The explosive demand generated by artificial intelligence is creating a supply crunch for memory chips so severe that it is predicted to last through 2027, fundamentally reshaping the semiconductor landscape and stock valuations. This scarcity is not merely a transient market swing but a deep structural shift that promises tremendous earnings power for memory producers while simultaneously causing significant pain for downstream manufacturers of consumer electronics.
CJ Muse, Senior Analyst at Cantor Fitzgerald, recently joined CNBC’s The Exchange to discuss the unprecedented momentum in memory stocks, emphasizing that the current dynamics are not merely cyclical but represent a structural, multi-year super-cycle driven entirely by AI adoption and capacity constraints in DRAM and NAND production. Muse dismisses concerns that the recent surge in memory stock valuations, like the over 230% jump seen in Sandisk Corp (SNDK) over three months, is a temporary, cyclical blip. He argues that traditional memory cycles, often tied to PC refresh rates, are being superseded by the insatiable appetite of AI infrastructure.
The analyst pointed out that the NAND space, where Sandisk operates, has been in "essentially a seven-year down cycle," meaning there has been no urgency among players to add capacity. This historical underinvestment now collides directly with the massive, sustained demand for high-bandwidth memory (HBM) required by AI servers. Muse stated plainly: “I do believe that this cycle is different… I think we’re going to be undersupplied for all of 26 and 27.” This projected deficit stems from the combination of robust content growth (the amount of memory per device) and inherent limitations in capacity expansion, setting the stage for a prolonged seller's market.
