GauchoRico | Filed: 2026-02-22 Quarter: Q3 FY25 (March–May 2025) | Reported: June 25, 2025
MU delivered a record Q3 with a massive beat on every metric — revenue $9.3B (vs. $8.7B guide, +6.9%), gross margin 39% [Non-GAAP] (vs. 36.5% guide, +250 bps), EPS $1.91 [Non-GAAP] (vs. $1.65 guide, +15.8%), FCF $1.9B+ (highest quarterly result in 6+ years). HBM run rate hit $6B+ annualized with 12-high ramping ahead of schedule. Data center revenue more than doubled YoY. Q4 guidance of $10.7B revenue (+15% sequential) with 42% gross margins implies continued acceleration. This is a legitimately impressive quarter — but I need to be honest with myself: MU is not my typical investment, and I have to apply my framework without getting caught up in the AI enthusiasm.
| | Q222 | Q322 | Q123 | Q223 | Q323 | Q124 | Q224 | Q324 | Q125 | Q225 | Q325 | Q426E | | | Feb-22 | May-22 | Nov-22 | Feb-23 | May-23 | Nov-23 | Feb-24 | May-24 | Nov-24 | Feb-25 | May-25 | Aug-25E | |---|---|---|---|---|---|---|---|---|---|---|---|---| | Revenue ($m) | 7,786 | 8,642 | 4,726 | 3,693 | 3,752 | 8,709 | 5,824 | 6,811 | 13,643 | 8,053 | 9,301 | ~10,700 | | QoQ % | +91% | +11% | -45% | -22% | +2% | +132% | -33% | +17% | +100% | -41% | +15.5% | +15% | | YoY % | +25% | +16% | +16% | -53% | -57% | +84% | +58% | +82% | +57% | +38% | +37% | ~40%+ |
Note: The cyclicality of this business is extreme. Revenue swings of 50-130% QoQ are not unusual. Q1 FY25 (Nov-24) of $13.6B followed by Q2 FY25 (Feb-25) of $8.1B — that's a $5.6B sequential revenue swing. This is commodity semiconductor, not SaaS. Never forget that.
| | Q322 | Q123 | Q223 | Q323 | Q124 | Q224 | Q324 | Q125 | Q225 | Q325 | | | May-22 | Nov-22 | Feb-23 | May-23 | Nov-23 | Feb-24 | May-24 | Nov-24 | Feb-25 | May-25 | |---|---|---|---|---|---|---|---|---|---|---| | Gross Margin [GAAP] | 46.7% | -0.7% | -32.7% | -17.8% | 38.4% | 18.5% | 26.9% | 56.0% | 36.8% | 37.7% | | Operating Margin [GAAP] | 34.8% | -23.9% | -62.4% | -46.9% | 25.0% | 3.3% | 10.6% | 45.0% | 22.0% | 23.3% | | Net Margin [GAAP] | 30.4% | -26.1% | -62.6% | -50.5% | 21.5% | 13.6% | 4.9% | 38.4% | 19.7% | 20.3% | | Net Income (m)|2, 626|−1, 234|−2, 312|−1, 896|1, 870|793|332|5, 240|1, 583|1, 885||FCF(m) | — | — | — | — | — | — | — | — | — | 1,900+ | | FCF Margin % | — | — | — | — | — | — | — | — | — | 20.4% |
The margin table tells the whole story of this business. Gross margins were NEGATIVE for three consecutive quarters in FY23 (Nov-22, Feb-23, May-23). A memory company with negative gross margins. That's how brutal the cycles are. The recovery to 37-39% GAAP gross margin (39% Non-GAAP) in Q3 FY25 is real progress, but it's nowhere near the 80%+ I want to see in my core positions. And that Q1 FY25 (Nov-24) spike to 56% GAAP gross margin already looks like it's normalizing — Q2 and Q3 FY25 came in at 37-38%. That sequential compression from 56% to 37% is exactly the kind of volatility that keeps me up at night.
