Micron FQ2 FY2025: Beat the Quarter, Lost the Tape on a Margin Guide-Down

Results vs. Consensus

Quality of Beat

• Revenue: Organic and mix-led. DRAM revenue of $6.1B was up 47% year-on-year with sequential price up mid-single-digits on a richer portfolio mix, even as DRAM bit shipments fell high-single-digits. The beat came from the high-value end of the book (HBM, high-capacity DIMMs, data-center LPDRAM), not from volume.
• Margins: The 37.9% non-GAAP gross margin was down 160 bps sequentially, with NAND pricing (down high-teens sequentially) and consumer mix the culprits. This is a clean, low-quality-of-earnings-risk margin: there is no one-time benefit propping it up, but there is also no sign yet of the NAND drag easing.
• EPS: The $0.13 beat versus consensus was helped by a 10.7% effective tax rate, below the guided rate on one-time items. Operating income of $2.0B (24.9% margin, up 21 points year-on-year) is the cleaner signal: the operating leverage off the trough is real. Adjusted EBITDA of $4.1B carried a 50.7% margin.

Segment Performance

Revenue by Technology

Revenue by Business Unit

Compute & Networking: the engine

CNBU set a quarterly record for the third consecutive quarter and now represents 57% of total revenue, a structural shift from the consumer-weighted Micron of prior cycles. The driver is HBM, where revenue grew more than 50% sequentially to clear the $1B milestone, alongside the combination of high-capacity DRAM modules and data-center LPDRAM that also exceeded $1B for the quarter. Micron remains the only supplier shipping low-power DRAM into the data center in high volume.

"In fiscal Q2, data center DRAM revenue reached a new record. HBM revenue grew more than 50% sequentially to a new milestone of over $1 billion of quarterly revenue. Our HBM shipments were ahead of our plans." — Sanjay Mehrotra, Chairman, President & CEO

Assessment: This is the part of Micron that justifies an AI-memory multiple. Data-center DRAM tripling year-on-year and HBM running ahead of plan are the proof points that Micron has closed the HBM gap to its Korean competitors. As long as CNBU compounds, the structural bull case is alive regardless of what NAND does.

Storage, Mobile, Embedded: the ballast

The three non-CNBU units fell a combined double-digit percentage sequentially. Storage dropped 20% as data-center customers digested prior storage investments and NAND pricing fell; Mobile dropped 30% on continued handset-customer inventory correction; Embedded slipped 3% on automotive inventory adjustment. Management frames all three as late-stage inventory normalizations that set up bit-shipment growth in the coming quarters.

Assessment: None of these declines is alarming on its own (each maps to a known inventory cycle), but together they explain why the blended gross margin is under pressure even as DRAM thrives. NAND-heavy SBU is the margin problem; the recovery thesis here is real but back-half-weighted and dependent on supply discipline holding industry-wide.

Key Topics & Management Commentary

Overall Management Tone: Confident on the structural story, careful on the near-term margin. Management led with the HBM milestone and the "best competitive position in our history" framing, then spent most of the call's airtime defending the gross-margin trajectory under sustained analyst pressure. The posture was forward-leaning on demand and deliberately non-committal on the timing of the margin recovery, the one place the answers stayed framework-level rather than quantitative.

1. HBM Crosses $1B and Stays Sold Out

The headline event of the quarter. HBM revenue grew more than 50% sequentially past $1B, ahead of plan, and management again raised the calendar-2025 industry HBM TAM (now over $35B). Micron reiterated that it is sold out of HBM for calendar 2025 and is already negotiating 2026 supply agreements, and that it expects to reach HBM share in line with its overall DRAM share on a run-rate basis by calendar Q4 2025.

"We have once again increased our HBM TAM estimate for calendar 2025 to over $35 billion. We remain on track to reach HBM share similar to our overall DRAM supply share on a run rate basis in calendar Q4 2025." — Sanjay Mehrotra, CEO

Assessment: The HBM ramp is executing better than the company's own plan, which is the single most important fact in the quarter. Reaching parity share with overall DRAM by year-end would validate that Micron is a genuine third force in HBM, not a tail supplier. This is the pillar the entire bull case rests on.

