Micron & the Supercycle — Faircurve Equity Pulse · 1 Jun 2026
Micron closed at $971, up nearly tenfold in twelve months and roughly +75% in May alone, on the sharpest memory up-cycle on record. The earnings are not in dispute: fiscal Q2 delivered $23.9B of revenue, a 75% gross margin and $12.08 of EPS, with DRAM prices up mid-60s% sequentially as HBM for AI accelerators met deliberately tight supply. The call turns on one distinction. AI demand looks structural — hyperscaler capex is committed years out in the trillions, high-bandwidth memory (HBM) is sold forward under multiyear agreements, and its supply is gated by advanced packaging rather than easily-added wafers — and Faircurve’s base credits that, valuing the HBM-era run-rate as the floor, far above any prior cycle. What it does not extrapolate is the pricing layer: a 75% gross margin guided to 81%, on selling prices the Street’s own consensus expects to fade. Stripping the pricing spike while crediting the structural demand places base fair value at $640; if pricing proves as durable as demand, the bull case is $1,150, modestly above spot. The initiation is SELL · $640 (−34%), at reduced conviction — a call on whether the pricing spike, not the demand, is permanent.
By Faircurve Research
Micron & the Supercycle
SPOT $971.00 (+5.1%)
SELL · $640
Micron closed at $971, up nearly tenfold in twelve months and roughly +75% in May alone, on the sharpest memory up-cycle on record. The earnings are not in dispute: fiscal Q2 delivered $23.9B of revenue, a 75% gross margin and $12.08 of EPS, with DRAM prices up mid-60s% sequentially as HBM for AI accelerators met deliberately tight supply. The call turns on one distinction. AI demand looks structural — hyperscaler capex is committed years out in the trillions, high-bandwidth memory (HBM) is sold forward under multiyear agreements, and its supply is gated by advanced packaging rather than easily-added wafers — and Faircurve’s base credits that, valuing the HBM-era run-rate as the floor, far above any prior cycle. What it does not extrapolate is the pricing layer: a 75% gross margin guided to 81%, on selling prices the Street’s own consensus expects to fade. Stripping the pricing spike while crediting the structural demand places base fair value at $640; if pricing proves as durable as demand, the bull case is $1,150, modestly above spot. The initiation is SELL · $640 (−34%), at reduced conviction — a call on whether the pricing spike, not the demand, is permanent.
The Snapshot
The Debate at $971
HBM is a genuine step-change
High-bandwidth memory is sold out through calendar 2026 under multiyear agreements and prices more like a custom-logic product than a spot commodity — structurally lifting the mid-cycle margin floor.
A disciplined three-player oligopoly
DRAM is now Micron, SK Hynix and Samsung. Capital discipline and surging AI demand have kept the market in deficit for six consecutive quarters.
AI demand is secular, not a quarter
Data-center is more than half of revenue; accelerator and high-capacity memory module bit-demand compounds with every training and inference cycle.
Record cash, net-cash balance sheet
Operating cash flow of $11.9B in the quarter funds a >$25B capex build and a raised dividend without leverage.
Peak margins always mean-revert
A 75% gross margin — guided to ~81% — is unprecedented for memory and has no durable historical analogue. The question is when pricing normalises, not whether.
Supply is being built into the peak
Micron is lifting FY26 capex above $25B while Hynix and Samsung expand. Memory down-cycles are made by the capacity ordered at the top.
A low multiple on peak earnings is the trap
Memory has historically been cheapest on P/E at the top. 9.6x forward is not a discount but a signal the market is not extrapolating the peak.
Elasticity is already biting demand
Management flags that memory pricing may crowd out content, with PC and smartphone units at risk of low-double-digit declines in calendar 2026.
The franchise debate is settled; the valuation debate turns on one distinction. AI demand durability looks genuinely structural — hyperscaler capex is committed years out in the trillions, high-bandwidth memory (HBM) is sold forward under multiyear agreements, and its supply is gated by advanced packaging rather than easily-added wafers. Faircurve credits that: the base values the FY26 HBM-era run-rate as the floor, far above any prior cycle. What it does not extrapolate is the pricing layer on top — a 75% gross margin guided to 81%, with DRAM selling prices +mid-60s% in a single quarter. Strip the spike, keep the structural demand, and fair value is ~$640. The call is whether the spike, not the demand, is permanent.
