DRAM & DDR5
DRAM contract prices are reported up more than 50% this quarter, though Micron's long-term supply pacts are reportedly capping some of the spot upside for customers locked into those agreements. (Gate News) Separately, Micron's CEO has stated that data center customers are seeking 50% more memory than the company can currently ship, and customers have committed approximately $22 billion in forward procurement to secure allocation. (The Motley Fool) For standard DDR5 procurement, the flat index reading today suggests near-term spot stability, but the demand signal from hyperscalers points to continued tightness.
NAND & SSD
NAND prices surged an estimated 70% in recent quarters; TrendForce currently projects a more modest 10–15% increase this quarter as the pace moderates. (The Motley Fool) AI server demand drove the top five NAND suppliers' combined revenue up 77% in Q2 2026. (SemiMedia) The SSD index sitting near par (99.8) suggests consumer and commercial SSD segments are not yet fully reflecting the enterprise NAND tightness, but procurement teams with Q4 refresh cycles should treat that gap as temporary.
HBM & AI Infrastructure Pricing Pressure
All three major suppliers — Samsung, SK Hynix, and Micron — are qualified to supply HBM4 for Nvidia's Vera Rubin platform, with Q3 delivery underway. (Gate News) SK Hynix reports 12-layer HBM4 in mass production with 16-layer in qualification. (Herald Business) Nvidia's reported 15% server price increase is expected to further strengthen Samsung and SK Hynix pricing leverage, with HBM prices potentially rising 50%+ in 2027. (TrendForce)
Procurement Implications
TrendForce projects DRAM and NAND will constitute 68% of major cloud service provider capex in 2027 — a figure that signals broad infrastructure repricing is already in motion. (TrendForce) OVHcloud has already raised prices, citing AI memory demand as a cost driver for non-AI infrastructure. (InfoQ) Teams planning server refreshes or cloud cost modeling beyond 2026 should factor in sustained memory cost inflation, particularly for any workloads competing for DRAM or NAND allocation with hyperscale AI deployments.