Kioxia has commented on the growing importance of memory and storage in AI data centers and has released its future roadmap.
Kioxia Unveils Roadmap for Memory and Storage Targeting AI Data Centers
This article is a translation. Read the Japanese original
As AI chips continue to evolve, expanding memory bandwidth has become a significant challenge. The company stated that it will provide solutions to expand DRAM capacity by leveraging CXL (Compute Express Link).
Specifically, Kioxia said that memory expansion using CXL will make it possible to significantly increase the memory capacity of CPUs and GPUs. Looking ahead, the company anticipates configurations that allow for the recognition of large-capacity DRAM, such as 225GB or 148GB, via CXL.
In the storage sector, the company plans to deploy XL-FLASH.
XL-FLASH is a storage technology that operates at high speeds, similar to memory, by utilizing CXL. Kioxia stated that it is advancing support for CXL in addition to existing PCIe interfaces.
Kioxia plans to expand 3D NAND stacking to over 100 layers by 2030. Furthermore, the company stated that it aims to achieve both high density and low cost through the adoption of QLC (Quad Level Cell) technology.
Source: キオクシアのフラッシュメモリとSSDはなぜAIデータセンターに必要とされるのか (ITmedia AI+, 2026-09-04)