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 percentage points to 87% and data center revenue more than doubled QoQ to a record $11.5B. Companywide revenue hit a record ~$41.5B (+346% YoY) and non-GAAP EPS rose to a record $25.11, with memory pricing strength driven by AI-linked HBM supply tightness (2026 supply already sold out under multi-year agreements). Despite the strong print and a raised planned U.S. investment to $250B+ through 2035, the stock trades at <7x forward earnings, implying investors expect the cyclical downturn to return (guided Q4 revenue ~$50B at ~86% gross margin).

Analysis

This is less a one-quarter earnings story than a supply-control story. The market is still underwriting Micron like a cyclical parts supplier, but if HBM stays allocation-constrained through 2027, MU starts behaving more like a bottleneck tollbooth inside the AI stack: pricing power translates directly into estimate revisions, buyback capacity, and multiple expansion. The biggest second-order beneficiary is NVDA insofar as memory scarcity validates the premium AI infrastructure cycle, but tighter HBM can also become a hidden tax on the rest of the AI supply chain by forcing customers to pre-commit capital earlier and concentrate orders with the few qualified vendors.

Near term, momentum can keep carrying the shares because the next catalyst path is still upward revisions, not peak numbers. Over 1-3 months, the trade is whether the market stops anchoring to historical memory downturns and starts capitalizing forward earnings at a higher mid-cycle multiple. The risk is that investors are confusing visible backlog with durable economics; memory pricing can remain strong until it suddenly cannot, so the stock can re-rate before the fundamentals truly prove out.

The contrarian miss is timing, not direction. Consensus seems to accept that HBM is structurally tighter, but may be underestimating how quickly OEMs, hyperscalers, and competitors will redesign around cost or accelerate qualification of alternate supply once margins look too attractive. If DRAM/HBM pricing flattens, inventory days rise, or management stops sounding sold-out on 2027 supply, the thesis weakens fast; that is the key falsifier.

Key Decisions for Investors

• Long MU on pullbacks over the next 2-6 weeks; target a re-rating from sub-7x forward earnings toward 8-9x if data-center gross margin holds above the low-80s. Use a tight stop if the next guide implies margin normalization or demand deceleration.
• Prefer 6-12 month MU call spreads over outright stock for new exposure. The bull case is durable scarcity; the risk is a later cycle turn, so defined risk is the better expression.
• If you want a relative-value hedge, own MU against a broader semiconductor basket rather than owning the whole AI complex. The edge is in pricing power at the memory bottleneck, not in generic AI beta.
• Set an alert on any commentary that 2027 HBM supply is no longer effectively sold out. That would be the cleanest warning that the current scarcity premium is getting pulled forward into a peak.