AI agents will soon outnumber the global human population, leading to an era where AI agents primarily interact with and query other AI systems rather than humans. This surge in AI-generated requests will demand much faster database and application processing, requiring broader computing architecture innovations. As a result, the value of memory chips is expected to move away from pure storage capacity (price per GB) toward processing performance, speed, and efficiency, especially the ability to handle rapid data access and token generation for AI inference workloads. Panelists from Samsung, Micron, SK Hynix, and others highlighted the limits of HBM alone and the rise of hierarchical/hybrid memory architectures that combine HBM, DRAM, and high-bandwidth NAND flash (HBF). The overall market is seen shifting from training-centric to inference-centric AI, making high-capacity, high-speed memory solutions increasingly critical. -FMS 2026 Event in Santa Clara, California. $MU $SKHY $DRAM $CRWD $NET $LITE