MU · US
MU.US: FQ3'26 update — HBM super-cycle & AI memory tight cycle
Micron posted FQ3 FY2026 revenue of $41.46B (+346% YoY), GAAP EPS $24.67 / Non-GAAP EPS $25.11, and operating cash flow $25.39B; data center revenue >$25B. HBM4 already shipped >$1B revenue; FQ4 guide revenue $50.0B ±$1.0B, GM ~86%, Non-GAAP EPS $31.00 ±$1.00. FY26 CapEx ~$27B. The “HBM / DRAM tight cycle for AI GPUs” thesis is intact, but cyclicality and China risk remain core variables. See NVDA demand, TSM / ASML capacity chain, and AMD GPU paths.
Cite a section with a deep link, e.g. /en/r/mu-us-research#thesis
FQ3 FY26 snapshot
- FQ3 revenue
- USD 41.5B
- Data center revenue
- USD 25.0B
- Non-GAAP EPS
- USD 25.11
- Operating cash flow
- USD 25.4B
FQ4 guidance
- FQ4 revenue guide
- USD 50.0B
- GM guide
- 1
- Non-GAAP EPS
- USD 31
- FY26 CapEx
- USD 27.0B
Thesis
Micron’s near-term pricing debate centers on the HBM / DRAM tight cycle for AI GPUs. FQ3 FY2026 validated the super-cycle with $41.46B revenue (+346% YoY), Non-GAAP EPS $25.11, operating cash flow $25.39B, and data center revenue >$25B. FQ4 guide of revenue $50.0B ±$1.0B, GM ~86%, and Non-GAAP EPS $31.00 ±$1.00 shows the cycle peak still extending.
The premium holds only if:
- HBM supply stays tight — HBM4 already >$1B revenue; HBM4E 2027 target validates next-gen demand;
- NVDA GPU shipment cadence holds — HBM demand anchors to Blackwell / Rubin cluster deployments;
- FY26 $27B CapEx returns are visible — capacity expansion amid tight TSM / ASML equipment supply.
But memory is deeply cyclical—do not price MU as a steady-state growth compounder. China DRAM/HBM policy risk, CapEx payback cycle, and DRAM spot-price inflection are mid-cycle key variables. See AMD GPU paths and NVDA demand chain.
Business
Model: IDM memory (DRAM / NAND / HBM) monetizing AI-driven HBM and DC DRAM pricing.
Flagships: HBM4/HBM4E; DC DRAM; mobile/other memory as cycle base.
Competitiveness: HBM4 volume; roadmap synced to NVDA GPU generations.
Strategy: HBM4E ~2027; FY26 CapEx ~$27B; capacity trade-offs vs commodity DRAM.
Value chain
Mid-stream memory IDM capturing HBM/DRAM cycle rents.
Upstream: ASML tools. Downstream: NVDA GPU ecosystem. Risks: HBM tool/packaging bottlenecks; China policy on DRAM/HBM.
Valuation
HBM super-cycle themes drove a massive 2026 move. As of this note (2026-07-30), valuation already prices some FQ4 $50B guide and ~86% GM peak—risk/reward is asymmetric at super-cycle peak.
Valuation frame
| Metric | Status | Read |
|---|---|---|
| 盈利增长 | 超级周期 | FQ3 营收 YoY +346%；GM 历史高位 |
| 毛利率 | 峰值 | FQ4 指引 ~86% |
| 估值 | 周期峰值定价 | 不宜用稳态成长股框架 |
| 相对 NVDA | 需求链绑定 | HBM 需求锚定 GPU 出货 |
| 催化剂 | HBM4E | 日历年 2027 目标 |
| 风险收益 | 高波动 | 周期下行回调幅度通常更大 |
Versus NVDA: Micron is a GPU demand-chain “memory layer” beneficiary—elasticity ties to GPU shipments but without CUDA-level software moat. Versus VRT and other AI infrastructure names: MU carries heavier cyclicality; drawdowns in downcycles are typically larger.
Financial trend (~24 months)
Eight-quarter revenue and gross margin with YoY and QoQ. Semi/AI hardware seasonality is secondary to supply, ASP, and order timing.
Revenue · last 8 quarters
Interactive chart available in the reader.
Gross margin · last 8 quarters
Interactive chart available in the reader.
Operations
FQ3 FY26 operations
FQ3 P&L and segments
| Item | FQ3 FY26 | Note |
|---|---|---|
| 营收 | $41.46B | 前季 $23.86B；YoY $9.30B |
| GAAP 净利润 | $28.24B | EPS $24.67 |
| Non-GAAP EPS | $25.11 | — |
| 经营现金流 | $25.39B | 现金流爆发 |
| 数据中心收入 | >$25B | HBM + DRAM 双驱动 |
Segment mix (illustrative)
| Segment | Revenue mix | Watchpoint |
|---|---|---|
| 数据中心（HBM + DRAM） | >60% | FQ3 >$25B |
| HBM | 最快增量 | HBM4 >$1B 已出货 |
| Mobile / Client | ~15% | 周期敏感 |
| Automotive / Industrial | ~10% | 长期增长但体量小 |
Takeaways:
- Revenue $41.46B nearly doubled sequentially from $23.86B; +346% YoY from $9.30B—super-cycle fully delivered.
