Cyclical trade, ~18-month horizon. DRAM supply squeeze is real — 3 producers, AI-driven demand outpacing supply, prices surging to record levels. Earnings will be enormous through FY2027. But memory cycles always turn. New capacity is already ramping, with major additions from all three producers contributing meaningful volume by H2 2027.¹ Discipline will eventually break, and pricing will normalize. Capture the earnings surge, exit before the turn.
Sell window: Late CY2026 to early CY2027.
| FY2025A | FY2026E | FY2027E (Peak) | FY2028E (Soft) | FY2028E (Hard) | |
|---|---|---|---|---|---|
| Revenue | $37.4B | $86B | $105B | $75B | $55B |
| YoY Growth | +49% | +130% | +22% | -29% | -48% |
| Gross Margin | 39.8% | 69.6% | 68.6% | 56% | 45% |
| Net Income | $8.5B | $47.3B | $57.2B | $30.7B | $16.1B |
| EPS | $7.59 | $42 | $50 | $27 | $14 |
| P/E at $420 | 55.3x | 10.0x | 8.4x | 15.6x | 30.0x |
| Q1 (Actual) | Q2 (Guided) | Q3 (Est.) | Q4 (Est., 14-wk) | FY2026E | |
|---|---|---|---|---|---|
| Revenue | $13.6B | $18.7B | $23.2B | $30.0B | $86B |
| COGS | $6.0B | $6.0B | $6.5B | $7.5B | $26.0B |
| Gross Profit | $7.6B | $12.7B | $16.7B | $22.5B | $59.5B |
| GM% | 56% | 68% | 72% | 75% | 69.6% |
| OpEx | $1.3B | $1.4B | $1.4B | $1.5B | $5.8B |
| Operating Income | $6.3B | $11.3B | $15.3B | $21.0B | $53.7B |
Revenue more than doubles (+$49B). COGS grows 16% (+$3.5B) — anchored by D&A (approximately 20% of beginning PP&E, consistent with FY2024: 20.1% and FY2025: 21.0%) plus variable manufacturing costs that scale with bit shipments (+20% YoY, management guided, with minimal cost-per-bit reduction as HBM packaging intensity offsets front-end improvements). COGS ramps in H2 as bit shipments accelerate: $6.0B (Q1-Q2) → $6.5B (Q3) → $7.5B (Q4, 14-week). Revenue follows from COGS and the gross margin target: at 72% GM and $6.5B COGS, Q3 revenue is ~$23B; at 75% and $7.5B, Q4 is $30B.
The margin expansion is mechanical. COGS is anchored by D&A (fixed, locked in by prior CapEx) while DRAM ASPs surge. Retail DRAM pricing confirms the trajectory — DDR4/DDR5 prices surged to new highs in late 2025/early 2026.²
The 75% ceiling is a modeling assumption — NVIDIA-like margins for a supply-constrained oligopoly with fixed costs. At 75% GM and ~$7.5B COGS (14-week Q4), quarterly revenue reaches ~$30B.
The demand-supply gap supports this trajectory: hyperscaler AI CapEx is growing ~70% YoY ($383B FY2025 → ~$650B FY2026 guided) while memory industry capacity investment is ~40% of PP&E.³ Industry HBM bit demand is growing ~40% annually through CY2025-2027 (SemiAnalysis estimate), while Micron’s bit supply grows only ~18% in CY2026.⁶ Despite the heavy investment, new capacity takes 2-3 years to come online, and HBM production requires ~3x the wafer capacity of conventional DRAM, compounding the supply constraint.⁵
| FY2026E | FY2027E | |
|---|---|---|
| Revenue | $86B | $105B |
| COGS | $26.0B | $33.0B |
| Gross Margin | 69.6% | 68.6% |
| OpEx | $5.8B | $6.3B |
| Operating Income | $53.7B | $65.7B |
| EPS | $42 | $50 |
GM starts FY2027 near 75% (momentum from Q4 FY2026) and compresses through the year toward ~62% as new capacity begins contributing.¹ Average ~69%. Supply squeeze holds through H1 (calendar late 2026); H2 (calendar early-mid 2027) softens.
