TrendForce Sees DRAM Up 10-15% and NAND 15-20% in 4Q26 as AI Server Demand Bites
TrendForce projects conventional DRAM contract prices to rise 10-15% quarter on quarter in 4Q26 and NAND Flash 15-20%, with enterprise SSDs the only category where price growth accelerates, as AI inference build-outs push 2026 enterprise SSD bit demand up more than 80% year on year.
TrendForce's latest memory research shows DRAM suppliers continuing to prioritise advanced-process capacity for high-performance server products in the fourth quarter, keeping the overall market undersupplied even as the pace of increase moderates. Conventional DRAM contract prices are projected to grow 10-15% quarter on quarter, while NAND Flash contract prices are seen up 15-20%. In the server DRAM market, improving server CPU availability has prompted cloud service providers and server OEMs to increase bit procurement to meet growing RDIMM demand for general servers supporting agentic AI workloads, and server DRAM is expected to remain undersupplied through the quarter.
Enterprise SSDs are the standout. Bit demand for enterprise SSDs is projected to grow by more than 80% year on year in 2026 as cloud providers expand AI inference infrastructure and shift from large language model training toward actual AI deployment, with the rollout of agentic AI driving exponential growth in data volumes for real-time retrieval and caching and QLC penetration in vector databases continuing to rise. Suppliers are expanding QLC capacity and ramping PCIe 6.0 alongside next-generation AI server platforms, but most incremental supply has already been committed, so price growth is expected to accelerate, making enterprise SSD the only memory category with a larger increase than in the prior quarter.
Consumer-side pressure is building at the same time. PC OEMs continue to procure aggressively on expectations that DRAM supply will stay tight in 2027, even as notebook inventories increasingly reflect higher component costs and PC DRAM supply is expected to decline next year as more capacity shifts to server applications. Mobile DRAM allocations remain tight and GDDR7 demand continues to be driven primarily by AI chips, with suppliers scaling back GDDR6 production. In parallel, NVIDIA's NVHBM architecture, announced in August, moves the memory controller into the HBM base die and is reshaping how memory makers source those dies, with Micron and Samsung already turning to external foundries including TSMC.