Date: 2026-02-22 | Period: Q1 FY26 (ended Nov-25, reported ~Dec 2025) Analyst: StockNovice (Joe)
No prior Joe analysis on MU. I own it at ~10.4% of portfolio (wsm007 allocation), which means it's effectively a starter position. My rough thesis entering Q1 FY26: HBM is real, the margin expansion is real, but memory is cyclical and I wasn't sure whether to trust the guide. Three consecutive massive beats will make you a believer — or they'll make you complacent right before the cycle turns.
| Metric | Q3 FY25 [GAAP] | Q4 FY25 [GAAP] | Q1 FY26 [GAAP] | Q2 FY26 Guide |
|---|---|---|---|---|
| Revenue | $9.3B | $13.6B | $13.6B | ~$18.7B |
| YoY Growth | +37% | ++100%+ | +56% | ~+50%+ |
| Gross Margin | 37.7% | 56% | 57% | ~68% |
| Operating Margin | 23.3% | 45% | 47% | — |
| EPS [Non-GAAP] | $1.91 | $4.50 | $4.78 | ~$5.68 |
| FCF | $1.9B | $3.0B | $3.9B | — |
Beat history (revenue vs. prior guide):
That's not a beat — that's a demolition.
Here's how I'm thinking about it. Three months ago I might have called HBM a "could be" story. At this point, the data won't let me. HBM is "is":
The Q2 FY26 guidance is what makes my jaw drop: $18.7B revenue at ~68% GM. If Micron delivers Q2 anywhere near that, they'll be on an annualized run rate of roughly $65-70B with gross margin north of 60%. At a $108B market cap (from the balance sheet snapshot), that's under 2x sales and a forward P/E in the single digits. That's not a "could be" — that's a screaming "is" trading at a "no one believes it" multiple.
The one thing I hold onto: memory is cyclical. The last trough (Q2 FY23) saw -32.7% gross margins. Anyone who says "this time is different" in memory deserves skepticism. But the structural case for HBM being less cyclical than commodity DRAM is actually solid — long-term agreements, limited supplier base (three companies: SK Hynix, Micron, Samsung), and hyperscaler demand that isn't discretionary.
DRAM ($7.1B, 76% of revenue, +51% YoY): QoQ bit growth >20%. Pricing better-than-expected (CFO's words — that's a tell). HBM driving the mix shift. Legacy DRAM (D4/LP4) being EOL'd in 2-3 quarters, which naturally pushes revenue per bit up. Accelerating.
NAND ($2.2B, 24% of revenue, +4% YoY): Still weak. Capacity being reduced 10% by end FY25. NAND pricing has been terrible industry-wide. But Micron is taking the right steps — cutting supply rather than dumping product to fill capacity. The data center SSD business is actually the bright spot here (record market share, #2 position, 3 consecutive record quarters).
HBM (run rate $6B+):
**Data Center (~5BQ3FY25, > doubledYoY) : * * Thisisthecleanstory.EverydollarofHBMgoestodatacenter.AItrainingandinferencearememory − hungryandgettinghungrierwitheachmodelgeneration.TheTAMprojections(35B CY25 → $100B CY28) seem aggressive but the trajectory is undeniable.
Sanjay Mehrotra has been methodical. No hype, no big promises without follow-through. Compare the guidance history:
The pattern matters. They're either sandbagging heavily (which I now believe) or they have genuine visibility into demand that's better than their own models. The "sold out" commentary on HBM has been consistent for multiple quarters. They're not guessing.
The strategic reorganization into 4 BUs (Cloud Memory, Core Data Center, Mobile/Client, Auto/Embedded) is notable — they're restructuring around end markets, not product lines. That's a mature, customer-centric operating model. I like it.
Management credibility: High. No miss-and-lower in recent history. Beats getting bigger, not smaller.
Going in, I owned MU as a cyclical recovery + HBM optionality play. The thesis has evolved into something more durable:
Old thesis: Memory cycle recovery with HBM as upside kicker Updated thesis: Structural margin regime change driven by HBM mix shift, with data center DRAM as secular anchor and commodity memory as optionality, not the story
The $200B U.S. investment commitment over 20+ years is the government's seal of approval on Micron's strategic importance. Idaho Fab 1 first wafers by mid-CY2027 adds U.S. manufacturing capacity that carries geopolitical value on top of economic value.
What could break the thesis:
None of these are happening right now. The thesis is strengthening.
At $108.4B market cap and Q2 FY26 guidance of 18.7B(onequarter), annualizedrevenuerunratewouldbe 65-75B by H2 FY26. That puts trailing P/S below 2x and forward P/E in the 4-6x range depending on margins.
This is not a growth multiple. It's a trough cyclical multiple being applied to a company in peak earnings expansion. The market is pricing in cycle mean-reversion. Maybe that's right. But I'm not going to sell a company beating guidance by 20-27% per quarter and guiding to 68% gross margins because "memory is cyclical."
P/S of 2.9x (from the balance sheet snapshot — likely stale, actual price may differ) is cheap relative to any semi comps except Samsung. SK Hynix trades at ~3.5x. Micron deserves to trade at a premium to SK Hynix, not a discount, given U.S. manufacturing advantage, HBM4 speed lead, and customer diversification.
My take: Cheap on every metric that matters right now. The market is offering a free call option on "HBM doesn't get cyclical."
Already a starter at 10.4% of portfolio. No change needed — let it run. If Q2 FY26 delivers $18.7B with 68% margins, I'd consider adding. The only thing that would make me trim is if the Q2 beat pattern breaks or Samsung cracks the HBM qualification problem ahead of schedule.
The position sizing feels right. Heavy enough to matter if this plays out, small enough not to blow up the portfolio if memory cycles hard.
Action: Hold / Monitor
As usual, thanks for reading.