Report by Morgan Stanley
Just a few days ago, Apple announced a 20 per cent price rise on some of its products due to the sharp increase in memory and storage costs, a phenomenon which, according to Shawn Kim, an analyst at Morgan Stanley, is becoming one of the clearest signs of the growing “inelasticity” of the AI-driven economy. After decades in which the cost of memory had been falling almost structurally, the market is now experiencing a significant break from that trend. Memory prices have risen more than sixfold over the past year and the total market could grow from approximately $220 Bn in 2025 to around $890 Bn in 2026 – a rate of growth that exceeds the individual total addressable market (TAM) of the smartphone, PC or server markets. Kim emphasises that this is not a response to a traditional semiconductor cycle, but rather a structural demand shock driven by AI, where hyperscalers continue to invest aggressively despite the sharp rise in costs. Demand remains highly inelastic, whilst supply continues to be constrained by wafer capacity, EUV tools and long lead times for capacity expansion, creating an exceptionally tight pricing environment.
The expert points out that the problem is not solely one of capacity, but of allocation. The three major DRAM manufacturers, which control approximately 90 per cent of the market and virtually all HBM production, are prioritising higher value-added memory for AI and data centres, diverting capacity away from traditional consumer markets. This is creating a two-speed market, where hyperscalers secure supply through LTAs and prepayments, whilst the rest of the buyers compete for an increasingly limited and volatile supply pool.
Shawn Kim points out that chip-driven inflation is not yet fully reflected in the CPI, but is beginning to show up in the IPP, corporate costs, cloud bills, capex budgets and delays in technology roll-outs. For investors, the main takeaway is that the economic impact of this ‘chipflation’ may still be underestimated, as much of the cost shock is still being passed on slowly through business channels and could take years to become fully visible in profits, margins and final prices.
