Micron and Sandisk Shares Are Crashing. Is It Time to Buy the Dip?

The key market mechanism is not "AI demand" in the abstract, but who captures the scarcity rent. Memory vendors are the obvious winners, yet the second-order losers are the system builders and hyperscalers that sit one layer downstream and have less ability to pass through BOM inflation; that can quietly compress server gross margins even while capex headlines stay strong. Within memory, MU is the cleaner relative winner because diversified exposure and larger scale should let it monetize tightness with less single-product risk than SNDK.

The immediate selloff looks more like position cleanup after an extreme run than a thesis break, but the next 1-3 months will be about inventory and pricing confirmation, not narrative. If spot pricing or channel checks soften, these names can de-rate 20-30% quickly because the market is still paying for scarcity, not just earnings. Over 6-18 months, the upside case depends on whether supply additions lag demand long enough for this to become a durable earnings step-up rather than a classic semiconductor peak.

The contrarian miss is that memory is the least forgiving corner of semis: high returns invite aggressive capacity response, and the supply response can arrive faster than consensus expects once utilization and pricing stay elevated. That makes SNDK the more vulnerable of the two after its outsized move, since pure-NAND exposure leaves it with less cushion if AI-related storage demand proves less elastic than DRAM demand. Falsifiers: sequential DRAM/NAND ASP declines, rising inventory days, or any guide that implies the next quarter is troughing earlier than expected.