According to a July 12 report, the shift in artificial intelligence from training large language models to powering real-time reasoning is transforming NAND flash memory demand. While the first wave of generative AI fueled demand for high-bandwidth memory (HBM) and DRAM, the next phase relies on NAND for storing vast datasets AI systems retrieve instantly. The average price of a standard 128Gb MLC NAND product reached $28.80 at the end of June, extending an 18-month rally, with prices increasing more than fivefold since the end of 2025, outpacing a roughly 2.3-fold increase in benchmark PC DRAM prices. NAND has traded above DRAM prices for four consecutive months since March. This reflects reasoning models requiring enormous information repositories for retrieval-augmented generation (RAG), which reduces hallucinations by consulting external databases stored on NAND-based systems. NAND offers lower cost, data retention without power, and energy efficiency for hyperscale AI data centers, prompting replacement of hard disk drives with enterprise solid-state drives. Market researcher Omdia projects global NAND revenue will jump from about $73 billion in 2025 to $360.1 billion this year, then to $489.2 billion in 2027. Samsung Electronics held a 29% share of the NAND market in the first quarter and plans to expand production through its P5 and P5-2 fabrication plants.