AI Memory Stocks Micron and Sandisk Are Up 200% in the Last 3 Months. History Says This Will Happen Next.

Memory chip prices surged on an AI-driven supply shortage, lifting Micron shares by 203% and Sandisk shares by 217% over three months. The backdrop is strong near-term fundamentals—NAND/DRAM prices up 200%/300% YoY, Micron May-quarter sales +345% with non-GAAP net income up 1,200%+, and Sandisk March-quarter sales +251% with non-GAAP profit of $23.41/share—yet the article warns the cycle may peak around 2028, with consensus earnings declines of 27% for Micron (fiscal 2029) and 54% for Sandisk. It highlights valuation dispersion (Micron at 24x earnings vs Sandisk at 67x) and flags a risk of sharp drawdowns if the shortage resolves—potentially up to ~50% or more.

Analysis

The key market mechanism is not "AI demand is good for memory"; it is that a commodity supply squeeze temporarily transfers pricing power to the memory vendors until procurement teams, foundry capex, and product redesigns force the cycle back toward equilibrium. That argues for near-term outperformance in MU and WDC/SNDK, but the second-order winners are actually downstream buyers that can delay purchases or redesign around costly memory content; the longer prices stay elevated, the more incentive hyperscalers have to push qualification toward alternate vendors, lower memory intensity, or slower AI deployment cadence.

Relative positioning matters more than the absolute bullish case. MU has a more diversified mix and should survive a normalization better than SNDK, which is cleaner NAND beta and therefore more vulnerable if spot pricing rolls over. The market is likely underestimating how fast an earnings multiple can compress once investors stop capitalizing peak margins and start discounting the next capex wave; in prior memory troughs, the stock move was driven less by the first revenue miss than by the abrupt recognition that supply discipline never lasts.

Catalyst timing is mostly 1-3 months around pricing commentary and forward guidance, not days. The bullish case remains intact as long as DRAM/NAND ASPs keep rising and management teams talk about constrained supply, but the bearish setup strengthens if memory pricing flattens while customer inventory builds. Contrarian takeaway: the consensus may be too eager to call the top because AI memory content is structurally different from legacy PC memory, so the cycle could stay tight longer than history implies; shorting too early risks fighting a powerful scarcity trade.

Key Decisions for Investors

• Long MU / short SNDK as a 3-12 month relative-value pair: MU has better survivability if the cycle extends, while SNDK carries more downside if the market starts discounting peak earnings; add only on pullbacks after earnings or when spot memory pricing stops accelerating.
• If using options, buy 6-9 month SNDK put spreads on strength rather than outright shorting the stock; the thesis is not immediate collapse but eventual multiple compression once pricing momentum breaks. Risk is capped to premium, and the trade only works if ASP growth decelerates into the next two quarters.
• Use WDC as a broader hedge against a memory downturn: a small short or put spread can express the view that commodity NAND is the first place where supply normalizes and pricing rolls over. Falsify if NAND pricing and company guidance remain tight through the next two reporting cycles.
• Do not chase an outright MU short today; the better entry is a tactical alert on any 10%+ gap-up that occurs without a fresh pricing surprise. If DRAM/NAND prices keep rising and capex stays disciplined, the bearish thesis is premature.