A chief executive who did not claim pricing power
Cristiano Amon spent most of Qualcomm's earnings call on 29 July talking about the parts of his company that are growing, and then said something plain about the part that is not. Asked about the price increase Qualcomm had sent to its customers, the chief executive did not describe a product that had earned the right to cost more. "We're just passing through big cost increases that we have," he said. Vendors do not usually volunteer that. A price rise a company chooses to take is a sign of strength. A price rise it is handing on is a sign that someone further up the chain took the margin first.
The mechanics are simple and they carry a date. Every Snapdragon chip shipped after 1 September 2026 carries a double-digit percentage increase. Bloomberg reported the customer notice on 24 July and Qualcomm confirmed the increase five days later alongside its results. In that notice the company said the semiconductor industry is experiencing a broad-based increase in input costs, across wafer fabrication, assembly, test, advanced packaging, memory and other materials, and that it has exhausted its ability to absorb higher costs from suppliers. Qualcomm has not published the exact percentage, and nobody should quote one as though it were confirmed.
The decline arrived before the increase did
The important date is not 1 September. It is the quarter that ended before it. Qualcomm's handset chip revenue for the third fiscal quarter was 5.086 billion dollars, against 6.328 billion a year earlier. That is a fall of just under 20 per cent, and all of it happened at the old prices. Whatever the increase does to demand from here, it cannot be blamed for a decline that was already booked before it took effect.
That ordering matters to anyone about to negotiate. A supplier raising prices into rising demand is rationing scarce output, and the buyer has very little leverage. A supplier raising prices while its own volume falls by a fifth is doing something different: recovering a cost it cannot avoid, on a book of business that is shrinking underneath it. The second case is the one Qualcomm described, and it is the more awkward of the two, because the company needs its remaining customers to absorb an increase at the precise moment there are fewer of them.
One honest limit belongs here. Qualcomm does not disclose handset unit volumes, so how much of that 20 per cent is fewer chips and how much is a cheaper mix is not public. The direction is not in doubt. The composition is.
Ninety-nine point seven per cent of one company's profit
The cost Qualcomm says it cannot absorb is visible as profit one layer up the stack. Samsung reported its own second quarter within a day of Qualcomm. Consolidated revenue reached 171.5 trillion won and operating profit 89.5 trillion won, both all-time highs, and the memory business set records for quarterly revenue and operating profit. The Device Solutions division, which is where memory sits, produced 89.2 trillion won of operating profit on its own.
Set those two figures side by side and the picture is stark. Device Solutions accounted for 99.7 per cent of Samsung's entire operating profit for the quarter. Everything else the company does, from phones and displays to televisions and appliances, netted 0.3 trillion won between them. Memory is not one contributor among several this year. It is very nearly the whole company.
This is what makes the pass-through credible rather than opportunistic. When a supplier blames an input cost, the useful test is whether that cost turns up as somebody else's margin, and here it does, at a scale that is hard to argue with. It also tells a buyer where the money is going and therefore how long it is likely to keep going there. Memory contracts for the coming cycle are largely signed already. This is not a negotiating posture that evaporates if you push back on it.
Automotive covered three fifths of the gap
Qualcomm's answer to a shrinking handset business is to sell into other ones, and the quarter shows real progress. Automotive revenue reached 1.588 billion dollars, up more than 61 per cent, and the internet-of-things line reached 1.830 billion, up nearly 9 per cent. Together those two grew about 28 per cent, which is the figure the company chose to highlight.
Growth rates flatter, though, and the dollars are the better guide. Automotive and IoT together added 753 million dollars year on year. Handsets gave up 1.242 billion. The two growing businesses replaced a little under 61 per cent of what the shrinking one lost, which is why total revenue still fell about 4 per cent to 9.9 billion dollars despite two strong segments. Amon told investors he expects growth in non-handset revenue to accelerate from 24 per cent this fiscal year to more than 60 per cent next.
The other half of the problem is a single customer. Amon said Qualcomm's share of the coming iPhone will be well below the 20 per cent the company had previously estimated, and that Apple product revenue should fall roughly 50 per cent from the September quarter to the December quarter. He attributed the faster decline to the same supply constraints, which cut the component share Qualcomm won in that launch. A supplier losing ground at its largest handset customer while handset revenue falls by a fifth has every reason to protect margin wherever it still can.
What to settle before the first of September
The increase applies to chips shipped after 1 September, not to devices sold after it, and that gap is where the practical decisions sit. Handsets reaching European shelves this autumn were largely built from silicon bought at the old price. The devices that carry the new cost are the ones being specified now for launch in 2027, which means a fleet refresh signed this quarter and one signed next spring are priced off different inputs even if the model names barely change.
Three things are worth doing while the date is still ahead. Ask every device supplier to state in writing whether a quoted price is fixed to a shipment date or open to component pass-through, because that single clause is where a double-digit chip increase reaches your budget. Pull forward the part of a 2027 refresh that genuinely can be pulled forward, and only that part, since buying ahead of a known cost rise is sensible and buying stock you do not need is not. And treat the memory line as its own negotiation, because it is the input Qualcomm named and the one now showing up as record profit at Samsung.
What does not work is waiting for the increase to be withdrawn. A supplier with falling volume does not discount its way back to growth. It holds price and lets a better mix carry it, which is exactly what the move into automotive and IoT is for.
Read next: Google Stopped Absorbing Your Memory Price | Samsung Just Split the Fold Into Two Shapes
