Ouch! iPhone memory costs may quadruple, says JP Morgan

JP Morgan says memory could rise to 45% of iPhone component costs next year from about 10% today, implying a sharp margin headwind for Apple and other smartphone makers. The report highlights intensifying competition for RAM from AI data-center buyers such as Nvidia, which is pushing up prices and reducing smartphone vendors’ bargaining power. Apple may offset some pressure through split production runs, recycling, and a margin hit, but competitors using more memory are likely to be squeezed more.

Analysis

The first-order read is not that Apple loses pricing power; it is that memory inflation becomes a relative-advantage event. In a constrained component market, the OEMs with the strongest balance sheet, tighter SKU mix, and the ability to absorb margin for share can defend unit volumes, while weaker Android vendors face a worse combination of higher bill of materials and less differentiation. That sets up a likely bifurcation over the next 2-4 quarters: premium platforms can choose to eat some cost, but value-tier vendors are forced into price hikes or spec cuts that risk demand destruction.

The more important second-order effect is on the supply chain allocation stack. If AI infrastructure continues to bid aggressively for high-bandwidth and related memory, handset makers are effectively competing for leftover capacity, which can create sudden lead-time extensions and forcing orders ahead of schedule. That typically benefits the memory suppliers more than the end-device makers, but it also raises the odds of inventory overbuild if handset OEMs over-hedge into a rising market, setting up a margin unwind 1-2 quarters later if demand normalizes.

For Apple specifically, this is less about near-term earnings pressure and more about protecting ecosystem share. Apple can use service attachment, financing, and mix management to offset hardware gross margin compression, whereas weaker peers have fewer offsetting levers and may see ASP increases translate directly into unit declines. The contrarian risk is that the market may be overestimating how durable the memory spike is: if AI procurement moderates or memory capex catches up faster than expected, the pricing shock can fade quickly, making this a better tactical than structural thesis unless lead times keep widening.

The cleaner expression is to own the suppliers of constrained memory capacity and fade the most commoditized handset exposure. The highest-risk setup is a delayed reaction from OEMs: margins can look fine for one or two quarters because of channel inventory and prior buys, then reset abruptly when contract repricing rolls through. That makes the next earnings season the key catalyst window, not today’s spot price move.

Key Decisions for Investors

• Long NVDA vs short a basket of handset OEMs/consumer electronics assemblers for 1-3 months: AI infrastructure demand should continue to crowd out smartphone memory allocation, but prefer the side with pricing power and backlog visibility. Risk: if memory supply normalizes faster, the relative trade fades.
• Buy AAPL on a 3-6 month horizon on any post-earnings weakness, with the thesis that margin compression is manageable and share gain is the more likely strategic response. Risk/reward is better than the market’s first-order read because services and financing can offset hardware pressure.
• Avoid or underweight lower-end Android OEM exposure for the next 2 quarters; if forced, pair against AAPL or against memory suppliers. These names have the least ability to absorb BOM inflation and the most risk of unit erosion from price pass-through.
• If accessible, buy out-of-the-money calls on AAPL into the next earnings cycle and finance via short-dated puts only if implied vol remains reasonable. The catalyst is management commentary on mix, channel inventory, and margin defense; upside comes from a less-bad outcome than consensus fears.
• Watch for a reversal indicator: memory lead times flattening and capex commentary from suppliers easing. If that happens, rotate out of supplier longs quickly; this is a 1-2 quarter tactical dislocation, not necessarily a multi-year supercycle.