Stanford's historical memory price dataset shows DRAM cost decline from $/bit in 1960 to fractions of a cent today
Tags Infrastructure · Hardware
A Stanford Digital Archaeology Museum (DAM) project presents a comprehensive historical dataset of memory prices spanning 1960 to 2026, showing the exponential decline in storage cost over six decades. The visualization demonstrates how memory costs dropped by approximately 12 orders of magnitude during this period, following a trajectory similar to Moore's Law. The dataset caught significant attention on Hacker News (307 points, 111 comments) as both a historical reference and a lens for understanding semiconductor economics. The current trend shows signs of price stabilization and even increase in recent years, partly driven by AI-related demand concentration among fewer suppliers.
Technical significance
The dataset provides empirical grounding for understanding semiconductor cost curves that underpin AI economics. The recent price plateau and uptick is particularly significant — it suggests that the decades-long trend of exponentially cheaper memory may be ending due to supply chain consolidation and AI-driven demand, which has direct implications for the cost trajectory of AI training and inference infrastructure.