The AI tax on your next phone: how memory costs are reshaping smartphone pricing

There is a number that explains almost everything happening to phone prices right now, and it comes from IDC’s senior research director for consumer devices, Nabila Popal, describing the second quarter of 2026:

Memory costs are up nearly 300% year over year, and now account for over 65% of the bill of materials at the low end.Nabila Popal, senior research director for consumer devices, IDC

Sit with that. On a budget smartphone, two-thirds of the manufacturing cost is now the memory chips alone — before the processor, the screen, the camera, the battery, the casing, the assembly, the shipping, the marketing, or anyone’s profit. A category of device that survived on thin margins and high volume has had its cost structure inverted by a component that used to be an afterthought. This is not a shortage in the conventional sense. Nobody bombed a factory. The chips are being manufactured in enormous quantities. They are simply going somewhere else.

The mechanism: why AI ate the memory

The smartphone in your pocket and the servers running AI models draw from the same finite pool of wafer capacity, controlled by three companies: Samsung, SK Hynix, and Micron.

Those companies face a straightforward commercial choice. High-bandwidth memory (HBM) for AI accelerators commands vastly better margins than the LPDDR chips that go into phones. Cloud providers are signing multi-quarter and multi-year purchasing agreements, locking in volume years ahead. Phone makers buy on shorter cycles, at lower prices, for a market that is shrinking. Given that choice, the manufacturers have reallocated production lines toward HBM, server DRAM, and enterprise SSDs. Consumer memory gets whatever capacity is left. Counterpoint’s Tarun Pathak described the resulting dynamic without diplomatic softening: memory companies are asking smartphone vendors to stand in line behind the hyperscalers, with allocation to phone makers deprioritized in favour of AI.