The AI sell-off is exposing a big divide: infrastructure demand versus semiconductor earnings expectations. Micron is reportedly 48% below its all-time high. SK Hynix is down 39%. Yet these are two of the companies most directly exposed to the high-bandwidth memory powering AI accelerators. The valuation debate is getting interesting: If Micron really trades at roughly 5x forward earnings, is the market overlooking its role in the AI buildout — or assuming today’s memory profits are close to a cyclical peak? That’s the critical distinction. AI demand can keep growing while memory prices, margins and earnings eventually normalise. A low P/E doesn’t automatically mean a stock is cheap if the earnings denominator is temporary. Meanwhile, Marvell and Astera Labs have also taken heavy hits in the infrastructure sell-off, even as some of these names remain significantly higher year to date. The market is forcing investors to separate long-term AI demand from the sustainability of the profits being priced in today. #AI #Micron #SKHynix #Semiconductors #Investing