Earnings call transcript: Micron tops Q4 2026 estimates as AI demand stays hot

Micron reported fiscal Q4 2026 adjusted EPS of $33.42 and revenue of $54.23B, beating consensus by 7.25% and 7.49%, respectively. Management expects AI-led memory tightness to persist through 2028, with more than 75% of fiscal 2027 shipments already covered by supply and capacity agreements and margins expected to expand through the remainder of fiscal 2027. HBM demand, including a custom HBM4E/NVHBM effort with NVIDIA, supports pricing and long-term visibility, though double-digit near-term cost growth, fab startup expenses and elevated construction CapEx remain offsetting risks. Shares rose only 0.16% after hours, indicating much of the strong result had been priced in.

Analysis

The muted post-print response implies MU’s valuation now requires forward estimate revisions rather than another beat. The key earnings-power question is not spot HBM demand but contract economics: price floors protect utilization, while ceilings and periodic repricing can cap incremental upside if memory inflation persists. Near-term gross-margin expansion can therefore continue, but the upside skew is likely smaller than headline supply-tightness rhetoric suggests.

The more durable implication is that memory is becoming a system-level gating item for AI deployments, raising the value of DRAM content per accelerator and per CPU server. This supports MU and Hynix (000660 KS), but also creates a second-order beneficiary in NAND/enterprise storage—SanDisk (SNDK) and WDC—if constrained DRAM forces customers to optimize architectures with more local storage. NVDA benefits from higher-value system configurations, although custom-memory designs may shift some economics toward NVDA and TSMC rather than accrue entirely to MU.

Over 1-3 months, the catalyst is consensus revising FY27/FY28 pricing, mix, and margin assumptions; the risk is that the stock already discounts a multi-year supercycle. Over 6-18 months, construction spending is economically ambiguous: delayed wafer additions preserve scarcity now, but a synchronized capacity-equipment ramp can turn the cycle once demand normalizes. The thesis is falsified by declining contract-price benchmarks, an increase in industry supply-growth guidance, or evidence that AI customers reduce memory content rather than absorb available supply.

Contrarian view: the market may be underestimating the demand elasticity created by memory-constrained AI racks—additional supply could be absorbed through higher memory configurations rather than trigger an immediate price collapse. But investors should not treat customer agreements as equivalent to fully uncapped take-or-pay revenue; disclosure on minimum volumes, termination rights, and price-reset terms is the critical missing diligence item.

Key Decisions for Investors

• Maintain a tactical long MU for the next 1-3 months only if FY27 consensus EPS and gross-margin estimates begin moving higher; use a 12-15% trailing stop from entry, as the rerating case is vulnerable to a modest guidance shortfall after a large prior run.
• Express the memory-content thesis through a pair trade: long MU / short NVDA in equal dollar beta-adjusted amounts for 3-6 months. MU has more direct sensitivity to memory-price and mix upside, while NVDA is more exposed to accelerator-demand expectations already embedded in valuation; exit if NVDA-led AI capex guidance materially accelerates or MU contract pricing weakens.
• Build a watchlist long in SNDK or WDC rather than immediately chasing: enter only after evidence of enterprise SSD price increases or upward revisions to NAND contract pricing. This is the cleaner second-order trade if AI infrastructure demand broadens beyond HBM, with a 6-12 month horizon.
• Avoid adding broad semiconductor-equipment longs solely on MU’s construction plans. ASML, LRCX, and AMAT benefit only when clean-room projects convert into tool orders; monitor order backlog and 2027 capex guidance, since construction-heavy spending delays their revenue realization.
• For downside hedging on existing MU exposure, buy 6-month put spreads rather than outright puts; the relevant risk window is the next two earnings cycles, when double-digit cost growth, startup expenses, or weaker price realization could compress the premium multiple.