$MU #MU
OVERVIEW
Memory/storage semiconductor giant — DRAM, NAND, HBM. One of three major DRAM makers globally (with SK Hynix, Samsung). 52W range $61.54–$471.34 (ATH March 18). ~$376, down 20% from ATH post-earnings. Cap ~$431B. P/E 18.67x. Up 350%+ over the past year. Dividend speculative — the most important AI infrastructure stock outside Nvidia.
FQ2 2026 (March 18): Revenue $23.86B — nearly tripled YoY (+196%), crushed $20.07B consensus. Adj EPS $12.20 (beat $9.31 by 32%). Net income $13.79B. EBITDA $37B, 49% margin. Gross margin guided ~81% for Q3. Q3 guide: $33.5B rev (vs $24.3B consensus), $19.15 EPS (vs $12.05) — exceeds Micron’s entire FY2024 revenue in one quarter. Cloud memory rev up 160%+ to $7.75B. Mobile/client tripled to $7.71B. DRAM prices up mid-60% YoY, NAND up high-70%. FY2026 capex raised above $25B; construction capex $10B+ in FY2027 (Idaho + NY fabs).
BULL CASE
AI memory supercycle — Micron at the epicenter. HBM is the AI bottleneck, not GPUs. Every Nvidia generation (Blackwell Ultra, Vera Rubin, Feynman) packs more memory, creating structural shortage. Micron’s 2026 HBM capacity 100% sold out under binding long-term contracts. HBM4 production begun for Vera Rubin. HBM4e ramps 2027. Custom base dies co-designed with customers turn memory from commodity into bespoke product.
The “Memory Fortress” thesis: boom-bust cycles replaced by sustained high-margin AI-driven growth. 81% gross margins unheard of — historically peaks at 40–50%. Bloomberg: HBM market grows 42% annually through 2033. AI data centers expected to consume 50%+ of memory shipments this year. Deloitte: AI data center workloads tripling/quadrupling annually through 2030.
32 analysts, consensus Strong Buy. Avg PT $443, high $525 (RBC), Wedbush $500, Bernstein $510. At $376: 18–40% upside. At 18.67x P/E on exploding earnings, valuation reasonable — if Q3 hits $19.15 EPS, forward P/E drops to ~10x annualized. CHIPS Act supports Idaho/NY buildouts ($100B NY campus). Only top-10 U.S. tech company UP in 2026.
BEAR CASE
Stock dropped ~15% after the best quarter in company history — market pricing perfection. Capex above $25B with $10B+ construction in FY2027 is enormous — any AI spending slowdown = overbuilding risk. Google’s TurboQuant (cutting LLM memory needs up to 6x) is a direct demand threat. SK Hynix/Samsung aggressively ramping competing HBM — supply shortage won’t last forever.
Memory is cyclical. DRAM prices up 60%+ YoY have never been sustainable in 50 years. When HBM supply catches demand (likely 2027–28), pricing power erodes fast. Stock went $61→$471 in one year (665%). A normal 30% correction = $330. Morningstar: trading at 498% premium to fair value. 11.6% dividend yield looks generous but payout is small ($0.15/quarter) — yield high because stock ran fast. PC/smartphone memory neglected as capacity shifts to HBM — could create margin compression in non-AI segments once HBM normalizes. China restrictions limit access to second-largest market.
PLAN
Fundamentally different from everything else in portfolio. $431B blue-chip semi in a once-in-a-generation demand cycle, not speculative micro-cap. Post-earnings selloff from $471→$376 (-20%) is the entry if you believe the AI memory supercycle has legs.
Invalidation: Sustained break below $300 = HBM supply glut, AI capex pullback, or memory cycle peaking. At that level the 665% rally unwinds and “this time is different” thesis fails.
1
1
178
