Smartphone makers face a margin squeeze as DRAM and NAND costs have exploded due to AI data-center demand. DRAM average selling prices have quadrupled and NAND prices have tripled over the past 12 months, according to Counterpoint Research. Memory vendors are redirecting capacity to AI accelerators and HBM for hyperscalers (Microsoft, Google, Meta, AWS), starving consumer electronics of supply. Apple is moving to TSMC's 2nm node for its A-series SoCs, one of the foundry's most expensive processes, adding further pressure.
Smartphone unit shipments are expected to fall to 950 million in 2026 from 1.2 billion in 2025, with DRAM cost increases as the biggest factor. Apple's silicon cost at 2nm will reach $80, roughly 10x higher than when Apple used TSMC's 28nm process a decade ago. TSMC's 2nm wafers now exceed $30,000. Smartphone makers face two choices: absorb losses or fight for share amid narrow margins. Apple can likely raise prices without losing share (all competitors face the same crunch), while Chinese OEMs like Xiaomi, with average selling prices around $160, struggle to pass costs to price-sensitive customers.
The market is reshuffling: premium brands like Apple and Samsung benefit from the ability to raise ASPs and maintain margins, while mid-range and budget segments face extinction or margin collapse. Samsung is particularly shielded due to vertical integration (makes its own memory and logic). Chinese competitors and lower-cost Android makers take the hardest hit as memory now represents 15-20% of BOM for mid-range devices. For investors tracking capex, this validates the reality that AI infrastructure is pulling the lion\'s share of wafer capacity away from consumer electronics.