Concerns & Risks -- 8/10

MU trades at ~10.5x FY2026E earnings -- optically cheap for a semiconductor company generating record revenue and margins. But memory is deeply cyclical: at prior peaks (FY2018, FY2022), MU traded at 5-8x peak P/E before correcting 50%+. The central question is whether AI-driven structural demand, HBM supply constraints, and the new Strategic Customer Agreement model genuinely break the historical boom-bust pattern, or if 75%+ gross margins and $33.5B quarterly revenue simply mark the top of the greatest memory cycle ever. Six concrete catalysts support the bull case, but seven material risks -- led by 40 years of cyclicality -- temper the outlook. Weight: 15%
Forward P/E (FY2026E)
~10.5x
Optically cheap but peak-cycle earnings
FY2027E P/E
~7.6x
Consensus $48 EPS -- assumes continued growth
EV / EBITDA
~6.3x
Low, but cyclical peak EBITDA (~$65B est)
FCF Yield
~5%+
Est ~$20B+ FY2026E FCF; capex intensive
IMPORTANT CAVEAT: P/E on cyclical peak earnings is misleading for memory stocks. At the FY2018 peak, MU traded at ~4-5x earnings before a 50%+ drawdown. At the FY2022 peak (~$8.35 EPS), MU fell from ~$98 to ~$49 despite appearing cheap on trailing multiples. EPS swung from +$8.35 to -$4.45 (a 152% reversal) in just four quarters. The current 10.5x on consensus FY2026 EPS of ~$35 must be evaluated against the real possibility of earnings normalization -- not taken at face value.
Valuation Snapshot
Metric Value Commentary
Price $366.24 22% below 52-week high of $471.34
Market Cap $413B
P/E (FY2026E) 10.5x Cheap for semis, but peak-cycle earnings
P/E (FY2027E) 7.6x Consensus $48 EPS -- assumes cycle extends
EV/EBITDA (FY2026E) ~6.3x Low, but on ~$65B cyclical peak EBITDA
EV/EBITDA (FY2025A) ~21.6x More normal period -- shows cycle distortion
P/B (est) ~6x Well above historical avg of ~2-3x
FCF Yield (FY2026E) ~5%+ ~$20B+ est FCF; strong but capex intensive
Key Catalysts (Bull Case)
# Catalyst Timing Impact
1 FQ3 2026 earnings June 2026 HIGH
Guide for $33.5B revenue and $19.15 EPS. Another massive beat could drive re-rating. FQ3 revenue alone exceeds entire FY2024 annual revenue ($25.1B).
2 SCA expansion 2026-2027 HIGH
More 5-year Strategic Customer Agreements could structurally de-risk the business model and justify higher multiples. First SCA signed in FQ2 2026.
3 HBM4E ramp CY2027 HIGH
Customizable base logic die offers differentiation and higher margins vs commodity HBM. For NVIDIA Vera Rubin platform.
4 NAND inflection H2 2026 MEDIUM
Data center SSD revenue more than doubled sequentially in FQ2. NAND supply tightness could make it a meaningful profit contributor beyond just DRAM.
5 US manufacturing advantage Ongoing MEDIUM
Only US-based memory manufacturer. CHIPS Act support and customer preference for domestic supply chain in an era of geopolitical tension.
6 Robotics / edge AI Long-term EMERGING
CEO highlighted humanoid robotics as a "20-year growth vector." Early stage but enormous long-term TAM expansion for memory content per device.
Key Risks (Bear Case)
# Risk Severity Detail / Mitigant
1 Cyclicality -- the #1 risk CRITICAL Memory has NEVER sustained peak margins for more than 4-6 quarters historically. From FY2022 peak ($8.35 EPS) to FY2023 trough (-$4.45 EPS) was a 152% swing. 40 years of boom-bust cycles. Mitigant: HBM trade ratio, greenfield lead times, and SCAs may be structurally different.
2 Capex over-investment HIGH FY2026 capex guided above $25B; FY2027 to step up "meaningfully" with $10B+ more in construction. If demand slows, this becomes a massive drag on FCF. Memory companies have a long history of over-investing at peaks.
3 China risk (dual) HIGH (a) China sales: Mainland China + HK = ~10% of FY2025 revenue; export restrictions limit advanced product sales. (b) Taiwan manufacturing: significant fab operations in Taiwan create geopolitical exposure. Tongluo site acquisition adds capacity but also risk.
4 Samsung competition MEDIUM Samsung investing heavily to regain HBM share (qualified at major customers for HBM3E). A Samsung comeback in HBM could pressure pricing power. Samsung holds ~33% DRAM share vs Micron at ~26%.
5 NAND oversupply risk MEDIUM NAND cycles are historically shorter and sharper than DRAM. Currently tight, but NAND could reverse faster. Micron holds only ~14% NAND share vs ~33% for Samsung.
6 Customer concentration MEDIUM Data center/AI customers (hyperscalers + NVIDIA) are an increasing share of revenue. CMBU + CDBU reached 56% of revenue in FQ2. Concentration risk if any single customer shifts supplier.
7 Tariffs / trade policy MEDIUM Management explicitly excludes tariff impacts from guidance. Any escalation of US-China trade tensions or broader tariff actions is not reflected in numbers. Could disrupt both supply chains and end-market demand.
Assessment: Is This Cycle Structurally Different?
Factor Bull Case Bear Case
HBM supply constraint 3:1+ trade ratio permanently constrains DRAM bit supply per wafer. Cannot just add capacity. Trade ratio advantage erodes as all three players scale HBM capacity on new fabs.
Greenfield fab lead times 3+ years to build a new fab limits supply response to demand surge. All three players building aggressively NOW. Supply arrives in CY2028-2029 and could overshoot.
AI demand durability AI is secular, not cyclical. Training + inference demand compounds. Not a one-time build. AI capex could decelerate if ROI disappoints. Hyperscaler spending is historically lumpy.
SCAs and pricing 5-year agreements with specific commitments may provide margin floors unprecedented in memory. No details disclosed on downside protection terms. Could be weaker than implied. Customers have leverage in a downturn.
Verdict: The structural arguments are more compelling this cycle than any prior memory peak. But "this time is different" are the four most dangerous words in investing. The prudent approach is to assign meaningful probability to both outcomes: a genuinely extended super-cycle (bull) and a historically normal reversion within 12-18 months (bear). The 10.5x P/E provides some cushion, but not enough to survive a 50%+ earnings decline if the cycle turns.
Score Rationale

