One of the defining market stories of the past 12 months has not been AI chips that compute, but the chips that remember. Equity analyst Shan Rui Yeo explains how memory works, from DRAM and NAND to high bandwidth memory, and how an industry that destroyed wealth for four decades became disciplined after consolidating to three players in 2013. He then walks through what changed: AI inference has made memory the key bottleneck, memory content is climbing with each new generation of GPUs, and new supply takes three to four years to build. With prices up sharply and customers signing long-term agreements, Part 1 of this three-part conversation lands on a commodity industry whose business model is changing in real time.