The memory trade is not the next leg of AI by default
Investors are rushing into memory as the obvious follow-on to the AI trade, but that shortcut ignores what memory has always been: a cyclical business with brutal pricing swings and limited differentiation. That matters now because semis just suffered their sharpest weekly drawdown in over a year, and the search for the next AI winner is colliding with stretched expectations in **MU**, **SNDK**, and **WDC**.
The market is making a familiar mistake with memory: taking a real demand tailwind and turning it into a clean secular story. Yes, AI is lifting demand for DRAM, HBM, and related storage, and that part is not in dispute. The problem is that investors are increasingly treating memory as the automatic next leg after AI leadership wobbled, when this is still the part of the semiconductor stack most exposed to supply response, pricing reversals, and fast multiple compression. In other words, memory may benefit from AI, but that does not make every memory name the next NVDA