Three signals settle the durability question: the DRAM pricing trajectory into the seasonally softer first half of calendar 2026; industry capex additions from all three makers, the mechanism that has ended every prior up-cycle; and the breadth of HBM demand beyond the current accelerator roadmap. A second consecutive quarter of sequential price gains would push the read toward the bull case; the first sign of inventory rebuild at hyperscalers would confirm the base.
Q2 FY26 — the peak, in numbers
Micron’s February quarter was not merely a record; it was a different order of magnitude. Revenue of $23.9B rose +196% year-on-year and +75% sequentially, as DRAM contract prices climbed mid-60s% in a single quarter and HBM shipped at volume into the latest accelerator roadmap. The gross margin reached 74.4% — a level memory has never sustained — and management guided the next quarter higher still, to roughly 81%. The signal worth holding is precisely that: a margin this far above the franchise’s own history is the definition of a cyclical peak, not a new normal.
Operational dashboard, Q2 FY26 (quarter ended 26 Feb 2026)
| Metric | Q2 FY26 | Sequential | Year-on-year |
|---|---|---|---|
| Revenue | $23.86B | +74.9% | +196.3% |
| DRAM revenue | $18.8B (79%) | strong | record |
| NAND revenue | $5.0B (21%) | strong | record |
| Gross profit | $17.75B | — | — |
| Gross margin | 74.4% | +18 pts | +37 pts |
| Operating margin | 67.6% | — | — |
| Diluted EPS | $12.08 | +163% | +757% |
| Operating cash flow | $11.9B | +42% | record |
| Free cash flow | $6.9B | +77% | record |
| Capex (quarter) | $5.0B | — | — |
Eight-quarter trajectory — revenue & non-GAAP gross margin
| Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |
|---|---|---|---|---|---|---|---|---|
| Revenue ($B) | 6.8 | 7.8 | 8.7 | 8.1 | 9.3 | 11.3 | 13.6 | 23.9 |
| Gross mgn (NG) | 28% | 36% | 38% | 38% | 39% | 46% | 57% | 75% |
| EPS (NG, $) | 0.62 | 1.18 | 1.79 | 1.56 | 1.91 | 3.03 | 5.20 | 12.08 |
The cyclicality, in one table
| Fiscal year (Aug) | Revenue | Gross mgn | EBITDA | Net income | Dil EPS | Capex | FCF |
|---|---|---|---|---|---|---|---|
| FY21 | $27.7B | 37.6% | $12.9B | $5.9B | $5.14 | $10.0B | $2.4B |
| FY22 | $30.8B | 45.2% | $16.7B | $8.7B | $7.74 | $12.1B | $3.1B |
| FY23 | $15.5B | -9.1% | $2.2B | -$5.8B | -$5.34 | $7.7B | -$6.1B |
| FY24 | $25.1B | 22.4% | $8.9B | $0.8B | $0.70 | $8.4B | $0.1B |
| FY25 | $37.4B | 39.8% | $18.5B | $8.5B | $7.59 | $15.9B | $1.7B |
| FY26E (cons) | $108.6B | ~70% | — | — | $58.87 | >$25B | — |
No table makes the case more economically than this one. In the space of three fiscal years Micron swung from a $8.7B profit (FY22) to a $5.8B loss with a negative gross margin (FY23) and back to record profitability — the signature of a commodity whose price, not its volume, drives the income statement. The current cycle is larger in amplitude than any before it: FY25’s $7.59 of EPS is on track to multiply roughly eightfold to a consensus $58.87 in FY26 and above $100 in FY27. That step-change is real, and HBM is a genuine reason the trough should be higher than in the past. It is not a reason to treat the peak as the baseline — the FY23 column is only thirty months old.
