Micron's 29% Drop Has Nothing to Do With Fundamentals

The key mechanism is not “AI demand is strong” — that’s already consensus — but that compute scarcity is forcing model builders to buy more memory per unit of effective compute. That makes MU a levered beneficiary of AI server densification, especially in HBM and high-capacity DRAM, where mix can expand faster than unit demand. The second-order effect is important: if customers are turning away users because capacity is tight, they will over-order infrastructure to avoid future chokepoints, which supports pricing and lead times for at least the next 1-2 quarters.

The risk is that this is a cyclical memory business wearing an AI narrative. A strong current quarter does not prove a durable step-up in end demand; it can just as easily reflect inventory prebuilds and qualification shortages. Over a 6-18 month horizon, the usual memory-cycle response — Korean and Chinese capex, faster HBM qualification, and eventual contract-price normalization — is the main threat to margin durability and to the multiple expansion the stock would need to rerate materially.

Contrarian view: the market may be right to keep MU cheap if AI bottlenecks sit upstream in power, packaging, or GPU supply rather than memory. In that case, memory demand remains healthy but not scarce enough to justify paying up, and the upside becomes purely earnings-driven. The best falsifier is any sign that HBM/DRAM pricing, lead times, or customer order pull-ins flatten before the next print; that would cap the near-term trade even if the secular AI story stays intact.