2. The Gross Margin Guide-Down

Fiscal-Q3 non-GAAP gross margin was guided to 36.5%, down from 37.9%, the opposite of what a market expecting an AI up-cycle wanted to see. Management attributed the step-down to a higher mix of consumer-oriented volumes, lower calendar-Q1 pricing in consumer markets, and continued NAND under-utilization, partly offset by HBM growth. They committed only that fiscal-Q4 margin would be "up somewhat."

"We do expect gross margin to be up somewhat. There's always tailwinds and headwinds... we do see NAND underutilization as we talked about. And actually, since our capacity has come down structurally, we're going to see less of those costs in the third quarter on period and see more of those costs hit us as inventory clears in the fourth quarter." — Mark Murphy, CFO

Assessment: This is the crux of the Hold. A sequential margin decline with under-absorption costs explicitly migrating into fiscal Q4 means the trough in blended margin may not be behind the company yet. "Up somewhat" is not the language of a clean inflection. Until we see DRAM mix and HBM scale overwhelm the NAND drag in the reported number, the margin remains the binding constraint on the multiple.

3. NAND Run for Discipline, Not Growth

Management was explicit that NAND is being managed down: wafer output is already down mid-teens percent from prior levels, and a planned reuse of underutilized equipment for capital-efficient node conversions will take structural NAND wafer capacity down over 10% exiting fiscal 2025 versus fiscal 2024. The strategy is to trade near-term volume for industry supply-demand balance.

"This strategy results in over 10% structural reduction of NAND wafer capacity exiting fiscal 2025 compared to levels exiting fiscal 2024. We will continue to prudently manage our NAND supply." — Sanjay Mehrotra, CEO

Assessment: Strategically correct, near-term painful. Under-loading fabs protects pricing but loads the income statement with under-absorption costs. The market will reward this only once NAND pricing inflects; until then it reads as a self-inflicted margin headwind that the company is choosing for the right long-term reasons.

4. 1-Gamma EUV: Extending DRAM Leadership

Micron launched its 1-gamma node, its first DRAM node using EUV, and shipped the industry's first 1-gamma-based DDR5 products. The node delivers 20% lower power, 15% better performance, and over 30% higher bit density than 1-beta. This is the cost-and-performance engine that underpins both standard DRAM margins and future HBM competitiveness.

Assessment: Process leadership is the durable moat in memory, and a clean 1-gamma launch matters more for the three-year story than any single quarter's margin. It is also the lever that lets Micron grow bit supply below industry demand while still taking high-value share. A quiet but important positive.

5. Data-Center DRAM Records Beyond HBM

The data-center story is broader than HBM alone. High-capacity DRAM modules plus data-center LPDRAM together exceeded $1B for the quarter, with Micron the sole high-volume supplier of low-power DRAM into the data center. SOCAMM, co-developed with NVIDIA for the GB300, extends that LPDRAM franchise into a new server form factor.

"Micron remains the only company in the world to ship low-power DRAM into the data center in high volume, showcasing our pioneering innovation and deep partnership with our customers." — Sanjay Mehrotra, CEO

Assessment: The second $1B data-center franchise is underappreciated relative to HBM. It diversifies the AI-memory exposure beyond a single product and deepens the NVIDIA relationship across both HBM and LPDRAM, which raises switching costs at the most important customer in the industry.

6. Consumer Recovery: Back-Half Weighted

Management reiterated that PC and smartphone customer inventories are normalizing on schedule, with content growth from AI (16GB-plus AI PCs, 12GB-plus AI phones) layered on top of modest unit growth. The Windows 10 end-of-life in October 2025 is flagged as a PC-refresh catalyst. Bit-shipment growth in mobile is expected to resume in fiscal Q3.

Assessment: Credible and consistent with prior guidance, but second-half-weighted, which means it does little for the near-term margin debate. The AI content-growth angle is the more durable driver; the cyclical inventory recovery is the timing variable.

