Micron's Data Center Gross Margin Hit 87% Last Quarter. Here's What It Means for the Stock.

Micron’s core data-center gross margin jumped ~12 points to 87% in fiscal Q3 2026, while data-center revenue more than doubled QoQ to a record $11.5B, lifting the segment to a ~$45B annualized run-rate (vs. ~$6B a year ago). Companywide revenue hit ~$41.5B (+346% YoY) and non-GAAP EPS rose to a record $25.11, supported by AI-driven memory tightness and multi-year supply agreements; the company guided fiscal Q4 revenue near $50B at ~86% gross margin. Despite the results, the stock is still valued at <7x forward earnings, implying investors expect the commodity-style downturn eventually—so the setup is bullish on AI supply tightness but remains timing/cycle-dependent.

Analysis

This is less a “memory is hot” story than a supply-constrained pricing regime where the profit pool is shifting upstream. If HBM remains sold out into 2027, the economic winner is not just MU; it’s also the equipment stack that enables capacity adds, while AI server assemblers and hyperscalers absorb the cost inflation in their BOMs and may see model deployment economics tighten at the margin. The market is still pricing MU like a cyclical commodity name, but the mechanism here is closer to a quasi-duopoly bottleneck with multi-quarter contract visibility.

The main risk is timing: the current margin structure can stay elevated for several quarters, but the setup is vulnerable to a capex wave from Samsung/SK Hynix and eventual yield normalization. That reversal would likely show up first in spot/contract DRAM spreads and forward gross margin guidance, not in reported demand, so the next 1-3 months matter less than the 6-18 month supply response. If hyperscaler capex gets rephased or AI accelerator demand pauses, MU’s multiple can compress fast because the stock is still being valued as if the cycle has already peaked.

The contrarian view is that the market may be underestimating duration, not magnitude. A sub-7x forward multiple implies investors think earnings roll over quickly; if pricing merely plateaus rather than collapses, the equity can still rerate materially on steady revisions. The real tell will be whether management can keep calling out sold-out HBM capacity and tight DRAM/NAND conditions into the next two earnings prints; if that language softens, the thesis breaks.

For positioning, the cleanest expression is long MU on pullbacks versus a broad semis basket, with discipline around any guide-down in data-center gross margin or HBM pricing. The higher-conviction catalyst is the next earnings cycle, where revisions can still move materially if contract pricing remains intact. If you want optionality, buy MU calls into volatility compression only if the stock fails to re-rate despite another upward revision, because that would signal the market is still under-owned and under-exposed to duration.

Key Decisions for Investors

• Go long MU on any post-print consolidation or market-wide semis pullback; thesis is that the stock is still priced for a fast cycle turn, while earnings revisions likely remain upward for at least 1-2 quarters.
• Pair trade: long MU / short SOXX for 3-6 months to isolate memory scarcity versus the rest of semis; thesis breaks if MU gross margin steps down meaningfully or memory pricing rolls over.
• Use MU calls only as a tactical expression into the next earnings event if implied volatility compresses; target upside is a rerating toward peer multiples if management reiterates sold-out HBM supply.
• Set a hard alert on DRAM/HBM spot pricing and data-center gross margin guidance; a sustained decline or a guide below the current margin range would falsify the “longer-for-longer” thesis.
• Watch SK Hynix and Samsung capex commentary as the structural risk over 6-18 months; accelerating supply additions would be the earliest warning that the margin peak is becoming tradable rather than durable.