Two of the largest companies in the world reported within a day of each other and the market sent them in opposite directions. Apple fell about 6% and kept falling, touching $300 on Friday after closing at $333.43 the session before. Amazon rose more than 15%. The convenient story is that one company is winning the AI trade and the other is losing it. The more useful observation is that both CFOs stood up and named the same line item, and it was memory.
Andy Jassy raised Amazon’s 2026 cash capex guidance from roughly $200 billion to roughly $220 billion, and he said out loud on the call that higher memory cost is what pushed the number up. Twenty billion dollars of incremental spend, attributed to an input price. On the Apple side, Kevan Parekh walked through the gross margin bridge and said that more than 100% of the sequential decline in June-quarter gross margin, from 49.3% in March to 48.1% excluding the tariff refund, is explained by memory cost. Same input, two companies, one raising a capex number and the other lowering a margin number.
That symmetry is the story, and most of the coverage this week missed it because it was busy sorting megacaps into winners and losers.
It is worth being precise about what actually hit Apple’s guide, because the press has flattened it. The September-quarter revenue guidance of 9% to 11% growth, against a 12% consensus, was not primarily a memory problem. Parekh attributed the deceleration to about 2.5 points of FX headwind plus a sequential increase in supply constraints, and Cook was explicit that the binding constraint is availability of advanced nodes for Apple’s own SOCs. He went further and said it is not a supply problem at all in the usual sense. It is a demand forecast issue. iPhone and Mac are selling far better than Apple planned for, and there is very little flexibility left in the chain to fix that inside a quarter. iPhone revenue of $54.25 billion was up 22%, a June-quarter record. Mac was up 29% to $10.35 billion, the best June quarter the Mac has ever had. Total revenue of $109.42 billion, up 16%, EPS of $2.02, up 29%. That is a company that undershot its own forecast of how much people would want to buy.
Memory is the margin story, and the margin story is the one that persists. Cook confirmed Apple has paid rising memory costs for three consecutive quarters and expects to pay more in the fourth. He called it a 100-year flood on memory pricing. The September gross margin guide of 47% to 48% works out to about 46.5% at the midpoint once the expected one point of tariff benefit is stripped out, down from 48.1% adjusted in June, and Parekh warned that the carry-in inventory cushioning some of it will shrink over time.
Here is what makes Apple structurally different from every other company in this cycle, and it is the thing Wall Street has been arguing about for a year in the wrong direction. Apple’s fiscal 2026 capex is somewhere around $13 to $14 billion. The four largest hyperscalers have guided to a combined $700 billion plus. Apple rents compute and partners for it rather than building it. For most of this year that discipline was treated as a feature, and it is a large part of why capital parked in Apple as a safe haven and pushed it briefly above $5 trillion on July 28, past Nvidia, at an all-time closing high of $340.08. What the June quarter showed is the cost of that position. Apple absorbs the input inflation that the buildout created and books nothing on the other side of the ledger to offset it. Amazon pays more for the same memory and sells the capacity it goes into. AWS grew 37%, its fastest in years, with segment operating margin at 39%, up 650 basis points year over year, and the AI and custom-chip businesses each crossed $25 billion in annualized run rate. Google Cloud grew 82% and Azure grew 43% in the same window. Amazon is capacity constrained through at least 2027 and said so.
So the question for Apple is not whether the moat holds. It does. The installed base, the switching costs, the services attach on top of hardware, the ability to raise prices without losing the customer, all of that is intact and the June quarter demonstrated it. The question is narrower and more testable: how much of the memory bill can Apple hand to the consumer, and how fast. It already ran the experiment once. On June 25 it raised MacBook and iPad prices more than expected, and the stock had its worst session in over a year before recovering to records. The iPhone 18 Pro is the real test. TechInsights has floated something in the neighborhood of a $270 increase to hold the current margin structure; JPMorgan thinks $50 is closer to the ceiling. The gap between those two numbers is the entire bull-bear case, and nobody will know the answer until September.
On the stock, the sell-side response was a shrug rather than a downgrade. Goldman had raised to $370 days before the print and pulled back to $360. Morgan Stanley trimmed from $364 to $360 and kept Overweight. TD Cowen’s Krish Sankar has been at $350 since WWDC. Baird sits at $330. Consensus is roughly $330, which means the average target was already below the stock the week before earnings and is now above it. Base case holds Apple in the $310 to $340 band into the September launch, with the fall Siri release and the pricing decision as the two catalysts. Bull case gets back to $370 to $380 if Apple prices the iPhone 18 Pro aggressively and the upgrade cycle absorbs it without unit damage. Bear case is $270 to $290, and the mechanism there is not company-specific. It is the safe-haven megacap cohort derating together once the market accepts that avoiding AI capex does not mean avoiding AI costs, with the John Ternus transition on September 1 removing the one thing that always steadied that cohort.
The number to hold onto is Jassy’s twenty billion. Apple lost margin to memory and Amazon added capex for memory, and both dollars ended up in the same place, on the income statements of three memory suppliers who now have two of the largest buyers on earth publicly confirming they cannot get relief. That is not an Apple story or an Amazon story. It is a Micron and SK Hynix story that two other companies reported for them.