Apple just posted its best demand in years and the stock fell about 6%.
That is not a contradiction. It is the whole point. Demand has stopped being Apple's problem. iPhone and Mac accelerated. Greater China grew more than 20%. Every region grew at a double-digit rate. But Apple cannot make enough product to fill those orders, and memory prices are climbing into its margins. Investors looked past a strong June quarter and priced the September one.
The beat borrowed more than it earned
Apple earned $2.02 a share in its fiscal third quarter, up from $1.57. Revenue rose 16.4% to $109.42 billion. Both cleared expectations.
Now take out the tariff refund. Apple said refunds added about $0.11 a share. That leaves roughly $1.91 against a $1.89 consensus. The beat shrinks to about two cents. Apple reports on a GAAP basis and offers no adjusted figure. But the ex-refund number is the better read on recurring earnings.
Gross margin tells the same story. Apple reported 50.1%. Refunds contributed about two points of that. Underlying margin was near 48.1%, close to the roughly 47.9% analysts expected. It was also down from about 49.3% in the March quarter.
Consensus was not even one number. Providers put revenue near $108.65 billion or about $109.04 billion. The gap is small, but it changes how large the beat looks. Any single consensus figure deserves caution here.
Two smaller items pulled in opposite directions. The tax rate rose to 17.9% from 16.4%, which hurt. Share count fell about 2% on buybacks, which helped. The release did not address the tax move.
Apple sold everything it could build
The demand evidence was not soft. iPhone revenue rose 21.7% to $54.25 billion. Mac jumped 28.7% to $10.35 billion, far above the roughly $8.74 billion expected. Greater China rose 22.4% to $18.82 billion.
That China figure came in below one reported estimate near $19.6 billion. A 22% gain is still not weakness. China has moved from drag to contributor.
The strength was broad. Every region grew at a double-digit rate and set a June-quarter revenue record. Active devices passed 2.5 billion.
Only one category fell. iPad revenue dropped 5.9% to $6.19 billion, against a hard comparison with last year's A16 launch. Its installed base still hit a high.
Then came the guide. Apple expects September revenue up 9% to 11%. The midpoint implies about $112.7 billion. Consensus sat near $114.9 billion. Even the top of Apple's range falls short.
The gap is not mostly about demand. Management named two causes. Currency will cost about 2.5 points of growth. Supply limits will worsen significantly, affecting iPhone, Mac and iPad. Apple still expects iPhone revenue to grow in the mid-teens.
Tim Cook said the cause was Apple's own forecast, not a supplier failure. Demand simply ran ahead of plan. Apple had already pulled supply forward, and that flexibility is nearly used up.
Investors did not fully accept the framing. Unfilled demand is not revenue. Some buyers wait. Others take an older model, or leave. Apple also declined to guide the December quarter, which makes it harder to tell a one-quarter squeeze from something longer.
Two problems, not one supply story
These constraints get merged constantly. They work differently.
The first is capacity at leading-edge chip nodes. Apple cannot sell a device it cannot build. That caps revenue.
The second is memory pricing. DRAM and NAND costs are rising fast, and management called the move an extreme shock. That raises the cost of every unit produced. It caps gross profit instead.
The remedies diverge. Node scarcity needs capacity and better forecasting. Memory inflation needs procurement, mix, redesign or higher prices. Cook noted the DRAM market has just three suppliers. More would help both availability and cost.
The margin path shows the pressure. Ex-refund gross margin ran near 49.3% in March, then 48.1% in June. September guidance implies about 46.5% once the expected one-point refund benefit is stripped out. That is roughly 280 basis points of underlying decline across two quarters.
The finance chief said memory costs explained more than all of the June step-down. Carry-in inventory, cheaper non-memory parts and mix softened the blow.
Where the margin sat matters too. Products margin rose to about 40.1%, helped by more than 2.5 points of tariff benefit. Services margin fell about 110 basis points to 75.6% on mix. So the reported strength was concentrated in hardware, and it was on loan.
Apple has already raised Mac and iPad prices. Cook called that reluctant and said it is early to judge the demand effect. That leaves the elasticity question open. Push prices and risk volume. Absorb costs and lose margin.
One balance sheet line deserves attention. Inventories nearly doubled from the fiscal year end, to $11.09 billion. Management separately credited carry-in inventory for cushioning memory costs. Apple has not connected the two. The 10-Q should clarify.
Spending is tilting as well. Research costs rose 32.3%, roughly twice the pace of revenue. Nine-month capital spending fell 28.2%. Apple runs a hybrid model using its own data centers and outside cloud capacity, so reported capex does not capture all AI spending.
Services still grows, just not on autopilot
Services revenue rose 12.1% to $30.74 billion. That trailed reported estimates near $31.2 billion, though one outlet set the bar at $31.4 billion.
Management named several drags. Currency was the main sequential factor. There was no film release to match last year's. App Store gaming softened. Business model changes in some countries and a US court ruling on link-out purchases also weighed.
Some of that passes. Some does not. App Store rules shape take rate and payment control. Apple's appeal on the link-out ruling is going to the Supreme Court.
Currency gets worse from here, with roughly 2.5 more points of Services headwind expected in September.
The base still compounds. Paid subscriptions passed 1.5 billion, a deep pool to monetize even when growth cools.
What Apple has to prove now
The quarter swapped one risk for another. Demand risk faded. Execution and margin risk took its place.
Apple's new Siri arrives this fall. It will not launch in the EU at first for iPhone and iPad, and China cleared only the earlier feature set. Apple disclosed no AI revenue. So the jump in research spending is investment, not return.
This was also Cook's final earnings call. John Ternus, the hardware engineering chief, takes over. He said little publicly. The handover arrives with node supply, memory costs, pricing and App Store rules all unsettled.
The burden of proof has shifted. Apple no longer has to show that people want its products. It has to show three other things. That it can build enough at leading-edge nodes. That it can carry or pass along memory costs without losing volume. And that Services can hold its pace through regulation and currency.
Until then, the June quarter stands as proof of a stronger franchise. It is not yet proof of better earnings.