3 Stocks That Get Paid More When Memory Prices Rise. I'd Buy Only 1 of Them.

Record DRAM and NAND prices are driving exceptional revenue and margin expansion for Micron, Sandisk, and SK Hynix, despite only modest shipment-volume growth. Micron's DRAM revenue rose 67% sequentially while bit shipments increased only low single digits; it reported 84.6% gross margin and guided for roughly $50B of Q4 revenue, up about 21% sequentially. Sandisk's revenue rose 51% sequentially to $8.97B, with roughly two-thirds of growth from pricing, while SK Hynix reported about KRW79T in Q2 revenue and a 76% operating margin. The article cautions that memory pricing is cyclical despite long-term supply agreements, favoring SK Hynix at roughly 6x expected 2027 earnings versus Micron at 7x and Sandisk at 9x.

Analysis

The relevant signal is not AI demand alone but operating leverage to memory ASPs: current earnings power embeds an unusually high contribution margin on price rather than bits. That makes MU, SKHY and SNDK effectively leveraged duration trades on 2027 contract-price discipline. The key competitive variable is supply behavior: incremental capacity from Samsung Electronics (005930.KS), CXMT, or an easing in HBM supply tightness would pressure commodity DRAM/NAND pricing well before reported revenue rolls over.

SK Hynix is the cleaner long among listed memory exposures because its HBM mix creates a relative scarcity premium and tighter attachment to NVDA/AI-server deployments. MU is more exposed to the market assigning a peak-cycle multiple to peak-cycle earnings; even modest ASP normalization could drive simultaneous EPS cuts and multiple compression. SNDK has the least diversification and should be viewed as the highest-beta NAND-price instrument rather than a durable AI compounder; enterprise SSD qualification gains would be the principal upside exception.

Over the next 1-3 months, watch quarterly contract-price surveys, lead times, customer inventory commentary from hyperscalers, and Samsung utilization/capex signals. Over 6-18 months, take-or-pay structures may reduce volatility at the margin but do not eliminate customers' ability to defer volumes, renegotiate at renewal, or shift sourcing; these contracts should not be capitalized as permanent protection without disclosure of duration, price bands, and counterparty concentration. Consensus may be underestimating the near-term scarcity premium in HBM but overestimating how much of broad DRAM/NAND pricing can remain structurally detached from a historically cyclical supply response.

Key Decisions for Investors

• Prefer long SKHY versus short MU as a 3-6 month relative-value trade: SKHY retains greater HBM scarcity exposure while MU has more downside if commodity-memory ASP expectations flatten. Target 10-15% relative outperformance; exit if SK Hynix HBM revenue mix/qualification commentary weakens or MU demonstrates sustained share gains in high-end HBM.
• Avoid adding outright SNDK exposure after price-led margin expansion. Use it as a downside hedge against a NAND contract-price reversal: initiate only after TrendForce/industry data show two consecutive downward monthly pricing revisions, targeting a 20% drawdown with a stop on renewed enterprise-SSD pricing acceleration.
• For existing MU longs, reduce delta into earnings or buy 3-6 month downside puts rather than extrapolating spot margins. A guide implying sequential ASP deceleration, higher customer inventories, or capex acceleration should be treated as thesis invalidation; the risk is an earnings reset larger than the initial revenue slowdown suggests.
• Monitor NVDA hyperscaler capex commentary as a leading demand read, but do not use NVDA strength alone as confirmation for commodity-memory longs. The actionable confirmation is rising HBM allocation alongside stable conventional DRAM/NAND contract prices; divergence between the two favors SKHY over MU and SNDK.