BlockBeats news, June 16: Citrini Research, the issuer of the "AI Doomsday Report," stated that AMD and Apple are currently advancing solutions to replace DRAM with flash memory in their AI products. AMD has acquired MEXT to optimize flash memory, bringing its performance close to that of DRAM and reducing memory costs in AI data centers; Apple is achieving similar optimizations on-device through its "LLM in a Flash" technology.
According to Citrini’s latest research report, the high memory demand of KV caching in AI inference, combined with the “memory tax” pressure from HBM accounting for 25% of DRAM capacity, highlights the opportunity: flash memory costs only 1/55th of DRAM, and through controller optimization, NAND stacking, and cell pattern adjustments, it can provide a viable alternative in capacity and bandwidth for edge AI.
The above report provides theoretical support for the recent sustained surge in storage stocks, led by SanDisk.
