AI boom gives memory chipmakers greater pricing leverage as Apple, Nvidia and Qualcomm compete for supplies - Firstpost

The pricing power shift in memory is fundamentally an AI-driven structural crowding-out of consumer electronics supply. HBM's TSV stacking shares wafer capacity with conventional DRAM—every wafer Nvidia secures for HBM is one fewer available for Apple's or Qualcomm's LPDDR5. This is a zero-sum technical trade-off, not a simple supply gap. Samsung, SK Hynix, and Micron are replaying the 2017 server-demand squeeze on mobile memory, but AI inference's exponential bandwidth hunger makes this cycle structurally different in magnitude. Apple's most rational counter is accelerating in-house memory controllers and custom advanced packaging—shifting leverage from "buying chips" to "defining specifications." Qualcomm is more exposed; Snapdragon's BOM carries the highest elasticity to memory pricing, likely forcing a 15-25% premium acceptance or a delayed flagship cadence. On the compliance front, tighter US export controls on advanced packaging could paradoxically free some HBM capacity for consumer tiers in the short term, while amplifying "friend-shoring" fragmentation costs over the medium term. Within 18 months, a 30%+ upward repricing of memory is near-certain. The deeper tail effect: AI is permanently rewriting memory's valuation logic from cyclical to infrastructure-driven. Downstream players must either capitalize upstream to lock capacity or architect around lower memory dependency. There is no third path.