02 (up 29%; includes $0.11 from tariff refunds) |
| iPhone | — | up 22%, a June-quarter record |
| Services | ≈$31.2B | ≈$30.7B, a record, but below what analysts expected |
| Gross margin | — | 50.1% (about 2 points of it from tariff refunds) |
| Capital spending, quarter | — | $3.4B; Microsoft had reported $41B the day before |
| Own guidance, Sept quarter | — | weak, with supply constraints given as the reason |
Apple declared a dividend of $0.27 per share. This was Tim Cook's last results call as chief executive: he becomes executive chairman on 1 September, and John Ternus takes over.
The three clocks at the top of this page ask three questions: is AI revenue growing fast enough (demand), is the company still committed to the spending (capex), and are investors still willing to wait (patience). Apple is not one of the five biggest AI builders whose spending this site's model tracks: Alphabet, Amazon, Meta, Microsoft and Oracle. That is exactly why its results matter. Apple's AI plan needs little capital spending, the money spent on things that last for years: chips, buildings, power. It runs AI models on the customer's own device, and it partners with other firms for the largest models. It spends about $11 billion a year on capital, against the $840 billion the five have said they will spend in 2026. That makes Apple the closest thing this bet has to a control group. It faces the same customers and the same demand, and its balance sheet is as large as the builders', but it has chosen not to place the bet. If the bet's logic is right, and the winners must own the computing hardware themselves, Apple's results should in time fall behind the builders'. If Apple can keep posting record quarters without owning a single data centre, then owning the hardware is worth less than the spending implies.
The quarter beat analysts' forecasts, but look at what made up the numbers before giving Apple full credit. Earnings per share rose 29%. Of that rise, $0.11, about a third, came from tariff refunds, which also lifted the gross margin by roughly two points. The iPhone did the real work: sales up 22%, a June-quarter record. Services set a record too, but came in below analysts' forecasts, and Services is the part of Apple that carries its high profit margins. So this was a real beat, but a flattered one. The engine underneath is the cycle of new phone sales, not a stream of AI revenue. That fits the control-group reading; it is not evidence against it.
The market's reaction is the most telling single fact of this results season. Alphabet spent heavily and its shares fell. Meta spent heavily and its shares fell further. Amazon spent heavily, raised its own spending guidance by $20 billion, and its shares rose. That same evening Apple's shares fell 8% in after-hours trading on weak guidance for the September quarter, which Apple blamed on supply constraints. Across five sets of results the pattern is this: the market is not punishing capital spending, and it is not rewarding restraint. It is judging how sure each company is about demand. A company that shows evidence of demand is forgiven the bill. A company that shows doubt is punished, whether or not it is building. For the front page's patience clock, this changes the reading slightly: investors' patience is not running out with the spenders in particular. It is running out with uncertainty.
One quarter proves little either way. Apple's decision to sit out only becomes evidence once the gap builds up over time. If the builders' cloud businesses keep growing at 35% to 90% a year while Apple grows at the pace of phone sales, then owning the hardware pays, and the control group is quietly losing. Nothing in these results forces that conclusion yet. The plain reading tonight is that the bet's arithmetic does not apply to Apple. Not the revenue floor, the revenue the AI spending must earn every year to pay for itself (the front page's §2). Not the wall of depreciation charges as the hardware wears out. Not the debt that must be borrowed again when it falls due. No AI capital spending, so no wall and no debt to roll over. Apple has no clocks running. What it has instead is the same question in another form: is AI you rent worth as much as AI you own?
Amazon reported the same evening; full analysis here. It answered the demand question and raised its spending guidance at the same time. SpaceX followed on 4 August, and it is Apple's opposite: a sixth builder spending $2.35 of capital for every dollar of revenue, where Apple spends three cents. The July US jobs report, the other side of the ledger and the front page's test of whether AI revenue and jobs are moving apart, lands on 7 August. Nvidia, whose sales show how much the builders are spending, reports on 26 August. Earlier this results season: Alphabet, Microsoft, Meta.