Micron Technology

▲ Bull Case

• ◆HBM as structurally re-rated semiconductor product with 5-year supply agreements and first-to-volume HBM4 positioning (unprecedented in memory history). If contracted pricing holds through FY27–FY28 cycle test (HBM GM floor 55–60%), MU has created a specialty-semi business inside a cyclical shell deserving 16–20x multiple on through-cycle EPS. With Bull FY28 EPS of $32 × 18–20x = $580–640/share.
• ◆CHIPS Act creates durable US-only cost moat: $6.1B+ grant plus 25% Investment Tax Credit on Idaho/NY capex reduces effective capex by ~24%—a permanent subsidy Korean competitors cannot access. As US AI infrastructure prioritizes domestic memory supply, MU's strategic positioning compounds over time.
• ◆Operating leverage and balance sheet inflection: Q2 FY26 net cash +$3.1B (vs. -$5B FY25), FCF margin 23.1%. Combination of $5.5B/qtr FCF + capex moderation from FY27 onward + buyback resumption supports 16–18x multiple. Through-cycle FCF $20B with WACC 11% = SOTP floor $487/share.

▼ Bear Case

• ◆Memory always cycles, and HBM is not exempt from supply laws. All three players adding ~50%+ wafer starts in 2026–2027; TechInsights forecasts H2 2027 oversupply inflection. 5-year HBM agreements untested through downcycle; historical precedent is renegotiation. If HBM4 ASPs decline 25–35% in FY2028, HBM GM falls to 35–40%, and FY28 EPS falls to ~$13. At cyclical-trough 10–12x multiple = $130–160/share—80% downside.
• ◆Samsung's re-entry compresses MU's share and pricing power. Samsung HBM3E yield recovery already underway; HBM share rising to ~35% in recent quarters. If Samsung HBM4 250K-wafer/month target materializes vs. MU's smaller ramp, MU's HBM share could compress to 15–17% by FY2028. Each 5pp share loss = ~$3–4B revenue, ~$2–3B net income.
• ◆Sell-side consensus at $657.69 is BELOW market $751 with 33 analysts rated Buy—unusual reluctant-bull pattern historically preceding correction. Analysts are bullish on business but formal targets below market price. Market is paying ahead of fundamentals at cycle peak.

“Primary debate: Is HBM a structural moat earning durable above-cycle margins, or just another DRAM product where Samsung re-enters and competes margins back to commodity levels? Bull thesis cites 12-Hi 3D stacking atomic-precision, 12–18 month customer qualification creating switching costs, 5-year supply agreements unprecedented in memory history, HBM TAM growing 40% CAGR, AI architecture progression requiring more HBM per GPU each generation, and HBM3E 30% lower power efficiency. Bear thesis counters: Samsung HBM3E yield recovery proves technology reproducible; memory has always broken pricing discipline at every cycle peak (FY18, FY22); contracts untested through downcycle with historical renegotiation precedent; all three players simultaneously add ~50%+ capacity; HBM still made on DRAM wafers so cost structure not fundamentally different. Step 19 verdict: HBM is partially structural—real but incomplete transformation. 5-year contracts will raise trough floor materially (35–45% GAAP GM vs. -9% historical) but will not eliminate cycle. Market at $751 likely overstates structural component by 30–50%. Empirical test is FY28 GM: >50% wins HBM-floor thesis; <40% loses it.”

• ◆Q3 FY26 earnings (June 2026): Beats $33.5B guide + Q4 raised guidance triggers consensus revision to $800–900; miss or Q4 caution triggers -15–20% correction
• ◆HBM4 volume ramp confirmation (Q3–Q4 FY26): Premium pricing sustained validates competitive advantage; volume shortfall signals mid-tier positioning
• ◆Q4 FY26 + FY27 capex guidance (Sept 2026): Capex moderation <$22B supports FCF expansion; ≥$25B signals margin pressure
• ◆First standard DRAM price weakness (H2 FY27): HBM holds while DRAM falls validates HBM-floor; both fall together proves cyclical correlation
• ◆Q1 FY28 gross margin (Dec 2027): THE CRITICAL TEST — >50% wins HBM-floor thesis, triggers major re-rate to $700+; <45% loses thesis, triggers bear/severe scenarios $200–350
• ◆Analyst price target revisions (June–August 2026): 33 analysts at Buy but avg target $657.69 below $751; reluctant-bull pattern historically precedes correction
• ◆Hyperscaler aggregate AI capex guidance (ongoing): >25% cumulative cut from AWS/Azure/GCP/Meta collapses HBM demand backlog
• ◆Nvidia Blackwell and next-gen adoption (H2 2026 into 2027): Ramp pace and HBM-per-GPU progression confirm or refute multi-year demand thesis

• ◆Memory cycle downturn with oversupply in FY27–FY28 (55–65% probability): All three players adding ~50%+ capacity; TechInsights forecasts H2 2027 oversupply. If HBM GM compresses to 35–40%, FY28 EPS ~$13. -$400 to -$600/share. CRITICAL.
• ◆Samsung and SK Hynix regain HBM share (35–45% probability): Samsung HBM3E yield recovery already underway; HBM share rising to ~35%. If Samsung 250K-wafer target materializes, MU share compresses to 15–17% by FY2028. -$120 to -$180/share.
• ◆HBM contracted pricing breaks or step-down provisions activate (20–25% probability): 5-year agreements untested through downcycle; historical precedent is renegotiation when pricing pressure builds. -$200 to -$300/share.
• ◆AI demand disappoints from hyperscaler capex cuts or Nvidia delays (15–25% probability): Hyperscaler capex includes speculative growth component. Recession or margin pressure triggers deferrals. -$300 to -$450/share.
• ◆Multiple compression without earnings change (30–40% probability): Stock de-rates from 22x forward to 12x on cycle-peak anxiety. -$300 to -$400/share.
• ◆US–China export control escalation (20–25% probability): Advanced-node memory or HBM restrictions to China/Taiwan. -$50 to -$80/share.
• ◆Fab execution risk on Idaho/NY CHIPS Act projects (20–25% probability): Cost overruns, construction delays, or yield challenges on greenfield fabs. -$30 to -$60/share.
• ◆CHIPS Act grant clawback or conditions tightened (5–10% probability): Political/budgetary pressure to reduce funding or impose stricter conditions. -$15 to -$30/share.