7. Capex Discipline at $14B

Fiscal-2025 capex was held at approximately $14B, with the overwhelming majority directed at HBM, plus multi-year facility builds (Idaho fab, Singapore advanced packaging, Taiwan test). Idaho is expected to provide meaningful DRAM output starting fiscal 2027; Singapore advanced packaging expands capacity from calendar 2027.

Assessment: The capex is front-loaded into capacity that monetizes in FY27 and beyond, which is the right posture for a structural HBM ramp but a near-term free-cash-flow headwind. FCF of $857M against $3.9B of operating cash flow shows how much the build is consuming. Capital discipline is being maintained; the absolute number is still large.

8. Balance Sheet De-Risked

During the quarter Micron extended debt maturities via a $1B 10-year senior note and a $1.7B term loan (used to retire 2026 notes and the prior term loan), ending with $14.4B total debt, low net leverage, and a weighted-average maturity of 2032. Post-quarter, the revolver was renewed and upsized to $3.5B, adding $1B of liquidity.

Assessment: Sensible liability management ahead of a heavy capex cycle. Pushing maturities out to a 2032 weighted average and adding revolver capacity means the balance sheet is not a constraint on the HBM build. One less thing to worry about.

Guidance & Outlook

The shape of the guide is the story: record revenue paired with a lower margin. Revenue at the $8.8B midpoint would be an all-time quarterly record, driven by bit-shipment growth across both DRAM and NAND. But the margin guide to 36.5% (down from 37.9%) reflects the consumer-mix and NAND-utilization drags discussed above, and the higher tax rate offsets the revenue growth to leave EPS essentially flat sequentially at $1.57.

Implied trajectory: Management would only commit to fiscal-Q4 gross margin being "up somewhat" from fiscal Q3, with under-absorption costs migrating into Q4 as inventory clears and the first DRAM start-up costs (Idaho, new node) beginning to appear. The all-in DRAM cost is guided roughly flat for FY25; all-in NAND cost down low-double-digits.

Street at: Consensus had modeled higher fiscal-Q3 gross margin (closer to 38–39%), so the guide is a genuine reset, not a sandbag. EPS consensus was broadly in line with the $1.57 guide.

Guidance style: Historically Micron guides realistically rather than conservatively. A margin reset of this size is best read as management telling the market the trough is not yet fully behind, not as a setup for an easy beat.

Analyst Q&A Highlights

The shape of the margin recovery beyond fiscal Q3

The dominant line of questioning on the call was where and when gross margin inflects. Management acknowledged fiscal Q3 is down sequentially on consumer mix and NAND, but committed to fiscal Q4 being up "somewhat," with NAND under-absorption costs migrating from period costs into inventory and then clearing through Q4, partly offset by new-node and construction start-up costs beginning to appear.

Q: "You did anticipate an improved gross margin profile beyond this quarter, fiscal Q3. So is that still the case that we should see gross margin improvements maybe starting in fiscal Q4 and potentially beyond?"
— Harlan Sur, JPMorgan

A: "We're not providing guidance on the fourth quarter. However, we do expect gross margin to be up somewhat... we do expect market conditions to improve. We do expect HBM and other high-value products to grow and contribute to mix improvement. Some headwinds, we do see NAND underutilization... we would expect fourth quarter margins to be up somewhat from third quarter."
— Mark Murphy, CFO

Assessment: The answer is the Hold thesis in miniature. "Up somewhat" with explicit headwinds named is not the language of a confident inflection. Management is being honest, but honesty here means the margin trough is not yet clearly in the rear-view mirror.

Is the pricing improvement real demand or tariff pull-forward?

A pointed question probed whether recent memory-price firmness reflected genuine end-demand or customers pulling orders forward ahead of potential tariffs. Management leaned hard on the demand side, citing consumer inventory normalization, AI content growth in PCs and phones, and strong data-center DRAM, while acknowledging the goal of driving a pricing inflection in calendar Q2.

