Memory markets do not behave like competitive markets, and the current pricing crisis makes that structural reality impossible to ignore. Understanding why prices rise and stay elevated requires abandoning the assumption that new entrants, substitutes, or demand destruction will eventually force correction. None of those mechanisms operate effectively here.
DRAM production is controlled by three firms — Samsung, SK Hynix, and Micron — who collectively hold roughly 95% of global supply. NAND flash adds a handful of additional players but remains similarly oligopolistic. In a genuinely competitive market, elevated prices attract new entrants who expand supply. In memory, the barriers to entry are so prohibitive that no firm has successfully broken into commodity DRAM production in decades. A leading-edge fab costs $15–20 billion to build and requires years to reach volume production. That capital threshold functions as a permanent moat, insulating incumbents from competitive discipline.
The oligopoly structure enables producers to manage output without formal collusion. When prices fall, producers cut capex and reduce wafer starts; when prices rise, they expand slowly and selectively. This supply discipline — described in industry language as rational capacity management — replaces the chaotic output competition that drives prices down in genuinely competitive industries. Even when a producer decides to expand, it takes 18–24 months to bring new fab capacity online, meaning supply cannot respond quickly to price signals. Elevated prices persist long after the signal is clearly present.
Geopolitical insulation compounds the structural problem. US-China semiconductor export controls have effectively walled off Chinese producers — CXMT and YMTC chief among them — from fully competing in global markets. This removes what might otherwise be the most significant source of new competitive supply pressure. On the demand side, memory is a non-substitutable input for smartphones, servers, PCs, and AI accelerators. Buyers absorb price increases rather than redesign products or delay production. Demand does not soften enough to punish producers for high pricing, and switching from one memory supplier to another requires lengthy re-qualification processes that further reduce buyer leverage.
A Different Kind of Crunch
Previous memory downturns resolved when producers cut supply and demand recovered. The current crunch has a different structural driver. The voracious demand for high-bandwidth memory by hyperscalers — Microsoft, Google, Meta, Amazon — has forced Samsung, SK Hynix, and Micron to reallocate limited cleanroom space and capital expenditure toward higher-margin enterprise-grade components. Every wafer allocated to an HBM stack for an Nvidia GPU is a wafer denied to the LPDDR5X module of a mid-range smartphone or the SSD of a consumer laptop. According to IDC, the result is that 2026 DRAM and NAND supply growth will come in below historical norms at 16% and 17% year-on-year respectively — far short of what a demand environment driven by AI infrastructure expansion requires.
The pricing data reflects the severity of the imbalance. Conventional DRAM contract prices in Q1 2026 rose 90–95% quarter-over-quarter, the steepest quarterly increase in the history of the product category. PC DRAM prices at least doubled in the same period. NAND Flash rose 55–60% QoQ in Q1, and TrendForce projects a further 70–75% QoQ increase in Q2 2026, with conventional DRAM adding another 58–63% in the same quarter. Suppliers are simultaneously reallocating capacity toward server and HBM applications, implementing catch-up pricing to narrow gaps across product segments, and actively limiting shipments to consumer channels to support price floors.
The consumer device market is absorbing these increases through a combination of specification downgrades and retail price increases. Android brands targeting mid-to-low segments are raising launch prices and modifying product roadmaps. The most acute impact is in the low-end smartphone segment, where base configurations are reverting to 4GB RAM — a step backward that reverses years of democratization in consumer mobile hardware. For high-end ultrathin notebooks, where mobile DRAM is soldered directly onto the motherboard, there is no specification escape valve; price increases flow through directly to the retail shelf.
Timeline for Resolution
Meaningful capacity expansion is unlikely before late 2027 or 2028. The structural constraint is straightforward: leading-edge fabs require steep capital commitments under conditions — high interest rates, elevated energy costs, EUV-intensive processes — that make producers more cautious, not less, about adding conventional DRAM and NAND capacity. The current upcycle is therefore more prolonged and less volatile than previous boom-bust periods because the supply discipline is not merely strategic; it is partly dictated by economics.
Legacy DRAM generations will not provide relief. By 2027, DDR4 supply will be heavily concentrated at Nanya and Winbond, and pricing will reflect scarcity rather than cost. Buyers dependent on DDR4 beyond that point should treat it as a specialty component requiring multi-year contracts rather than a commodity available at spot. The market is not cycling back to the pricing environment that prevailed before the AI infrastructure buildout began.
Even when new capacity eventually comes online and the acute shortage eases, the structural reorientation of the industry toward HBM and enterprise memory means conventional consumer memory is permanently lower priority than it was in the previous decade. The crunch eases at the margins in late 2027 at the earliest; a full normalization of consumer-grade pricing is unlikely to occur at all. What the industry is experiencing is not a temporary dislocation. It is the permanent repricing of a commodity that was never as competitively supplied as it appeared.
