The memory crunch
Every wafer that goes to an HBM stack is a wafer denied to a phone. Units fall, revenue holds, and the incremental consumer dollar moves to the memory supplier — if the forecasts are right.
The sold-out HBM book has a downstream invoice. Incremental DRAM supply is being prioritized for high-capacity server modules. A wafer is a rival good. Send it to a stack for an Nvidia GPU and it does not become the LPDDR5X in a mid-range phone or the SSD in a laptop.
every wafer allocated to an HBM stack for an Nvidia GPU is a wafer denied to the LPDDR5X module of a mid-range smartphone
The allocation line from the consumer-shortage research
Memory is now more than 30 percent of the bill of materials in some devices, according to CCS Insight’s Ben Hatton. Some entry-level smartphones have already seen prices rise more than 50 percent. Between the third quarter and mid-fourth, memory prices jumped 30–50 percent. A DDR5 16-gigabyte module ran from $40 in July to $105 in October to $180 in November; spot reached $200–220, and HP was reportedly buying at $200. Dell hiked commercial PCs 10–30 percent from December 17, 2025. C.C. Wei, looking at TSMC’s own mix, said consumer and price-sensitive end markets are being challenged by rising component prices.
IDC’s 2026 forecast is the tell, if it lands: PC units down 11.3 percent, PC revenue up 1.6 percent to $274 billion; smartphone units down 12.9 percent, revenue down 0.5 percent. Units fall. Revenue holds. The consumer’s incremental dollar transfers to the memory supplier. The OEM sells fewer boxes at higher prices for the same money and a worse mix.
The incidence is regressive. Memory is 15–20 percent of mid-range BOM against about 10–15 percent for flagships. Apple and Samsung are filed as structurally hedged — cash and long-term supply agreements that let them buy 12–24 months ahead. Thin-margin Android and PC names, the list that includes HP, Dell, Lenovo, Xiaomi, Transsion, have to pass the cost through or cut the spec. Lenovo’s Winston Cheng has said a new memory fab takes two to three years. His company, in one report, stockpiled RAM 50 percent above usual, enough to last through 2026. A hoard defers the hit. It does not cancel it.
Who eats the increase
Contract structure, not the calendar, decides the residual claimant. TrendForce has US cloud buyers on multi-year agreements that restrict suppliers from raising prices for those clients. From the third quarter of 2026, the primary source of server DRAM increases shifts toward customers without those agreements. A capped price makes pre-buying pointless. A phone OEM does not have that cap.
An unverified report that SK Hynix scrapped the caps would blur the seam. The July 29 call did not confirm it. Official language is about ten long-term agreements, deposits, and pricing that is “not uniform” — designed to address volatility. It did not say take-or-pay. Variable LTA pricing is already a softer seam than TrendForce’s “restrict suppliers from raising prices.” It is not the same as caps gone. The wiki records both and does not flatten them.
The shock may already be rolling over. Third-quarter DRAM contract increases are decelerating to 13–18 percent quarter-on-quarter from about 58–63 percent in the second quarter and about 90–95 percent in the first — because consumer makers are hitting an affordability limit. Demand destruction capping the price is the same squeeze working in reverse. It is also a reason the OEM margin hit may be nearer its end than its beginning.
The used-phone funnel
CCS Insight has the secondary market up 15.4 percent in 2026 against a primary smartphone market down 14.8 percent. First-quarter mobile trade-in value was $1.63 billion, up 31 percent year on year. The second quarter printed $1.43 billion returned to US consumers. Assurant said it serviced more than 7 million devices, about 1.8 million more than a year earlier, on new reverse-logistics programs. Those numbers are value and units, not Assurant revenue. Biju Nair names rising smartphone prices as the reason trade-ins matter more. No disclosure sizes how Assurant’s profit and loss scales with trade-in volume. The argument names a public company. It does not size it.
Omdia reported the smartphone market down only about 4 percent year on year in the second quarter of 2026, against IDC’s 12.9 percent full-year cut and CCS Insight’s 14.8 percent. The back half has to be much worse for those forecasts to land. Direction has two independent forecasters. Magnitude is still a projection.