For & Against

Claude View

What's Next

Micron reported Q2 FY2026 on March 18, 2026 – revenue of $23.9B crushed estimates by 24% and EPS of $12.07 beat by 39%. The next catalyst window is narrow and high-stakes.

What the market is watching most closely: Q3 FY2026 gross margin. Management guided to approximately 81%, which would be the highest in memory industry history. If Micron delivers, FY2027 estimates of $96 EPS look conservative and the stock re-rates higher. If margins compress even modestly from the 74.4% posted in Q2, the cycle-peak narrative takes hold immediately. DRAM contract pricing data in May will provide the earliest signal before the earnings print.

The secondary catalyst is HBM4 production readiness. Micron delivered HBM4 samples in FY2025 and volume production is underway for NVIDIA's Vera Rubin platform. SK hynix is the current HBM leader at 62% share; if Micron executes on time, it secures its position in the 2027 order book. A delay hands share to competitors and undermines the structural demand narrative.

For / Against / My View

For

HBM supply constraint is real and priced too cheaply. Micron trades at 8x FY2026 consensus EPS while HBM production is sold out through 2027 under long-term agreements. Adding HBM capacity requires 3x the wafer area of standard DRAM, physically limiting the supply response. Warren's analysis shows CMBU revenue went from $1.9B to $13.5B in two years with a 45% operating margin – this is not a commodity business anymore. At 8x forward earnings, the market is pricing a margin reversion that HBM supply physics may prevent for another 18-24 months.

The valuation gap versus peers is extreme and specific. Quant's peer scatter shows Micron at 49% revenue growth and 8x forward P/E versus NVIDIA at 15% growth and 40x, and TXN at 2% growth and 40x. The PEG ratio of 0.22 is not a generic "it's cheap" argument – it quantifies a market that prices Micron's earnings as entirely cyclical while pricing slower-growth semis as durable. If the new mid-cycle gross margin is 50% rather than the historical 35%, normalized EPS is approximately $35 and the stock trades at 13x – still cheap for a semiconductor oligopolist.

Management credibility through the cycle supports the thesis. Historian scored management 7/10 on credibility, noting they hit every technology promise: HBM3E on time, HBM4 samples delivered, $130M/quarter cost savings achieved during the downturn. The 30% dividend increase in Q2 FY2026 is a capital allocation signal from a team that suspended bonuses and cut 15% of headcount just two years ago – they know what a cycle peak feels like, and they are choosing to signal this is not one.

The only informed insider buyer is the most qualified person on the board. T. Mark Liu, former TSMC Executive Chairman with 30+ years in semiconductor manufacturing, voluntarily purchased $7.8M in Micron shares in January 2026, weeks after joining the board. Sherlock flagged this as significant – Liu understands memory economics at the deepest possible level, and he put personal capital to work while every executive was selling.

Against

Every named executive sold shares in the past six months – $124M total. Sherlock documented broad-based insider selling: CFO Murphy ($28M), CBO Sadana ($21M), CTO DeBoer ($18M), CPO Arnzen ($18M), CEO Mehrotra ($18M). The breadth matters more than any single transaction. No executive bought shares during the March 2026 dip to $322 – a 31% drawdown. If management truly believed margins would sustain above 70%, the absence of buying at that drawdown is a tell worth noting.

Capex is consuming all the cash at the top of the cycle. FY2025 FCF was just $1.7B on $37B revenue – a 4.5% FCF margin during a supercycle. FY2026 capex is guided to $20-25B, and CHIPS Act grants of $6.4B are spread over years. Micron is building fabs in Idaho, New York, India, and Singapore simultaneously. The Boise fab has already slipped from CY2025 to CY2027, and Clay, NY has not broken ground. If the cycle turns while these fabs are under construction, the operating leverage that produced $13.8B in quarterly profit works in reverse – FY2023's $5.8B annual loss happened on a smaller fixed-cost base.

Technology risk to memory demand is not theoretical. Google's TurboQuant algorithm demonstrated a 6x reduction in AI memory requirements in March 2026, briefly sending MU down 10%. Historian notes that inference – which uses less memory than training – is likely to become the dominant AI workload. If memory intensity per workload flattens even as AI adoption grows, the demand trajectory disappoints without any supply overshoot required. The market recovered from TurboQuant quickly, but the underlying risk has not been resolved.

Samsung and SK hynix are not standing still. Samsung's profits jumped 700% in early 2026 on memory demand, and the company is aggressively expanding HBM capacity. SK hynix is reportedly exploring a US listing, bringing direct competitive scrutiny and capital access. Warren notes that the moat in memory depends on maintaining technology parity at each node transition – Samsung fell behind on HBM3E yields, but one generation does not make a permanent advantage. Chinese entrants CXMT and YMTC continue building capacity with government backing, adding longer-term pricing pressure on conventional DRAM and NAND where Micron still generates the majority of revenue.

My View

I'd lean cautiously bullish, but with less conviction than the $520 average analyst target implies. The For side carries more weight because the HBM supply constraint is a physical reality, not a narrative – you cannot wish 3x wafer area into existence in twelve months – and the valuation at 8x forward earnings provides genuine margin of safety even if estimates get trimmed 20-30%. What tips the scale is Liu's $7.8M purchase combined with management's clean execution on every technology milestone; these are people who have navigated memory cycles before and are choosing to invest and raise dividends rather than hoard cash. The Against side's strongest point is the broad insider selling, which introduces real ambiguity about how durable management believes current margins are – $124M in executive sales across six months is not a rounding error. The condition that would flip me cautious is Q3 FY2026 gross margin coming in below 70%; that would confirm the cycle is turning faster than the bull case assumes, and at $466 there is no room for a margin disappointment in a business with this much operating leverage.