Data compiled by Woofun AI shows that Bernstein has significantly raised price targets for Samsung Electronics, SK Hynix, and Micron, reflecting a structural shift in memory pricing dynamics. The investment bank notes that traditional DRAM prices have surged approximately 4.5 times since the third quarter of 2025, outpacing High Bandwidth Memory (HBM) adjustments due to annual contract constraints. This disparity is driving manufacturers to renegotiate HBM pricing with GPU and XPU vendors for 2027, with Bernstein predicting HBM prices could double or triple next year.

Concurrently, the report highlights that these cost pressures may amplify AI infrastructure expenses. If Nvidia aims to preserve a 75% gross margin, it may pass on HBM cost increases to customers at a fourfold multiplier. Bernstein estimates that combined hikes in HBM, traditional DRAM, and NAND could elevate total capital expenditure in AI data centers by nearly 30%. While maintaining 'Outperform' ratings for the three major memory producers, Bernstein warns that cloud providers will likely need to realign supply chain pricing and cost-sharing models to manage these escalating inputs.