Score of 8/10 reflects a favorable but cautious risk-reward assessment. The valuation at ~10.5x FY2026E earnings and ~7.6x FY2027E is optically cheap for a company guiding to $33.5B in quarterly revenue with 81% gross margins. EV/EBITDA of ~6.3x and FCF yield above 5% are attractive on current numbers. The stock at $366 sits 22% below its 52-week high of $471, suggesting meaningful skepticism is already priced in despite record-shattering results.

Six concrete catalysts support the bull case: FQ3 earnings (another potential blowout), SCA expansion (structural de-risking), HBM4E (differentiation), NAND inflection (broadening the earnings base), US manufacturing advantage (geopolitical tailwind), and robotics/edge AI (long-term TAM expansion). The Street consensus of ~$530 average price target implies ~45% upside.

However, seven material risks prevent a higher score. Cyclicality remains the existential concern -- memory has never sustained peak margins beyond 4-6 quarters in 40 years. The aggressive capex trajectory ($25B+ in FY2026, stepping up meaningfully in FY2027) echoes prior cycle peaks. China risk is dual-pronged (export restrictions and Taiwan fab exposure). Samsung is actively working to regain HBM share. NAND could reverse faster than DRAM. Customer concentration is rising. And tariff impacts are explicitly excluded from guidance.

The score is not lower because the structural arguments for this cycle are genuinely stronger than any prior memory peak: HBM trade ratios constrain supply, greenfield fabs take 3+ years, AI demand appears secular, and SCAs are a novel mechanism for margin stability. Management has earned credibility through consecutive massive guidance beats and disciplined navigation of the FY2023 downturn. For investors who believe the structural thesis, 10.5x on accelerating earnings with 45% Street upside is a compelling entry point -- but position sizing must reflect the cyclical tail risk.

Data sourced from Daloopa, Stock Analysis, and WallStreetZen.