Revenue growth of +22% ($86B → $105B) is driven by volume: bit shipments grow ~30% as new fabs ramp (M15X, Yongin Phase 1, Idaho fab #1), while ASPs moderate in H2 as supply catches up. COGS grows from $26B to $33B — D&A steps up $2.2B (per the PP&E schedule) and variable costs scale with 30% bit growth.
This is likely the earnings peak. The stock probably peaks before this number is reported.
| Soft Landing | Hard Landing | |
|---|---|---|
| Revenue | $75B | $55B |
| COGS | $33B | $30B |
| Gross Margin | 56% | 45% |
| EPS | $27 | $14 |
D&A is locked in at $13.8B (20% of $68.8B PP&E). Variable costs flex modestly with production cuts but remain elevated from the FY2027 ramp. The higher cost base — built for $105B revenue — becomes a drag when revenue falls.
| FY2025A | FY2026E | FY2027E | FY2028E | |
|---|---|---|---|---|
| Beginning PP&E | $39.7B | $46.6B | $57.3B | $68.8B |
| CapEx | $15.9B | $20.0B | $23.0B | ~$20B |
| D&A (20% of beg. PP&E) | $8.4B | $9.3B | $11.5B | $13.8B |
| Ending PP&E | $46.6B | $57.3B | $68.8B | $75.0B |
Near-term CapEx of ~$20B/yr is consistent with Micron’s $200B US investment plan announced February 2026 — $150B for manufacturing and $50B for R&D over 20 years, front-loaded for the AI/HBM buildout.⁴
D&A grows from $8.4B to $13.8B over 3 years — a $5.4B increase in fixed cost baked into the P&L regardless of revenue. This is the mechanism that eventually compresses margins: revenue is cyclical, but D&A only goes up.
| Period | Annualized | Context |
|---|---|---|
| Prior cycle peak (Q3 FY2022) | 0.93x | Top of 2021-2022 memory boom |
| Cycle trough (Q2 FY2023) | 0.37x | Revenue down 57%, PP&E flat |
| Current (Q1 FY2026) | 1.11x | 20% above prior peak — unprecedented |
| Q2 FY2026 (guided) | ~1.43x | 54% above prior peak |
The current pricing environment has no historical precedent. The forecast requires this ratio to hold. If it mean-reverts toward the prior peak (0.93x) against growing PP&E, revenue falls faster than the soft-landing scenario.
Every forecast input traces to management guidance or is an explicit modeling choice.
| Assumption | Value | Source |
|---|---|---|
| Q1 FY2026 revenue | $13.6B | Actual (reported) |
| Q1 FY2026 COGS | ~$6.0B | Actual |
| Q1 FY2026 gross margin | 56% | Actual |
| Q2 FY2026 revenue | $18.7B ±$400M | Guided (Q1 earnings call, Dec 17, 2025) |
| Q2 FY2026 gross margin | 68% ±100bps | Guided |
| Q2 FY2026 OpEx | $1.38B ±$20M | Guided |
| Q3 FY2026 OpEx | “Flattish” vs Q2 | Guided |
| Q4 FY2026 OpEx | “Up” from Q3 + extra week | Guided (14-week quarter) |
| FY2026 CapEx | $20B | Guided |
| FY2027 CapEx | “Up versus 2026” | Guided (CFO) — we model $23B |
| PP&E (FY2025 ending) | $46.6B | 10-K |
| D&A / beginning PP&E | ~20% | FY2024 actual 20.1%, FY2025 actual 21.0% |
| Tax rate | ~12% | Singapore incentives; rising to 13-14% with global minimum tax |
| Assumption | Value | Rationale |
|---|---|---|
| D&A (FY2026) | $9.3B | 20% of beginning PP&E ($46.6B). Historically consistent — FY2024: 20.1%, FY2025: 21.0%. |
| FY2026 bit growth | +20% | Management guided (Q1 FY2026 earnings call). From existing footprint efficiency and node transitions (1-gamma, G9), not new fabs. SemiAnalysis estimates +18% for CY2026, consistent.⁶ |
| FY2026 cost/bit | ~Flat (blended) | Front-end improvements (~9-10% ex-HBM) largely offset by HBM packaging intensity (~3x wafer trade ratio). Management declined to guide cost/bit for FY2026.⁵ |
| COGS trajectory (FY2026) | $6.0B → $6.5B → $7.5B/qtr | D&A anchors fixed costs. Variable manufacturing costs scale with bit shipments (+20% annual). Growth loads into H2: management said Q1 bit growth was “very modest” and Q2 “primarily driven by price,” implying most of the 20% annual bit growth falls in Q3-Q4. Q4 includes 14-week adjustment (~$7.0B on 13-week basis). |