What management said, eight quarters running
Theme frequency — mentions per call (oldest → newest)
| Theme | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 |
|---|---|---|---|---|---|---|---|---|
| HBM | 27 | 24 | 17 | 32 | 38 | 33 | 30 | 24 |
| Data center / AI | 44 | 41 | 30 | 47 | 49 | 45 | 42 | 46 |
| DRAM | 38 | 34 | 24 | 41 | 36 | 39 | 33 | 31 |
| Supply / tightness | 16 | 18 | 13 | 17 | 19 | 24 | 31 | 27 |
| Pricing / selling prices | 14 | 16 | 11 | 19 | 17 | 18 | 21 | 19 |
| Capex / capacity | 22 | 24 | 12 | 19 | 18 | 21 | 27 | 25 |
| Inventory | 11 | 13 | 18 | 12 | 8 | 6 | 4 | 4 |
| Long-term / supply agreements | 3 | 1 | 2 | 1 | 1 | 2 | 4 | 9 |
Recurring metrics management quotes
| Metric | Q3'24 | Q2'25 | Q4'25 | Q1'26 | Q2'26 |
|---|---|---|---|---|---|
| Total revenue | $6.8B | $8.1B | $11.3B | $13.6B | $23.9B |
| Non-GAAP gross margin | 28% | 38% | 46% | 57% | 75% |
| DRAM price (sequential) | +~20% | +mid-s.d. | +low-d.d. | +20% | +mid-60s% |
| NAND price (sequential) | +~20% | -high-teens | +high-s.d. | +mid-teens | +high-70s% |
| HBM run-rate | >$100M/q | >$1B/q | ~$8B ann. | record | record |
| Free cash flow | $0.4B | $0.9B | $0.8B | $3.9B | $6.9B |
The transcript arc is a near-monotonic march from recovery to euphoria: gross margin from 28% to 75%, HBM from a sub-$100M curiosity to an ~$8B annualised run-rate, and DRAM pricing re-accelerating to +mid-60s% in a single quarter. Two sub-surface signals matter for the call. First, the “inventory” line fades to near-silence while “supply / tightness” and “supply agreements” rise — the language of a peak. Second, management itself frames pricing as structurally driven yet concedes elasticity risk in PCs and phones. Both are consistent with a cycle that is extraordinarily strong and, by management’s own framing, extraordinary rather than normal.
Guidance cadence and three quotes
Next-quarter guidance track
| Guided for | Rev midpoint | GM guide | EPS guide | Framing |
|---|---|---|---|---|
| after Q3'24 | $7.6B | 34.5% | $1.08 | HBM sold out calendar 2024/25 |
| after Q2'25 | $8.8B | 36.5% | $1.57 | HBM >$1B; tight leading edge |
| after Q4'25 | $12.5B | 51.5% | $3.75 | records; HBM share = DRAM share |
| after Q1'26 | $18.7B | 68% | $8.42 | calendar 2026 sold out; tight beyond calendar 2026 |
| after Q2'26 | $33.5B | ~81% | $19.15 | first 5-yr supply agreement; supply far below demand |
Guidance has gone vertical — from a $1.08 EPS guide two years ago to $19.15 for the coming quarter — and the qualitative framing has hardened from “sold out” to “supply much less than demand.” That is the bull case in management’s own words. The one consistent counterweight, voiced by the CFO, is that memory remains a commodity: NAND weakness, underload charges and elasticity have all surfaced even inside this up-cycle. The durability question is whether the multiyear agreements convert “sold out” into a genuinely higher floor — or merely delay the point at which added supply meets cooler demand.
Where it trades vs the complex
Faircurve-curated peer set. Micron is anchored against the memory & storage complex that shares its narrative engine and cycle: the DRAM oligopolists SK Hynix and Samsung Electronics, the NAND pure-play SanDisk, and HDD-maker Western Digital as a storage-cycle adjacency. All multiples use each name’s next-fiscal-year (FY+1) consensus, every name on a price-over-consensus basis. Anchor years: MU FY27 (Aug’27), Hynix & Samsung FY27 (Dec’27), SanDisk & WDC FY27 (Jun’27) — all ≈ calendar 2027. Korean caps converted at ₩1,410/$. The crucial read: the DRAM/NAND makers all trade at single-digit forward P/Es — SK Hynix 6.2x, Samsung 6.5x, SanDisk 9.6x, Micron 9.6x — while HDD (WDC, 30.6x) sits much higher because it is earlier in its own ramp. Because two of four peers screen n/m on forward EV/EBITDA (SanDisk’s GAAP EBITDA is negative on post-spin amortisation), the valuation blend substitutes forward EV/Sales for the EV/EBITDA leg.
