Micron is tech's new margin king as memory crisis pushes company past Nvidia and Meta

Micron reported a record fiscal Q3 gross margin of 84.9%, up from 74.9% in the prior quarter and 39% a year ago, alongside revenue of $41.46B and net income of $28.24B. The company guided to roughly 86% gross margin in fiscal Q4 and said memory pricing should stay tight beyond 2027, reflecting strong AI-driven demand and pricing power. The stock was up more than 700% over the past year and gained another 14% in extended trading.

Analysis

Micron’s margin breakout is less about a cyclical upturn and more about a structural change in bargaining power: AI infrastructure buyers have converted memory from a cost item into a gating input, which means pricing can stay irrationally high longer than most semiconductor cycles. The second-order effect is that the entire AI stack is seeing a hidden tax re-routed from compute to memory, compressing downstream application margins unless cloud providers pass costs through or slow deployment pace.

The biggest incremental winners are the names with the least elastic demand and the cleanest ability to pass through component inflation. That argues for relative outperformance in hyperscaler-facing compute and networking where capex is still justified by model training demand, while consumer hardware remains the most exposed squeeze point because memory is a larger share of bill of materials and end-demand is far less price insensitive. The pressure on consumer OEMs is likely to show up first as weaker gross margin guidance, then as delayed launches and fewer promotions rather than immediate unit collapse.

The market is likely underestimating duration risk in memory pricing. If management is right that tightness persists past 2027, the real trade is not the next quarter’s beat but the possibility that a capacity response from the broader memory ecosystem arrives too late to matter for the current cycle, leaving spot prices elevated into 2026. The contrarian risk is that extreme profitability triggers a capital spending wave and customer design substitution, which would cap the duration of supernormal margins and compress the multiple before the earnings stream actually rolls over.

For now, this is a strong signal that AI capex winners are becoming increasingly bifurcated: compute and interconnect names can still win, but consumer tech and lower-end device makers may be forced into margin repair. That argues for staying long the constrained supply chain and short the exposed end-market where pricing power is weakest.

Key Decisions for Investors

• Long SNDK / short AAPL as a 3-6 month relative-value trade: memory scarcity supports Sandisk’s pricing power, while Apple faces the clearest margin compression and pricing pass-through risk; target 15-20% spread capture with a stop if consumer demand inflects positively.
• Add to NVDA and AMD on 1-3 month weakness, not strength: both remain exposed to memory cost inflation, but AI demand should keep revenue elasticity intact; use pullbacks as entry and hedge with a small short in AAPL to offset supply-chain inflation risk.
• Buy AVGO vs. MSFT on a 6-12 month horizon: Broadcom’s infrastructure-heavy mix and pricing discipline should better absorb component inflation than software-heavy peers if enterprise budgets tighten from AI hardware costs.
• Avoid or underweight consumer hardware names into Q4 guidance cycles: the risk/reward is poor because margin pressure typically shows up before unit weakness, and consensus may still be underestimating the delay in pass-through.
• For aggressive accounts, buy 9-12 month out-of-the-money calls on SNDK funded by selling near-dated upside in NVDA: this expresses the view that memory scarcity outlasts near-term compute enthusiasm, while limiting premium outlay.