Q4 FY25 guidance for 42% Non-GAAP gross margin is encouraging though — sequential improvement driven by more DRAM, more data center.
| Metric | Guidance | Actual | Beat | Quality |
|---|---|---|---|---|
| Revenue | ~$8,700m | $9,301m | +$601m (+6.9%) | Pricing stronger than expected |
| Gross Margin | 36.5% [NG] | 39.0% [NG] | +250 bps | "Better-than-expected pricing" |
| EPS | ~$1.65 [NG] | $1.91 [NG] | +$0.26 (+15.8%) | Margin beat + volume |
| FCF | Not guided | $1.9B+ | Best in 6+ years | Strong cash conversion |
Three consecutive quarters of beating on revenue and margins. Management has built a pattern of conservative guidance and beat-and-raise. This is exactly what I want to see from management. When Murphy says "better-than-expected pricing" — that's the most important phrase in the whole call. It means demand for HBM is pricing-inelastic right now, which is the hallmark of a product with real value-add.
| KPI | Q3 FY25 | Trend | Notes |
|---|---|---|---|
| DRAM Revenue | $7.1B | ↑ +51% YoY | 76% of total revenue |
| NAND Revenue | $2.2B | ↑ +4% YoY | Being managed for profitability |
| HBM Run Rate (annualized) | $6B+ | ↑ Ramping | Sold out for CY2025 |
| Data Center Revenue | ~$5B est. | ↑ 2x+ YoY | Record; >54% of total |
| Data Center SSD Market Share | #2 | ↑ | Third consecutive record quarter |
| DRAM Bit Shipments QoQ | +>20% | ↑ | Sequentially accelerating |
| NAND Capacity Reduction | -10% by FY25 end | ↓ Disciplined | Shifting capacity to DRAM/HBM |
| Customers (HBM) | 4 major | ↑ from 3 | GPU + ASIC platforms; expanding in CY26 |
The HBM story is the most compelling part. $6B+ annualized run rate from essentially zero two years ago. HBM is "sold out for calendar year 2025" — you don't hear that language often. The fact that MU is now shipping in high volume to 4 customers spanning both GPU (NVIDIA) and ASIC (Google TPU, Amazon Trainium, Microsoft Maia) platforms is meaningful diversification.
The 12-high HBM yield execution ahead of schedule is very good news. This is the higher-ASP product, and if yields are better than expected, margins should be better than expected. HBM4 samples shipped — generation transitions are where MU historically struggled (behind SK Hynix on HBM3), but the early HBM4 progress is encouraging.
| Factor | Assessment | Pass/Fail |
|---|---|---|
| 1. Revenue Growth >40% | Q3 FY25: +36.6% YoY. Below threshold. Q4 guidance implies 40%+ YoY. Rolling has HBM growth "significantly exceeding" DRAM demand in CY26. | ⚠️ BORDERLINE |
| 2. Trajectory | Decelerating: Q1 FY25 +57%, Q2 FY25 +38%, Q3 FY25 +37%. But absolute revenue is still growing sequentially. The YoY comp gets easier in H2 FY25. Rolling data shows Q4 FY25 at $13.6B — a big reacceleration. | ↓ then ↑ |
| 3. Gross Margins | 37-39% GAAP; 39% Non-GAAP. Far below my 80%+ threshold. This is a manufacturing-intensive commodity semiconductor business. Margins are cyclically high right now. | ❌ FAIL |
| 4. Competitive Advantage | HBM: SK Hynix leads (first mover), Samsung following. MU is #3 in HBM market share but gaining. DRAM: Oligopoly (Samsung, SK Hynix, MU). NAND: More competitive. The HBM oligopoly structure provides pricing power now, but it's not a software moat. | ⚠️ MODERATE |
| 5. Relative Valuation | EV/TTM Sales: ~2.9x (market cap $108B). For a semiconductor cyclical at peak margins, this seems reasonable. Not cheap, not expensive. CRM traded at 10-15x sales at comparable growth rates. But MU's margins aren't comparable. | ✅ REASONABLE |
| 6. Special Circumstances | YES — HBM is a genuine secular shift within memory. The AI infrastructure build-out is requiring dramatically more high-bandwidth memory per GPU. TAM is growing ($18B → $35B CY25 in HBM alone). Supply constrained through CY2025 minimum. | ✅ STRONG |
Six-Factor Verdict: MU fails my traditional software-centric criteria on margins and doesn't quite clear the 40% growth threshold in Q3. But the special circumstances factor is real. This is a secular upgrade cycle within an oligopolistic market, and MU is executing well.
