Search

The memory cycle, rewritten

Memory prices continue soaring. And this time, the cycle has a structural floor.

AI-driven demand has sent DRAM and NAND prices to multi-year highs, while a wave of long-term agreements and strategic investments is quietly reworking how the memory industry manages its next downturn. 

Yole Group’s article highlights today how AI-driven demand is pushing DRAM and NAND prices to multi-year highs, while fundamentally reshaping the memory industry. Written by Simone Bertolazzi, PhD, Principal Analyst, Memory at Yole Group, this snapshot underscores a key structural shift toward multi-year agreements and deeper supply chain integration, which could make future downturns less severe. At the same time, it points to potential risks in consumer markets, where rising prices and macroeconomic pressures may impact demand. 

These insights align with Yole Group’s comprehensive memory product portfolio, including the DRAM Market MonitorNAND Market MonitorStatus of the Memory Industry 2025 report, and Next-Gen DRAM 2026 – Focus on HBM and 3D DRAM 2026 report, among others. Leveraging deep expertise in memory technologies and markets, Yole Group’s analysts deliver detailed analysis of market trends, pricing, and supply–demand dynamics to support strategic decision-making. 

Enjoy your reading!

Simone Bertolazzi, PhD Principal Analyst, Memory at Yole Group
AI infrastructure demand is not only driving memory prices to multi-year highs, it is fundamentally reshaping the industry’s dynamics.

AI Supercycle Drives Record Profits Across Memory Industry 

The memory market entered 2026 at full tilt. Blended DRAM average selling prices (ASPs) surged approximately 27% quarter-on-quarter in Q4 2025, with momentum accelerating further to exceed 50% QoQ growth in Q1 2026. NAND followed, with blended ASPs also rising roughly 27% and more than 50% over the same periods, a sharp deviation from the modest single-digit growth anticipated in 2025 for early 2026. 

The driver is structural. AI infrastructure has rapidly expanded from model training to large-scale inference deployment, increasing demand not only for high-bandwidth memory (HBM) but also for mainstream server DRAM configurations and enterprise SSDs, at volumes that have outpaced available supply. Hyperscaler investment surged throughout 2025, driven by a wave of announced AI data centers and gigafactories, pushing AI server deployments well beyond earlier forecasts. 

This article is reserved for logged-in users. You have 80% left to discover.
up