Will Micron's average gross margin for FQ3-FQ4 FY2026 remain above 45%?

The 45% threshold is materially below current trajectory. FQ1 delivered 56.8% and FQ2 is guided at 68% — both well above the threshold. For the FQ3-FQ4 average to fall below 45%, margins would need to compress ~23pp from the FQ2 guided level within two quarters. While memory margins are famously cyclical (-9.1% to 68% in 3 years), historical downcycles show 10-15pp quarterly declines at worst, not 23pp in one step. DRAM pricing is surging 55-60% QoQ in Q1 CY2026, and management expects continued expansion. The main risks — Samsung HBM competition (>500bps compression) and Idaho fab startup costs — would erode margins 5-8pp combined, still leaving margins well above 45%. However, 75% of revenue reprices quarterly, creating genuine rapid-adjustment risk if demand weakens sharply in the June-August window.

The structural case for margins holding above 45% through H2 FY2026 is strong but not unassailable. FQ2 at 68% provides a massive buffer — even a 15pp sequential decline to ~53% in FQ3 followed by another 10pp decline to ~43% in FQ4 would average 48%, still above threshold. The risk scenario requires BOTH a sharp pricing reversal AND cost inflation simultaneously. Samsung HBM pricing competition is the most credible threat at >500bps, but even aggressive Samsung entry is unlikely to compress Micron's blended gross margin by more than 5-8pp in a single quarter given HBM is only ~20-25% of total revenue. The NARRATIVE_REALITY_GAP = DIVERGING signal from Myth Meter and heavy insider selling give me pause — those who know the business best are selling aggressively. But even a moderately pessimistic scenario puts FQ3-FQ4 average around 50%, well above threshold. The 75% quarterly repricing creates a tail risk of rapid repricing but the current pricing environment is strongly favorable.

I want to steelman the NO case more carefully. Micron's gross margin history shows it can swing 30+ percentage points in a year (22% FY2024 average vs 68% FQ2 FY2026). The 5-year average is 22%, meaning current levels represent a 3-sigma event. The Revenue Revealer rated durability as 'weak CONDITIONAL near FRAGILE boundary' — margin sustainability is contingent on demand conditions that could shift. If DRAM contract pricing turns negative in CY2026 H2 (which TrendForce has not forecast but is possible if AI capex moderates), combined with rising Idaho fab costs and NAND drag, a scenario where FQ3 drops to 52% and FQ4 drops to 38% (average 45%) is within the realm of possibility. But this requires multiple negative catalysts converging simultaneously within a 6-month window, which is improbable given current demand strength. The staleness note (101-day-old fundamentals) adds uncertainty — FQ2 results on March 18 will be critical. I weight this at 0.68 rather than higher due to the extreme historical volatility and the unresolved debate about whether margins are structural or cyclical.

This is a high-threshold question with a massive margin of safety. At 68% guided for FQ2, margins would need to collapse by more than 20pp on average across two quarters to breach 45%. DRAM pricing is in a strong upcycle with 55-60% QoQ increases. Management explicitly expects continued expansion. The memory oligopoly (Samsung/SK Hynix/Micron) has pricing discipline. Even the worst-case Samsung HBM competition scenario only shaves 5pp. The only realistic NO scenario requires a complete demand collapse in AI/data center spending — which contradicts every forward indicator in the context. Insider selling is irrelevant to the specific 45% threshold question; they're selling because the stock is high, not because margins are about to halve.

The 45% threshold makes this a relatively easy YES case in the near term. Current margins are 56.8% actual and 68% guided — both massively above threshold. The question is whether something catastrophic happens in the June-August FQ4 window. Historical precedent shows Micron margins can drop 15-20pp in a single quarter during downturn transitions (FY2022-2023). But those transitions came after extended overbuild periods; the current cycle is characterized by supply discipline and AI-driven demand. The 75% quarterly repricing is the key risk vector — if contract pricing turns, margins adjust fast. But for average FQ3+FQ4 to drop below 45%, you'd need something like FQ3 at 50% and FQ4 at 39%, which requires a pricing collapse that isn't supported by current supply-demand dynamics. I'm at 0.73 rather than 0.80+ because the NARRATIVE_REALITY_GAP = DIVERGING finding means the committee itself identified this cycle as potentially over-extended.

The math strongly favors YES. If FQ3 margin is 58% (a 10pp drop from FQ2's guided 68%) and FQ4 is 48% (another 10pp drop), the average is 53% — well above 45%. Sequential 10pp declines would represent a severe downturn, yet still wouldn't breach the threshold. To get below 45% average, you'd need FQ3 at ~50% AND FQ4 at ~40%, requiring sustained 14-15pp quarterly compression. Memory downcycles of this severity take 3-4 quarters to develop, not 2. The Gravy Gauge's finding that memory shortage is deepening further supports pricing through the near term. Idaho fab startup costs and NAND drag are headwinds but measured in low-single-digit percentage points, not the 20+pp needed to breach threshold.

FQ2 guided at 68%, FQ1 actual at 56.8%. Even a dramatic margin decline would keep the FQ3-FQ4 average above 45%. DRAM pricing momentum is strong. The 45% threshold requires a margin collapse that contradicts every near-term indicator in the analysis context.

Strong YES case given the 23pp buffer between guided FQ2 margins and the 45% threshold. However, Micron's extreme historical cyclicality (-9.1% to 68%) means rapid reversals are possible. The 75% quarterly repricing and Samsung HBM competition create downside vectors. Insider selling at peak adds uncertainty. Still, the magnitude of compression needed makes NO unlikely in just 2 quarters.

The base case is clear: margins well above 45% through H2 FY2026 given current trajectory and demand environment. The bear case (Samsung pricing war + demand shock + cost inflation) is possible but requires multiple negatives simultaneously. Memory oligopoly pricing discipline and AI demand tailwinds support YES.