Sandisk SNDK and Micron Technology MU are turning into one of the more unusual stories in the AI trade, with both companies becoming cheaper on forward earnings even as their shares race higher. Sandisk has surged 482% this year, making it the top performer in the S&P 500 Index, while its price-to-estimated earnings ratio has fallen below nine from 23 just a few months ago. Micron has advanced 172% in 2026, ranking as the fifth-best performer in the S&P 500, yet it also trades at less than nine times forward earnings, down from about 12 in February. That compares with roughly 21 times forward earnings for the broader S&P 500, suggesting investors are paying a much lower multiple for two memory names whose earnings expectations have been rising faster than their share prices.

The reason sits at the center of the AI infrastructure boom: demand for memory remains intense, supply is tight, and expectations for future earnings have moved sharply higher. Analyst estimates for Sandisk’s 2027 adjusted earnings per share have climbed nearly 2,000% over the last 12 months, while Micron’s have risen 768%. Kim Forrest of Bokeh Capital Partners, which owns Micron shares, said high-bandwidth memory can be priced very high because of supply and demand, while Rob Thummel of Tortoise Capital, which holds Sandisk and Micron in its Tortoise AI Infrastructure ETF, said memory and storage demand could keep rising unless hyperscaler capital spending changes significantly. The Tortoise ETF is up 70% this year, while the Philadelphia Stock Exchange Semiconductor Index has gained 71%, showing how much momentum is already embedded in the broader semiconductor trade.

Still, the valuation reset carries a possible warning because memory stocks have long moved through boom-and-bust cycles. Randy Hare of Huntington National Bank, which holds Micron shares, said these companies cannot be viewed like steady earnings-growth businesses and suggested the easier money in Micron may already have been made, even though he still sees upside in the trade. Micron’s recent history shows the risk: it reported an adjusted loss per share of $4.45 in 2023, after its shares fell 46% in 2022 as earnings-growth expectations weakened, before rebounding as a new cycle began. Jed Ellerbroek of Argent Capital Management, which does not own memory stocks because of their cyclicality, warned that shortages eventually turn into gluts, though he said it may take a couple of years for supply to catch up. For now, bulls such as Jay Hatfield of Infrastructure Capital Management argue the boom remains supported by valuation, while investors are left weighing whether AI demand can keep pushing estimates higher before the cycle possibly turns.