Stanford DAM releases 66-year memory price database
Original: Historical memory prices 1960-2026
Why This Matters
Comprehensive historical memory pricing data supports AI infrastructure cost analysis and semiconductor industry research.
Stanford University's DAM laboratory published historical memory pricing data spanning 1960-2026, tracking DRAM, NAND flash, and HBM costs per gigabyte with interactive visualization and downloadable datasets updated monthly.
Stanford University's Data and Memory (DAM) laboratory released a comprehensive historical memory pricing dataset extending from 1960 to 2026, building on John C. McCallum's classic memory-price dataset. The interactive database tracks three primary memory types: DRAM, NAND flash, and HBM (high-bandwidth memory). DRAM data, sourced from the McCallum dataset through mid-2024 and supplemented by Keepa Amazon retail prices thereafter, shows cheapest consumer DIMM pricing across generations including Pre-DDR, DDR, DDR2, DDR3, DDR4, and DDR5. NAND flash pricing spans 2010-present, using Keepa's cheapest consumer NVMe SSD prices from 2016 onward with approximate anchor points before. HBM data consists of industry-analyst estimates from TrendForce and SemiAnalysis, reflecting confidential accelerator contracts without public spot market pricing. The database includes a quarterly accelerator cost breakdown model from Epoch AI, showing production-volume-weighted component costs across Nvidia, AMD, Google TPU, and Amazon Trainium accelerators, with HBM4 projections through Q3 2026. Pricing is presented in nominal USD as cheapest listed retail prices, not contract averages or inflation-adjusted values. Users can download raw CSV data, toggle visualization series, and export charts via interactive charts.