The rolling file shows the story has gotten significantly better since Q3 FY25:
| Quarter | Revenue | GM% | EPS [NG] | FCF |
|---|---|---|---|---|
| Q2 FY25 (Feb-25) | $8.1B | 38% | $1.56 | $1.5B |
| Q3 FY25 (May-25) | $9.3B | 39% | $1.91 | $1.9B |
| Q4 FY25 (Aug-25) | $13.6B | 56% | $4.50 | $3.0B |
| Q1 FY26 (Nov-25) | $13.6B | 57% | $4.78 | $3.9B |
| Q2 FY26 Guide | $18.7B | 68% | — | — |
Holy smokes. Q4 FY25 came in at $13.6B against guidance of $10.7B — that's a $2.9B beat, or +27% above guidance. And then Q1 FY26 held at $13.6B with further margin expansion. The Q2 FY26 guide of $18.7B at 68% gross margins — if that holds, we're looking at a genuinely transformed margin profile. 68% gross margins would be exceptional for any company, let alone a memory manufacturer.
The promise tracker shows management has delivered on every commitment made at Q3 FY25:
This is outstanding execution. Management said it and did it.
At 2.9x TTM sales, MU is not expensive for a company growing 40%+ with accelerating margins. The comparison to CRM Case Study is imperfect because CRM was a pure SaaS model with recurring revenue and different margin structure. But even as a cyclical semiconductor, 2.9x sales for a company in the middle of a supply-constrained HBM boom is reasonable. The risk is what happens when the HBM cycle turns.
If Q2 FY26 delivers $18.7B at 68% gross margins, we're looking at a radically different earnings power than what's implied in current valuation. The market is not pricing in $18.7B quarters.
Promises Made at Q3 FY25 Call (June 2025):
Management accountability score: Outstanding. Q4 guidance beat by 27% on revenue and 1,400 bps on margins. These aren't small beats — these are commitments made conservatively in a period of strong demand. I respect that.
Cyclicality is the elephant in the room. We've seen revenue swing from $8.6B to $3.7B in a single year (FY22 to FY23). The FY23 collapse was brutal: three consecutive quarters of negative gross margins. If the AI CapEx cycle pauses — if hyperscalers pull back on GPU purchases — HBM demand could compress rapidly.
HBM customer concentration. 4 customers. If one major hyperscaler shifts spend, it's felt immediately. This is not a diversified customer base.
SK Hynix competitive response. Hynix got to HBM3 first, holds #1 share, and has multi-quarter lead time on relationships. MU is gaining share but from a weaker starting position.
The $200B CapEx commitment. $14B FY25 CapEx, $20B FY26 CapEx, $200B over 20 years. This is enormous. If the memory cycle turns before these fabs come online (first DRAM output H2 2027), Micron could be building capacity into a downturn.
Margin sustainability. Q2 FY26 guide of 68% gross margins would be extraordinary. I am skeptical until proven. Memory companies have never sustained these margins.
Thesis: MU is not a traditional GauchoRico holding. Gross margins at 37-40% [GAAP] fail my framework. Revenue cyclicality is extreme. The moat is an oligopoly structure, not a software platform or network effect.
But the special circumstances are real. HBM is not conventional DRAM. It requires different manufacturing expertise, advanced packaging, and tight customer co-design. The TAM is growing from $18B (CY24) to $35B+ (CY25) with Mehrotra saying CY26 HBM bit demand growth will "significantly exceed" overall DRAM growth. Supply is constrained. Management is executing flawlessly. Every promise from Q3 FY25 was delivered or exceeded.
The rolling data showing $13.6B quarters at 56-57% gross margins — if sustained — would fundamentally change the margin thesis. At 57% GAAP gross margins, MU starts looking more like Intel at its peak than like a commodity memory company.
My position context: wsm holds 10.4% in MU. For a memory company — which I'd normally shy away from — this is a meaningful bet on the AI infrastructure cycle and HBM specifically.
Action: HOLD at current weight. Watch Q2 FY26 results (~March 2026).
Key metrics to watch for Q2 FY26:
We will see what happens... The Q2 FY26 report (~March 2026) will be the real test. If MU delivers $18B+ at 65%+ gross margins, this is a genuinely different company than what I typically invest in, and the allocation might be justified. If it misses materially, the cycle narrative cracks and I'd reconsider the position.
Prior Belief: No prior MU analysis (first review). Updated Belief: MU is a special-circumstance AI infrastructure play with cyclical risk. HBM is a real structural shift. Management executing at a high level. Hold at current wsm allocation (10.4%). Not adding until Q2 FY26 validates the margin expansion thesis.
GR