Citrini: AMD and Apple simultaneously promote the replacement of DRAM with flash memory in AI products BlockBeats News: On June 16th, Citrini Research, the publisher of the "AI Apocalypse Report," announced that AMD and Apple are currently simultaneously promoting the replacement of DRAM with flash memory in AI products. Among them, AMD acquired MEXT optimized flash memory, making its performance close to DRAM, thereby reducing the memory cost of AI data centers; Apple implements similar optimizations on the device side through the "LLM in a flash" technology. According to the latest research report by Citrini, the high memory demand of KV cache in AI inference, as well as the "memory tax" pressure of HBM accounting for 25% of DRAM production capacity. The cost of flash memory is only 1/55 of DRAM, and through controller optimization, NAND stacking, and cell mode adjustment, it can provide an alternative solution for edge AI in terms of capacity and bandwidth. The above report provides theoretical support for the sustained surge in storage stocks represented by SanDisk in recent times. [Original link]