Date: 2026-05-08 Price: $714.36 (52w high $716.72 — essentially at the top) Verdict: NOT A BUY at $714. Better business than INTC but classic shr-017 cyclical trap risk: trailing metrics flatter peak HBM/AI cycle profitability, all C-suite selling, trading 28% above mean analyst PT.
| Item | Value |
|---|---|
| Market cap | ~$806B |
| Shares out | 1,128M (mild SBC dilution ~0.4%/yr; no ATM) |
| 52w range | $84.68 – $716.72 |
| 1y / 3y / 5y return | +658% / +960% / +761% |
| Dividend | Intact, recently raised $0.46→$0.60 annualized (yield 0.09%) |
| Trailing P/E | 33.2x (TTM EPS $21.49) |
| Forward FY26 P/E | ~12.3x (consensus $58.11) |
| Forward FY27 P/E | 7.0x (consensus $101.78 — wide uncertainty band) |
| P/B | 11.1x reported / 15.2x tangible |
| Net cash | -$3.79B (i.e., $3.8B net cash position) |
| Filter | Result | Notes |
|---|---|---|
| Size | PASS | $37.4B FY25 revenue |
| Current ratio ≥2.0 | PASS | 2.90x |
| 10yr earnings stability | FAIL | FY23 loss -$5.83B; cyclical |
| 20yr uninterrupted dividend | FAIL | Suspended 1996-2021; only 4 years continuous |
| EPS growth +33% over 10yr | PASS | Strongly positive 10y view |
| P/E ≤15x | FAIL | Trailing 33.2x; forward 12.3x |
| P/B ≤1.5x | FAIL | 11.1x; P/E×P/B = 512 (cap of 22.5) |
Score: 3/7. Better than INTC but still 4 fails. LT debt vs working capital filter PASSES ($7.3B vs $27.1B at Q2 FY26).
| FY | Revenue | Net Income | Diluted EPS | Cycle Phase |
|---|---|---|---|---|
| 2021 | $27.7B | $2.7B | $2.43 | Recovery |
| 2022 | $30.8B | $8.7B | $7.75 | PEAK |
| 2023 | $15.5B | -$5.83B | -$5.34 | TROUGH |
| 2024 | $25.1B | $0.78B | $0.70 | Early recovery |
| 2025 | $37.4B | $8.54B | $7.59 | Up-cycle |
| 2026E | ~$95B run-rate | ~$58/sh consensus | $58.11 cons | Apparent super-cycle |
Q2 FY26 (Feb 2026): Revenue $23.86B vs ~$19B est, EPS $12.20 vs $9.16 cons, gross margin ~74%. HBM/AI driven.
Bull case: HBM (high-bandwidth memory for AI training) creates a structurally new demand vertical less tied to consumer cycles. NVDA Blackwell requires HBM in every GPU. Micron sold out through CY26. Net cash position. CapEx slows after 18A/HBM3E ramp. FY27 EPS $100+ → 7x P/E is genuinely cheap.
Bear case (shr-017): Memory has always looked cheap on trailing peak earnings. FY22 → FY23 swung $14B in net income on a 50% revenue collapse. Samsung + Hynix are ramping HBM capacity aggressively. Consensus FY26 EPS range is $28-$64 — a 2.3x spread reflecting deep uncertainty. China export controls are an ongoing tail risk. 42%-of-revenue capex limits real FCF.
| Year | Reported FCF | SBC | SBC-adj FCF | Notes |
|---|---|---|---|---|
| FY23 | -$6.12B | $0.6B | -$6.71B | Trough |
| FY24 | $0.12B | $0.83B | -$0.71B | Recovery |
| FY25 | $1.67B | $0.97B | $0.70B | Up-cycle |
| FY26 Q2 | +$5.52B (qtr) | — | — | Annualized ~$22B |
FCF compression driver here is capex, not SBC ($15.9B capex vs $1B SBC FY25). Q2 FY26 standalone FCF run-rate would put forward FCF yield at ~2.7%. Not cheap, not absurd.
Zero C-suite open-market buys in 12 months. Only buy: Director Mark Liu, 23,200 sh @ $336-337, ~$7.8M (Jan 2026 — at half the current price).
Sells (selected, last 12 months, all C-suite):
Aggregate >$150M C-suite sales vs $7.8M director buy. Even discounting 10b5-1 plans, the unanimity is the signal. Per shr-002, this is a strong bearish overlay; no countervailing independent open-market officer buying.
g = (P/E - 8.5) / 2
Per shr-003, the gap between trailing (12.4%) and forward (1.9%) is a profitability-inflection signal — consistent with a cycle peak being priced in. Per shr-005, PEG is unreliable here because FY26→FY27 EPS goes from $58 to $102 (75%+ growth) — a meaningless trailing growth rate. Use normalized through-cycle EPS instead.
Through-cycle normalized EPS (5y avg incl. FY23 loss): ~$3.50/share. At $714 = 204x normalized P/E. This is the discipline shr-017 demands.
The Graham play on memory is buying the trough, not the peak:
The cycle eventually turns. MU will trade below $200 again. That's the entry.
Better business than INTC by every metric (profitable, FCF-positive, growing dividend, net cash, real AI exposure). But same conclusion at this price: late-cycle AI momentum trade, not a Graham value entry. Watch for cycle softening before considering.