- GAAP net income $28.24B / EPS $24.67 at historic peaks; Non-GAAP EPS $25.11 remains extreme after adjustments.
- Operating cash flow $25.39B validates earnings quality; FY26 CapEx ~$27B (FQ4 ~$10B) reflects HBM capacity expansion.
- Data center revenue >$25B >60% of FQ3 revenue—HBM + data center DRAM are core drivers.
HBM & memory cycle
HBM is Micron’s core pricing variable today. HBM4 already shipped >$1B revenue in FQ3; HBM4E targeted for calendar 2027—product cadence syncs with NVDA GPU architecture iteration (Blackwell → Rubin).
Watch:
- HBM shipment volume and ASP — quarterly HBM revenue mix and ASP trends;
- HBM capacity allocation — HBM vs data center DRAM vs mobile DRAM trade-offs;
- TSM / ASML equipment delivery — advanced packaging and EUV capacity are HBM supply bottlenecks;
- Competitive landscape — Samsung / SK Hynix HBM share and pricing.
Typical memory cycle pattern: at super-cycle peaks, GM / EPS / valuation all sit at highs; inflection signals are usually HBM ASP decline, DRAM spot turning down, or hyperscaler inventory adjustments.
Data center revenue
Data center (HBM + DRAM) is Micron’s largest FQ3 segment at >$25B (>60% of FQ3 revenue). Drivers:
- HBM for AI GPUs — NVDA Blackwell / Rubin, AMD Instinct MI series HBM demand;
- Data center DRAM — high ASP server DRAM for AI training/inference clusters;
- Supply tightness — FY26 $27B CapEx expands HBM capacity, but TSM CoWoS and other advanced packaging remain industry bottlenecks.
If NVDA GPU shipment cadence slows or AI CapEx plateaus, data center DRAM/HBM demand will be the first segment adjusted.
Competition
Strategy / product / risk map
Three axes:
| Axis | Micron | NVDA | Samsung / SK Hynix |
|---|---|---|---|
| HBM | HBM4 >$1B; HBM4E 2027 | GPU defines HBM spec | Share competition; pricing pressure |
| Cycle | Deep cyclical; peak GM ~86% | Platform; GM ~75% | Memory cycle synchronized |
| CapEx | FY26 ~$27B | Asset-light fabless | Heavy IDM |
| China | DRAM/HBM policy risk | Export controls | Local capacity competition |
Micron’s role in the HBM super-cycle is “pricing beneficiary when supply is tight”—but cyclicality means peak valuation is unsustainable. Do not price MU like VRT-style AI infrastructure steady-growers.
Peer comparison
| Company | Share / role | GM | Strength | Weakness |
|---|---|---|---|---|
| Micron (self) | HBM / DRAM | ~86% peak | HBM4 ramp + DC mix | Cycle peak valuation |
| SK hynix | HBM leader share | High-cycle | HBM capacity lead | Korea geo / CapEx |
| Samsung Memory | DRAM / HBM | High-cycle | Scale IDM | HBM share recovery race |
Management
CEO Sanjay Mehrotra and CFO Mark Murphy are stable. No material 24m C-suite change. Stability: steady — cycle pricing/CapEx matter more than personnel.
Key management (24m)
| Role | Name | Since | 24m change |
|---|---|---|---|
| CEO / President | Sanjay Mehrotra | 2017-05 | No change |
| CFO | Mark Murphy | 2022-04 | No change |
| EVP Global Ops / Tech | Scott DeBoer / ops bench | — | No material C-suite turnover disclosed |
Outlook
Company guide · FQ4 FY26
| Item | Guide | Note |
|---|---|---|
| FQ4 营收 | $50.0B ±$1.0B | 环比 +20%+ |
| 毛利率 | ~86% | 处历史高位 |
| Non-GAAP EPS | $31.00 ±$1.00 | — |
| FY26 CapEx | ~$27B | FQ4 ~$10B |
| HBM4 收入 | >$1B 已出货 | HBM4E 目标 2027 |
Framework:
- Hitting $50.0B ±$1.0B would validate a FY26 full-year revenue path approaching $130B+—magnitude proof of the super-cycle.
- GM ~86% at absolute historic highs—any sequential GM decline will be read as cycle inflection.
- HBM4 >$1B already shipped validates product cadence; HBM4E 2027 anchors next-gen demand.
- FY26 CapEx ~$27B is double-edged—capacity expansion supports HBM supply but adds depreciation pressure in downcycles.
Scenarios
| Scenario | Conditions | Implication |
|---|---|---|
| Bull | FQ4 >$51B; HBM4E early ramp; NVDA GPU shipments accelerate; GM holds 86%+ | Super-cycle extends; upside continues |
| Base | FQ4 roughly hits $50B; HBM4 ramps steadily; GM ~86% | Peak-range volatility; cyclical pricing |
| Bear | FQ4 misses guide; HBM ASP declines; DRAM spot turns; China policy shock | Cycle inflection; 30–50% drawdown possible |
Positioning frame (research only): caution at cycle peak; avoid aggressive chase after FQ4 guide raise. Track HBM shipments/ASP, NVDA GPU demand, DRAM spot pricing, China policy.