| GM ceiling | 75% | NVIDIA-like margins for a supply-constrained oligopoly. Revenue derived: COGS ÷ (1 - GM%). |
| FY2026 revenue | $86B | Derived from COGS trajectory ($26B) and GM expansion toward 75%. Demand-supply gap supports the margin: AI CapEx ~+70% YoY while memory capacity investment ~40% of PP&E.³ |
| FY2027 bit growth | +30% | New fabs ramping: M15X, Yongin Phase 1, Idaho fab #1.¹ Volume-driven year. Note: SemiAnalysis estimates +60% for CY2027 — significantly above our FY2027 assumption. The gap is partly CY/FY timing (new fabs come online H2 CY2027, which is partly FY2028), but if supply ramps faster than modeled, both revenue and margin compression accelerate.⁶ |
| FY2027 COGS | $33B | D&A steps up to $11.5B (+$2.2B per PP&E schedule). Variable costs scale with 30% bit growth. |
| FY2027 revenue | $105B | Derived from COGS ($33B) and avg GM ~69% (75% → ~62% through year as new supply hits). |
| FY2027 OpEx | $6.3B | ~+8% YoY. |
| Supply squeeze | Through mid-2027 | All three producers ramping new capacity; meaningful volume by H2 2027.¹ |
| FY2028 revenue (soft) | $75B | Pricing moderates gradually. AI demand partially absorbs new capacity. |
| FY2028 revenue (hard) | $55B | Samsung breaks discipline. Pricing falls sharply. |
Revenue derivation. The model is cost-base driven. COGS has two components: D&A (~20% of beginning PP&E, historically consistent) and variable manufacturing costs (scaling with bit shipments). Bit growth of +20% (guided) with flat blended cost/bit (HBM packaging intensity offsets front-end improvements) pushes COGS from $6.0B/qtr (Q1-Q2) to $7.5B (Q4, 14-week). Revenue follows: COGS ÷ (1 - GM%). At 75% GM and $7.5B COGS, Q4 revenue reaches ~$30B — summing to ~$86B for the year.³⁵
| # | Risk | Severity | Probability | Timeframe |
|---|---|---|---|---|
| 1 | Timing — selling too late | High | Medium-High | Mid-2027 |
| 2 | Earlier-than-expected cycle turn | High | Medium | CY2026-2027 |
| 3 | Multiple compression caps upside | Moderate | Medium-High | Ongoing |
| 4 | Samsung breaks discipline early | High | Medium | CY2027 |
| 5 | AI CapEx disappointment | High | Low-Medium | 1-2 years |
#1 — Timing is the #1 risk for a cyclical trade. Memory stocks peak before earnings peak — the market prices the downturn 2-3 quarters early. In the last cycle, Micron’s stock peaked early 2022 while earnings didn’t trough until mid-2023. The stock dropped ~50% before the decline was obvious. Exit on leading indicators (contract prices, inventory days), not lagging ones (quarterly earnings).
#3 — Multiple compression is the “priced in” risk. Historical memory P/E at peak: 6-11x. If the market pays only 8x on $42 EPS (FY2026), that’s ~$336 — below $420 today. Upside depends on earnings exceeding expectations or the market granting a higher multiple.
#4 — Samsung is the historical discipline-breaker. Record quarterly profit, massive cash to invest, planning a ~50% HBM capacity surge in 2026 (~160K to ~250K wafers/month per TrendForce, Dec 2025). When prices start softening, the temptation to hold volume by cutting price is strong.
| Scenario | EPS | P/E | Implied Price | vs. $420 |
|---|---|---|---|---|
| FY2026 base | $42 | 11x | $462 | +10% |
| FY2027 peak (base) | $50 | 10x | $500 | +19% |
| FY2027 peak (market pays up) | $50 | 12x | $600 | +43% |
| FY2028 soft landing | $27 | 11x | $297 | -29% |
| FY2028 hard landing | $14 | 9x | $126 | -70% |
Upside: +19% to +43% to the peak. Downside: -29% to -70% if you hold through the turn. Higher peak earnings improve the upside; higher cost base worsens the downside. The trade works — if you exit in the window.