Q: "You've seen some of the memory prices improve of late. I'm just wondering how much of that is true end demand versus actually tariff-related pull-ins? And how sustainable do you think the industry pricing dynamics today are?"
— Krish Sankar, TD Cowen

A: "Demand trends are in a good place. And of course, on the supply side, leading-edge supply... as well as leading-edge DRAM supply is tight... on NAND, the supply actions by various players and underutilization in the fabs certainly is improving the supply picture as well. We are well focused on driving an inflection towards higher pricing in CQ2."
— Sanjay Mehrotra, CEO

Assessment: Management did not directly quantify tariff pull-in, which leaves a small unanswered risk that some calendar-Q2 strength is borrowed from later quarters. The leading-edge tightness argument is the more durable point and is corroborated by the HBM trade-ratio dynamic.

The structural margin question: why lower than the last upturn?

The sharpest challenge of the call asked why gross margin is roughly 10 points below the prior up-cycle at a comparable revenue base, and whether 50%-plus margin is even achievable again. Management pointed squarely at NAND as the weight, defended DRAM margins as healthy, and reiterated the mix-shift-to-high-value strategy without committing to a specific margin target or timeline.

Q: "Your revenue is between $7.8 billion and $8.8 billion... the last time that happened, your gross margins were 10 points higher. Why are gross margins like 10 points lower at essentially the same revenue base? And does this mean that we can forget about your gross margins ever going to the 50s again?"
— Chris Danely, Citi

A: "Our gross margins in DRAM have been healthy... NAND is what has weighed down on our margins... as we see a greater supply discipline, we would certainly fully expect that NAND fundamentals would improve in the industry as well... we remain very focused on continuing to strengthen the mix of our revenue toward higher-margin products."
— Sanjay Mehrotra, CEO

Assessment: This is the bear's best question and management did not fully close it. The implicit answer is that NAND has to heal for blended margin to reclaim the 50s, which the company can influence with supply discipline but cannot control alone. The market's 8% reaction is largely this exchange priced in.

Path to sub-120-day inventory

Questions pressed on the credibility of reducing days of inventory from 158 to below the 120-day target within two quarters. Management framed the reduction as DRAM-led, driven by AI demand and HBM trade-ratio tightness pulling down leading-edge inventory, with NAND managed via the supply actions already described.

Q: "You're at 158 days and you'll be below the target model, which I think is 120 days in 2 quarters... that seems like a lot of reduction. Can you talk about how much volume do you need to do that?"
— Joseph Moore, Morgan Stanley

A: "Conditions are tighter in DRAM than NAND... AI-driven growth and related to HBM and other product sets, but particularly HBM with the trade ratio is creating tightness in that market. The target we have for inventories... is 120 days, and we would expect to be below that on DRAM by... our fiscal fourth quarter."
— Mark Murphy, CFO

Assessment: A falling DIO would be a genuine positive signal and is consistent with the leading-edge tightness narrative. It is also the cleanest near-term proof point to watch: if DRAM days fall as promised, the pricing-power story is validated independent of the margin optics.

NAND under-absorption and construction cost cadence

A modeling-oriented question sought the cadence of NAND under-utilization period costs and incremental construction costs into the back half. Management indicated fewer period costs in fiscal Q3 (because capacity came down structurally, so more lands in inventory), with under-absorption weighing on fiscal Q4 and into FY26, and relatively small sequential start-up costs that grow through 2026 as Idaho approaches wafer-outs.

Q: "Is there a way to think about the underutilization charges and the period cost into kind of May and August on the NAND side? And then is there also a way to think about the incremental construction cost as we go into August and November?"
— C.J. Muse, Cantor Fitzgerald

A: "We had less period costs in our third quarter... So more of the underutilization charge will go into inventories, and then that will flush through... Those period costs... do weigh on gross margins in fourth quarter and into '26... as we approach wafer outs in Idaho, that number will increase through '26."
— Mark Murphy, CFO

Assessment: This is the most important detail for modeling the next two quarters. Under-absorption costs are not a Q3 event that clears; they migrate into Q4 and persist into FY26, and start-up costs build on top. It is the single best reason to wait for the inflection rather than anticipate it.

