Micron’s Pullback Tests the AI Memory Boom as Google Flags TurboQuant
Analysts say software gains may not erase tight supply for high‑bandwidth memory.
Overview
- Micron shares now trade about 19% below a recent peak after a post‑earnings surge that featured revenue roughly tripling and gross margin rising to 74.4%.
- The retreat reflects fresh questions about lofty expectations and whether Google’s TurboQuant compression research could change how much memory AI systems need.
- AI chips rely on high‑bandwidth memory, a stacked form of DRAM placed close to the processor for faster data flow, which has been scarce and costly.
- Producing high‑bandwidth memory uses about three times more wafer capacity than standard DRAM, which has tightened supply and boosted prices for memory makers.
- Micron is steering more output to high‑bandwidth memory and signing long‑term deals with minimum volumes and quarterly price resets, while one analysis pegs the stock at 3.3 times projected fiscal 2027 earnings and frames a potential 2030 triple as a prediction rather than a guarantee.