Sell window: Late CY2026 to early CY2027.
Exit triggers:
Do NOT exit on:
| Signal | Impact |
|---|---|
| GM reaches 75% by Q3 (not Q4) | FY2026 closer to $90B+. Raises peak. |
| DRAM prices flatten before mid-2027 | Cycle turning earlier. Accelerate exit. |
| AI CapEx guidance >$1T for CY2027 | Demand stronger than expected. FY2027 could exceed $105B. |
| Samsung announces aggressive price cuts | Discipline breaking. Exit immediately. |
| Micron inventory days rising >150 | Demand slowing (Q1 FY2026 baseline: 126 days). Tighten exit window. |
| GM fails to reach 72% by Q3 FY2026 | Pricing surge weaker than modeled. Revise down toward $80B. |
| Micron CY2027 bit supply closer to +60% than +30% | SemiAnalysis estimates +60%.⁶ Higher revenue but faster margin compression — peak may be sharper and shorter. FY2028 downside worsens with larger cost base. |
[1] New fab timelines. OpenMetal supply-demand analysis (openmetal.io) projects supply relief beginning 2027 with normalization 2028-2029. Key fabs: SK Hynix M15X (opened May 2026, ahead of schedule), SK Hynix Yongin Phase 1 (completed 2027), Micron Idaho fab #1 (production H2 CY2027, per Micron IR), Samsung P5 (late CY2027).
[2] DRAM pricing. TrendForce revised Q1 CY2026 contract price forecast to +90-95% QoQ for conventional DRAM (Feb 2, 2026; trendforce.com), up from +55-60% in the Jan 5 initial forecast. Server DRAM ~+90%, PC DRAM >+100%, mobile DRAM ~+90% — record highs across all categories. PCPartPicker retail trends (pcpartpicker.com/trends/price/memory) confirm the trajectory with a short lag.
[3] Demand-supply gap. Demand: Big 4 hyperscaler CapEx was ~$383B in FY2025 (Meta $72B, Amazon $131B, Alphabet $91B, Microsoft $88B). FY2026 guidance totals ~$635-665B (Meta $115-135B, Amazon ~$200B, Alphabet $175-185B, Microsoft ~$145B run-rate) — ~70% YoY growth. Supply: memory industry CapEx/PPE is ~40% across major producers (Micron: $20B / $46.6B = 43%), but new capacity takes 2-3 years from commitment to wafer output. Demand growing ~70% vs supply constrained by 2-3 year build cycles = wide near-term gap.
[4] Micron $200B US investment. Announced February 18, 2026. $150B for manufacturing (Idaho, New York, Virginia) + $50B for domestic R&D over 20 years. Idaho: two new 600K sq ft DRAM/HBM fabs — first opening mid-2027, second before end 2028. Funded by operating cash flow, CHIPS Act grants ($6.4B), and 35% investment tax credits. Near-term CapEx (~$20B/year) consistent with prior guidance. (Source: Micron IR)
[5] HBM wafer trade ratio and pricing. HBM production requires ~3x the wafer capacity of conventional DRAM, compounding the supply constraint (OpenMetal, ibid). DDR5 64GB pricing up 474% over 2 years; NVMe costs up 223% over 3 years. Management declined to guide cost-per-bit for FY2026 — Mark Murphy noted they’re “moving away from” cost detail because “mix makes it harder to follow” (Q3 FY2025 earnings call).
[6] SemiAnalysis bit supply/demand model. Industry HBM bit demand growing ~40% annually in CY2025, CY2026, and CY2027. Micron bit supply: +18% CY2026 (consistent with management’s +20% FY2026 guidance), +60% CY2027 (significantly above our +30% FY2027 assumption — the gap is partly CY/FY timing as new fabs come online H2 CY2027). (Industry estimate, not management-sourced.)
Analysis as of February 2026. Market data approximate. Not investment advice.