HBM capacity and share into 2026

Given the repeated "sold out for 2025" framing alongside rising TAM, analysts asked whether Micron can keep expanding HBM capacity to hold share into 2026. Management reaffirmed exiting CY25 at HBM share in line with overall DRAM share, declined to give a 2026 share target, but expressed high confidence in capacity expansion, the 8-high to 12-high transition, and the HBM4 ramp.

Q: "As we look into next year... do you still have the capability to increase your capacity to maintain that market share in HBM, which is equal to our overall share of DRAM as you go into next year?"
— Chris Caso, Wolfe Research

A: "As we exit calendar '25, our share in HBM will correspond to our industry DRAM shares. So if you look at that run rate... in calendar year 2026, our share would be higher for the full year basis versus 2025... we feel very good about our HBM position, our close relationships with our customers, our execution."
— Sanjay Mehrotra, CEO

Assessment: The implied 2026 full-year HBM share step-up is the cleanest forward positive in the call. Management's reluctance to put a number on it is conservatism rather than weakness, and the HBM4 timing aligned to customer requirements is the right competitive posture.

What They're NOT Saying

1. A fiscal-Q4 gross margin number: "Up somewhat" is the only commitment. The refusal to quantify the recovery, in a quarter where the Q3 guide already disappointed, is the loudest silence on the call and the core of the Hold.
2. The dollar size of NAND under-absorption: Management described the direction and cadence of under-utilization costs but never sized them. Without that number, the margin bridge into FY26 has a hole in it.
3. 2026 HBM pricing and share specifics: The company is "in discussions" on 2026 HBM agreements but disclosed no pricing direction or a firm 2026 share target, leaving the most important 2026 variable qualitative.
4. How much calendar-Q2 pricing strength is tariff pull-forward: The tariff-versus-true-demand question was answered with demand color but no quantification of any pull-in, leaving a small risk that near-term pricing is borrowing from later quarters.
5. The 1.1-point GAAP-to-non-GAAP gross-margin gap: GAAP margin (36.8%) trails non-GAAP (37.9%); the adjustments were not walked through in detail on the call.

Market Reaction

• Pre-print setup: MU closed at $103.00 on March 20, up 22.4% year-to-date but down 3.5% over the trailing 30 days and roughly 33% below its 52-week closing high of $153.45. The stock entered the print having already round-tripped much of its 2024 AI rally.
• After-hours move: Shares fell about 4% in the initial after-hours reaction on the evening of the print.
• Next-day session: The decline deepened to -8.0% on March 21, closing at $94.72 (a $8.28 drop), with an intraday low near $93.70 (-9.0%). Volume was 63.2M shares, 3.3x the 30-day average. The S&P 500 was flat (+0.1%) that session, so this was company-specific.

The move was a textbook "sell the news" on the guide, not the print. A beat-and-record quarter could not offset a sequential gross-margin guide-down at the start of an up-cycle, and the 3.3x volume confirms it was a positioning reset rather than noise. The stock had rallied into the print on AI-memory optimism; the margin reset gave fast money a reason to de-risk, and the absence of a quantified fiscal-Q4 margin recovery removed the obvious bid that would have caught the falling knife.

Street Perspective

Debate: Is the margin guide a one-quarter air pocket or a structural ceiling?

Bull view: The dip is mix and timing. DRAM margins are healthy, HBM is scaling, NAND under-loading is a deliberate choice that protects long-term pricing, and fiscal Q4 is already guided up. The 36.5% guide is the trough, and blended margin re-rates from here as HBM mix compounds.

Bear view: Blended margin is structurally 10 points below the last up-cycle at the same revenue, NAND is a perennial drag with no fixed timeline, and under-absorption costs migrate into fiscal Q4 and FY26. "Up somewhat" is not an inflection, and the 50%-plus margin era may not return.

Our take: The bull has the better three-year argument and the bear has the better two-quarter argument. With under-absorption costs explicitly still ahead and no quantified recovery, the near-term holds the rating. We side with patience: the inflection is plausible but unproven.

