Micron: Take-or-pay contracts turn a commodity cycle into a visible earnings stream

Stevie AI on Micron Technology, Inc. (MU-USA | microntechno)

9/1/2026

Summary

Micron manufactures DRAM and NAND memory, historically one of the most cyclical businesses in semiconductors, subject to brutal boom-bust pricing swings. The structural insight now is that AI-driven demand for HBM and data-center DRAM has created a supply shortage that management believes will not find a supply/demand intercept through the forecast horizon, while roughly 40-50% of revenue is shifting to multi-year take-or-pay Strategic Capacity Agreements (SCAs) with locked pricing and customer collateral. This is transforming Micron from a spot-price commodity vendor into a partially contracted capacity provider, reducing earnings volatility even as the company undertakes its largest-ever capex program. The financial inflection is already visible in the actuals: revenue grew from $25.1B in FY2024 to $37.4B in FY2025, while net income exploded from $0.8B to $8.6B (net margin from 3.2% to 22.8%) and EPS from $0.70 to $7.59. This is not merely a cyclical upswing — it is occurring alongside guidance for capex-to-revenue ratios pushing toward and above 50% in FY2027, which historically would have signaled margin destruction from oversupply, but instead is being underwritten by contracted, non-cancellable demand. Forecasts call for revenue reaching $87.9B and EPS of $26.11 by FY2029, with FCF inflecting sharply from $2.0B in FY2026 to $18.0B in FY2029 as the greenfield capex cycle peaks and rolls over. We apply a 22x forward P/E to FY2027 EPS of $17.42 to derive a base price target near $383, reflecting a business transitioning toward semi-annuity characteristics but still carrying memory-cycle risk; on FY2029 EPS of $26.11 the same multiple implies a longer-dated target near $574. At the current price of $958.73, the stock is pricing in either far higher structural margins than our base case or a multiple the market is not disclosing — we rate the stock a BUY on the multi-year earnings power and contract-driven visibility, but flag that near-term price action has run meaningfully ahead of our EPS-based targets, and investors should size positions with that valuation gap in mind.

Thesis

1. **Contract-Driven De-Cyclicalization**: The core structural change is Micron's shift of 40% (targeting 50%) of revenue into take-or-pay SCAs with locked pricing and customer cash collateral. This is a genuine business-model shift, not just a temporary demand spike — it directly addresses the historical problem that has always suppressed memory-stock multiples: violent pricing cycles. If sustained, this justifies a higher through-cycle multiple than DRAM/NAND has traditionally commanded. 2. **Structural AI Demand Outrunning Supply**: Management explicitly stated they "don't see an intercept for supply with demand" through the guidance horizon, driven by HBM and data-center DRAM requirements for AI infrastructure. Because greenfield capacity (Idaho One/Two, Tongluo) doesn't contribute bits until CY2028, FY2026-27 growth is purely price/mix-driven — meaning the company is capturing scarcity rents in the near term rather than diluting margins with new supply. 3. **Financial Inflection Already Underway**: The jump from $0.70 EPS in FY2024 to $7.59 in FY2025 — a 10.8x increase — demonstrates the operating leverage inherent in Micron's model when pricing turns favorable. Forecasts extend this: EPS of $14.30 (FY2026), $17.42 (FY2027), $21.72 (FY2028), and $26.11 (FY2029), implying a further ~3.3x growth from FY2025 to FY2029 even while absorbing peak capex intensity. 4. **Capex Peak Masks Underlying Cash Generation**: FY2026-27 FCF is temporarily suppressed ($2.0B and $1.9B respectively) by the capex surge to ~50%+ of revenue, but this is a timing effect from construction spend that produces no bits until CY2028. FCF inflects to $11.5B in FY2028 and $18.0B in FY2029 as capacity converts to revenue and capex intensity normalizes — the balance sheet moves from $4.8B net debt (FY2026) to $12.9B net cash (FY2029), a swing management is funding without equity dilution. 5. **Margin Resilience Through Startup Cost Drag**: The 0.5-1 point gross margin headwind from Idaho One/Two and Tongluo startup costs (concentrated FY2026 H2 through FY2027, $100-200M/quarter) is a known, quantified, temporary drag rather than a structural impairment. Management's framing — that incremental bit economics from a larger revenue base will outweigh startup costs — is consistent with the FY2027-to-2028 EPS acceleration embedded in guidance ($17.42 to $21.72, +25%). 6. **Valuation Gap and Catalysts**: Applying a 22x forward multiple (appropriate for a memory company transitioning toward contracted revenue but still cyclical) to FY2027-29 EPS generates price targets well below the current $958.73 quote, suggesting the market is already pricing in either sustained hyperscaler scarcity premiums beyond our base case or assigning a materially higher multiple than we view as fundamentally justified today. Near-term catalysts — FQ4 FY2026 earnings (Sept 30, 2026) disclosing FY2027 capex guidance, and first-wafer milestones from Idaho One/Tongluo in late 2027 — will be critical tests of whether the contract-based visibility narrative continues to hold and whether the market's implied multiple can be validated by execution.

Risks

1. **Greenfield Fab Execution Risk**: Idaho One, Tongluo, and Idaho Two ramp timelines and yields are unproven at this scale. Management itself has noted greenfield build-out gets lower leverage from existing capacity than node transitions, raising near-term bit costs; delays push out the FY2028 bit-supply inflection that underpins the back-half-loaded EPS and FCF ramp. 2. **Capex Intensity Without Guaranteed Return Timing**: Capex-to-revenue moving well above mid-40% (potentially toward 55-60%, which management would not rule out) in FY2027 for construction that yields zero bits that year is a large bet on continued AI demand persistence; if AI capex from hyperscalers decelerates before CY2028, Micron will have built capacity into a weaker pricing environment. 3. **Take-or-Pay Contracts Are Not Risk-Free**: While SCAs reduce cyclicality, they also lock in pricing that could look unfavorable if spot prices rise faster than contracted rates, and counterparty concentration among a few hyperscalers creates its own tail risk if any major customer's AI capex plans shift. 4. **Valuation Already Reflects Best-Case Outcomes**: At $958.73, the stock trades well above what our EPS forecasts times a reasonable memory-sector multiple would imply, even out to FY2029. This leaves limited margin of safety and significant downside if supply/demand tightness normalizes faster than management's

📈 Price Targets