Qualcomm reported fiscal third-quarter revenue of $9.947 billion on Wednesday, down 4% from $10.365 billion a year ago and comfortably above the roughly $9.67 billion analysts expected. Adjusted earnings came in at $2.21 a share against a $2.23 estimate. GAAP diluted EPS was $1.87, down 23%.
Then the guidance landed. Fourth-quarter revenue of $9.7 billion to $10.5 billion — roughly in line. Fourth-quarter adjusted EPS of $2.05 to $2.25 — against a consensus of $2.36. The top of the range sits below what the street had penciled in for the middle. Shares fell about 5% after hours.
The culprit is not demand for Qualcomm’s chips. It is the price of the parts that sit next to them. Handset chip revenue fell 20% to $5.09 billion, and CEO Cristiano Amon pinned the gross-margin compression on the cost of memory — the same DRAM and high-bandwidth memory that AI server builders are buying faster than three companies can make it. Amon called the margin hit “a temporary, short-term thing.” Qualcomm’s answer is a broad price increase across its chip lineup starting September 1.
Our take: This is the first major P&L where the AI buildout shows up as a cost line instead of a revenue line. Everything else this earnings season has been AI winners arguing about the size of their windfall. Qualcomm is the other side of the trade: a company with no AI datacenter business getting taxed by one. Memory is a commodity with three real suppliers and one insatiable customer class, and when that customer class outbids you, the shortage travels down the supply chain until it reaches a phone buyer. The September 1 price increase is Qualcomm passing the bill along. Watch how many other component-dependent businesses do the same over the next two quarters — because that is what AI capex looks like when it arrives as inflation.
The business underneath the headline
Strip out handsets and Qualcomm’s diversification story is working. Combined automotive and IoT revenue grew 28% year over year, and automotive posted its 23rd consecutive quarter of double-digit annual growth. That is a decade-long bet finally carrying weight in the mix.
The problem is scale. Handsets are still the volume engine, and a 20% drop there swamps a 28% gain in the segments meant to replace it. Management said the quarter represents the bottom for handset revenue. Reuters also reported the company expects its Apple revenue decline to accelerate — a separate, structural erosion that has nothing to do with memory and everything to do with Apple building its own modems.
So there are two clocks running. One is cyclical: memory prices normalize, margins recover, the price increase sticks or it does not. The other is permanent: the largest customer leaves. Wednesday’s guidance only speaks to the first one.
The memory squeeze has been visible for weeks. It already reached consumer phone prices, and on the supply side SK hynix printed a 76% operating margin selling into the same shortage. Meanwhile the hyperscalers funding all of it got very different verdicts on their capex the same afternoon.
What to watch
- Whether September 1 sticks. A blanket price increase into a slowing handset market is a test of pricing power, not a solution. If phone makers absorb it, Qualcomm’s margin math works. If they delay orders instead, the “bottom” call moves out a quarter.
- Memory contract pricing. Qualcomm’s fourth-quarter margin assumption is a bet on where DRAM settles. The next round of memory contract negotiations tells you whether this guidance was conservative or optimistic.
- The Apple line. Guidance for an accelerating decline is the number that matters for 2027 models. Auto and IoT growing 28% does not backfill it fast enough on the current trajectory.
- Who else blames memory. Every consumer electronics and PC maker reporting over the next month faces the same input cost. Qualcomm just handed them the script.
A revenue beat, an earnings miss by two cents, and a guidance cut caused by a shortage in a product Qualcomm does not sell. The AI trade has started billing the companies that are not in it.