Forward P/E (FY+1) — sorted high to low
Full peer table — forward FY+1 multiples
| Ticker | Mkt cap | Fwd P/E | Fwd EV/EBITDA | Fwd P/S | Fwd EBITDA mgn | Fwd rev gr |
|---|---|---|---|---|---|---|
| SK Hynix | ≈$1.20T | 6.2x | 8.2x | 3.8x | 45.8% | +37% |
| Samsung Elec | ≈$1.64T | 6.5x | 10.5x | 2.9x | 26.5% | +21% |
| SanDisk | $251B | 9.6x | n/m | 5.9x | n/m | +117% |
| Western Digital | $183B | 30.6x | n/m | 10.4x | n/m | +37% |
| Peer avg (ex-MU) | — | 13.2x | 9.4x | 5.7x | 36.2% | +53% |
| MU | $1.10T | 9.6x | 12.3x | 6.1x | 50.0% | +64% |
Memory-maker cluster only (Hynix, Samsung, SanDisk), the truest comparison for cycle phase: forward P/E mean 7.4x. Micron at 9.6x sits at the rich end of the DRAM/NAND oligopoly — there is no relative-value discount to lean on. SanDisk’s +117% forward revenue growth and negative GAAP EBITDA reflect a noisy post-spin estimate base and are footnoted as such; WDC’s 30.6x is an HDD name earlier in its own up-cycle, not a memory comp.
On its merits, Micron is a co-leader of a consolidated, structurally improved industry, with the strongest balance sheet of the three DRAM makers and a credible HBM franchise. The peer set does not, however, supply a valuation cushion. Every memory maker trades on a single-digit forward P/E, and Micron at 9.6x is the most expensive of them. On Micron specifically the low multiple is unambiguously a peak multiple: the Street’s own consensus has EPS cresting in FY27 (~$101.58) and fading to ~$69 by FY29 — so 9.6x is paid on earnings analysts already expect to decline. The relative-value case for the stock would require the peak to hold — the proposition the absolute valuation tests.
The math, line by line
Base-case derivation (normalised through the cycle)
| Variable | Formula · inputs · arithmetic | Base value |
|---|---|---|
| Normalised EPS | FY26 HBM-era run-rate (cons $58.87) held as the structural floor — AI-demand step-up credited, FY27 pricing spike ($101.58) stripped | $58 |
| Mid-cycle P/E | memory makers 6–10x on peak EPS → ~12.5x on normalised EPS (prior mid-cycles 10–14x; HBM-leader premium) | 12.5x |
| → P/E fair value | $58 × 12.5x | $725 |
| Normalised revenue | ≈ FY27 cons $178.5B less pricing-spike premium | ~$175B |
| Fwd EV/Sales (sub.) | memory-maker fwd mean ~4.2x, applied just below peak | 4.0x |
| → EV/Sales fair value | $175B × 4.0x = $700B EV + $2B net cash ÷ 1.14B sh | $616 |
| DCF — discount rate (WACC) | Cost of equity = 4.3% risk-free + (1.92 beta × 5.0% equity risk premium) = 13.9%; blended with 4.7% after-tax cost of debt (debt ≈1% of capital) | 13.8% |
| DCF — terminal growth | nominal-GDP-anchored | 3.0% |
| → DCF fair value | 5-yr FCF fade off the super-cycle at a 13.8% discount rate | ~$540 |
| Blended fair value | 40% P/E ($725) + 40% EV/Sales ($616) + 20% DCF ($540) | $644 → $640 |
Bull / bear flex
| Variable | BEAR | Bear: why | BASE | BULL | Bull: why |
|---|---|---|---|---|---|
| Normalised EPS | $32 | selling prices crack as supply lands; margins revert, though the HBM floor stays above prior cycles | $58 | $98 | pricing proves as durable as demand; earnings plateau near FY27–28 consensus |
| Mid-cycle P/E | 12.5x | trough on low EPS | 12.5x | 11.7x | durable plateau warrants a near-peak multiple |
| Fair value | $400 | $640 | $1,150 |
Blended fair value & what’s priced in
The valuation deliberately breaks from the standard forward template because Micron is a commodity cyclical at a margin peak, where applying any multiple to peak earnings endorses the top. The fix is to value normalised earnings power — the HBM-era FY26 run-rate (~$58 EPS) held as the floor, which already credits a structurally higher AI-demand base (the prior cycle earned $7.59) and strips only the FY27 pricing spike ($101.58) — at a mid-cycle multiple, substituting forward EV/Sales for EV/EBITDA where peers screen n/m. The 40/40/20 blend yields $644, rounded to a $640 target. Independent anchors corroborate the direction: FMP’s reference levered DCF prints $386; sell-side consensus PT is $625 (median $450; range $310–$1,625) — both below spot, though both may lag a stock that roughly doubled in May. Even probability-weighting the three cases (25% bear / 45% base / 30% bull) leaves fair value near $730 (−25%).
