Nanya Targets Edge AI With Custom UWIO DRAM

Artificial intelligence workloads are expanding from cloud-based model training to inference on edge devices, reshaping demand for memory that delivers high bandwidth with lower power consumption. While high-bandwidth memory, or HBM, dominates AI training systems, its energy requirements and architectural constraints are opening opportunities for customized DRAM products designed to work with advanced packaging and device-specific processors.

Nanya Technology recently detailed plans to develop customized UWIO DRAM integrated with 3D IC stacking technology, targeting memory demand for edge AI inference. The Taiwanese chipmaker expects mass production to begin at its new Taishan facility in early 2027. Nanya has not disclosed the investment amount, initial production capacity or prospective customers for the program.