Buy, Sell, or Hold: SanDisk at $2,184 and Micron at $1,134

Micron reported $23.86B in fiscal Q2 revenue, 22.28% above consensus, with non-GAAP EPS of $12.20 and Q3 guidance for $33.5B revenue and roughly 81% gross margin; SanDisk posted Q3 FY26 revenue up 251% year over year to $5.95B, gross margin expansion to 78.4%, and Q4 EPS guidance of $30-$33. Both names are executing strongly on AI memory demand, but the article stresses that valuations have outrun sell-side targets, with Micron trading about 39% above its $879.10 target gap and SanDisk about 820% YTD. The near-term setup is positive on fundamentals, but sentiment is tempered by late-cycle valuation risk and the potential for guide-down-driven derating.

Analysis

The market is starting to treat AI memory like a quasi-utility, but the hidden loser is the rest of the NAND/DRAM ecosystem: smaller non-AI consumers are being rationed out while hyperscaler contracts set the clearing price. That tends to extend near-term pricing power, yet it also creates a more brittle demand base because a handful of customers can abruptly pause orders if server deployment schedules slip. In that sense, the rally is less about broad end-demand and more about allocation scarcity, which can reverse faster than traditional cycle peaks.

The key second-order issue is supply response timing. High margins and zero-debt balance sheets are usually interpreted as durability, but in semis they are also the signal that capex will eventually chase the curve, first through equipment vendors and then through incremental wafer starts that pressure pricing 2-4 quarters later. Micron’s current setup is cleaner because the forward multiple still leaves room for another leg if guidance stays conservative, while SanDisk’s valuation is already anchored to flawless execution and no air pockets in consumer or enterprise demand.

The consensus seems to be underestimating how much of the move is positioning rather than fundamentals. With both names crowded into the same “AI memory scarcity” trade, a benign earnings print can still produce a sell-the-news reaction if guidance is merely good instead of exceptional. The asymmetric risk over the next 1-3 months is not collapse in demand, but multiple compression from any hint that capacity additions, customer digestion, or mix normalization will flatten the earnings trajectory.

WDC is the quiet relative loser: even if the stock has operating leverage to the cycle, the market is increasingly rewarding pure-play exposure to the scarce components, not legacy storage platforms with less direct pricing power. That makes this a second-order short if investors want expression against the theme without fighting the commodity upside in memory itself.

Key Decisions for Investors

• Hold MU long into the next earnings window, but only as a tactical trade: target a post-print upside extension if guidance stays tight; cut quickly on any in-line guide because the stock is crowded and can de-rate 10-15% in days.
• Avoid initiating fresh SNDK longs at current levels; if already long, tighten stops and consider selling upside via call spreads into strength, since the setup is priced for perfection and vulnerable to a 15-25% multiple reset on any guide miss.
• Pair trade: long MU / short WDC for 4-8 weeks. MU has the cleaner AI-memory torque and better re-rating runway; WDC is more exposed to a normalization in non-hyperscaler storage demand and should underperform if memory pricing cools.
• For more aggressive expression, buy MU call spreads 1-2 months out rather than stock. This preserves upside to a beat-and-raise while defining risk in a name where implied volatility is likely to stay elevated.
• Set a catalyst watch on capex commentary from both names over the next two quarters: if managements signal aggressive supply response, begin rotating out of the trade because pricing power will likely peak before reported earnings do.