What the market is pricing in
| At today’s $971 | Implied | Versus |
|---|---|---|
| Revenue growth priced in (5y to a 10% return, exit 4.0× EV/Sales) | ~+25%/yr | FY27→FY31 ($178B→~$440B, 2.5×); needs pricing to hold — bit demand alone is only ~15–20%/yr |
| Implied sustainable EPS at 12.5× | ~$78 | ~76% of the $101.58 FY27 peak, held permanently |
| EPS growth priced in (same basis, exit 12.5×) | ~+5%/yr | flattered by peak margins — ignore in favour of the revenue line |
| P/E paid on FY27 peak EPS ($101.58) | 9.6x | vs 5–8x historic memory peak multiple |
Reverse-engineering the price is the cleanest test of what is embedded. The single most judgeable figure is on the revenue line, which strips the peak-margin illusion: to clear a 10% annual return over five years — exiting at a normalised 4.0× EV/Sales — Micron’s revenue must compound at about +25%/yr, from ~$178B (FY27) to ~$440B by FY31. Against long-run DRAM bit demand of ~15–20%, that requires pricing to hold near the spike, not merely volumes to grow. In earnings terms the same price implies a sustainable EPS of ~$78 — roughly three-quarters of the FY27 peak held in perpetuity, and well above the ~$58 structural-demand floor the base credits; paying 9.6x the FY27 peak, when memory has historically commanded 5–8x at cyclical tops, means the market is treating the pricing spike, not merely the demand, as semi-permanent. The stock is not pricing the AI build-out so much as the build-out at today’s prices, without a down-cycle. That is the proposition Faircurve declines near $971 and would accept much closer to the $640 base — and the one the bull case, if pricing proves as durable as demand, renders roughly fair at $1,150.
Three cases, one call
What breaks the thesis, what triggers the move
Six principal risks (to our SELL)
Four near-term catalysts
- Q3 FY26 results (late June 2026) — the next read on pricing and margin: a guide that holds the ~81% gross margin and sequential price gains pushes toward the bull case; the first sign of moderation confirms the base.
- DRAM/NAND contract-pricing prints — monthly spot and contract data into the first half of 2026 are the highest-frequency tell on whether the spike is normalising; the mechanism that ends every memory up-cycle.
- Industry capex announcements — FY27 capital plans from Micron, Hynix and Samsung; the supply ordered at the top is what builds the next trough.
- Hyperscaler / OEM inventory commentary — any sign of inventory rebuild or elasticity in PCs and phones would confirm demand is being pulled forward by price.
Faircurve Equities · Independent Single-Name Research · MU · 1 Jun 2026 · Issue No. 016
Sources: Financial Modeling Prep (quotes, statements, key metrics, analyst estimates & price targets, transcripts, price history). 8 quarterly earnings call transcripts (Q3 FY24 through Q2 FY26). Micron Technology SEC filings (10-K FY25, 10-Q Q2 FY26). Korean peer caps converted at ₩1,410/$.
Methodology: Forward FY+1 peer comparison across the memory & storage complex. Because the subject is a commodity cyclical at a margin peak, the base values normalised mid-cycle earnings power (not the peak run-rate) at a mid-cycle multiple. 40/40/20 blend of Forward P/E, Forward EV/Sales (substituted for EV/EBITDA where peers screen n/m) and a cost-of-capital DCF (13.8% discount rate, 3.0% terminal growth), cross-checked against a reverse-DCF of what the price implies.
Disclaimer: Research, not investment advice. Author has no position in MU at publication.