Lexar is developing storage technology aimed at shifting some local AI workload demands from DRAM to NAND Flash. Chief Technical Officer Daniel Guo said DRAM is about six times more expensive to manufacture than NAND Flash, creating an opportunity for AI SSDs to lower the memory requirements of running large language models on local hardware. The Lexar AI Storage Core SSD is intended to support local AI deployments by offloading models to storage, reducing memory footprint by at least 40%.
In internal testing, Lexar ran the Qwen 3.5 122B AI model on a local PC. The company said its Lexar AI suite and AI Storage Core SSD reduced the DRAM requirement from 128 GB to 32 GB for a related configuration, running a model with 35 billion parameters at 15.6 tokens per second versus 5.2 tokens per second using traditional frameworks. When loading the 122B model on 32 GB of DRAM, Llama.cpp failed and crashed, while Lexar’s SSD offloading reached about 4.4 tokens per second.