Debate: HBM as a structural re-rating versus memory as a perennial cyclical

Bull view: HBM changes the character of the business: long-dated agreements, sold-out capacity, a rising trade ratio that tightens leading-edge DRAM, and a TAM that keeps getting raised. This is secular AI demand, and it deserves a higher and more stable multiple than memory has historically earned.

Bear view: Memory has promised "this time is different" before. HBM is real but is still DRAM under the hood; capacity additions across the industry eventually catch demand, and the cycle reasserts. Pay a cyclical multiple on mid-cycle earnings, not a secular one on a peak.

Our take: HBM genuinely improves the demand visibility and customer stickiness of the DRAM franchise, so the business deserves some re-rating. But the NAND half of the company remains fully cyclical, and the blended entity is not yet a secular grower. The right multiple is between the two camps, which is consistent with a Hold at the current price.

Debate: NAND supply discipline, fixable or perennial?

Bull view: Industry-wide under-loading, capex cuts, and delayed node transitions are finally rationalizing NAND supply. Micron's structural 10%-plus wafer reduction leads the discipline, and pricing inflects in calendar Q2. NAND swings from drag to contributor by FY26.

Bear view: NAND discipline has repeatedly broken down once pricing recovers, because the marginal producer always defects. Micron under-loading its own fabs is just ceding share and eating under-absorption costs while waiting for competitors who may not cooperate.

Our take: NAND is the swing factor for the rating. If calendar-Q2 pricing inflects and holds, the margin story improves materially and the stock re-rates; if discipline cracks, NAND stays a drag and the 8% sell-off looks cheap in hindsight. We need a quarter of evidence either way.

Model Update & Valuation Framework

Valuation framework: At the $94.72 post-print close, Micron has de-rated roughly 38% from its 52-week high. Annualizing the fiscal-Q2/Q3 non-GAAP EPS run-rate (around $1.56–$1.57 per quarter) puts the stock in the low-to-mid teens on a forward earnings basis, which is not demanding for a company at an apparent cyclical inflection with a structural HBM growth driver. The bull case is that earnings power is heading meaningfully higher as HBM scales and NAND heals, in which case the trailing multiple understates the de-rating. The bear case is that consensus FY26 earnings are too high if the margin recovery stalls, in which case the multiple is a value trap on peak-ish numbers.

Where we land: The risk/reward is genuinely balanced at this price. The valuation is undemanding and the structural driver is real, but the one variable that decides the next two quarters (gross margin) was just guided the wrong way with under-absorption costs still ahead. We initiate at Hold and would turn constructive on evidence the blended margin has inflected (a clean fiscal-Q4 step-up, or a NAND pricing inflection that holds), or on a further de-rating that prices in the bear case.

Thesis Scorecard: Establishing Coverage

As this is our initiation of Micron, the scorecard below establishes the bull and bear pillars we will carry and grade each quarter, marked to what fiscal Q2 revealed.

Overall: The structural bull pillars are all confirmed and the business has a genuine AI-demand engine, but the two near-term bear pillars (NAND and blended margin) are precisely what the guide put in the spotlight. The thesis is balanced at initiation.

Action: Initiate at Hold. Hold the position; do not chase the de-rating yet. The clearest catalysts to upgrade are a fiscal-Q4 gross-margin step-up that confirms the inflection and a NAND pricing recovery that holds into calendar Q2.

Bottom Line

Micron did the hard part: it beat, set a data-center DRAM record, and put HBM over $1B ahead of its own plan. The AI-memory franchise is no longer a promise, it is a P&L line that is compounding. But the quarter also exposed the catch in owning Micron today, which is that the same company running a world-class DRAM and HBM business is choosing to under-earn in NAND, and the blended gross margin is paying the bill. Guiding next-quarter margin down at the start of an up-cycle, and committing only to "up somewhat" after that, is why a beat-and-record quarter shed 8%.

We are initiating at Hold, not because the story is broken but because the price already reflects a lot of the good news and the single variable that moves the stock from here was just guided the wrong way. This is a stock to own once the margin inflects, and the de-rated valuation means we will not have to be heroic to get constructive. For now, the right posture is patience: let the fiscal-Q4 margin and calendar-Q2 NAND pricing tell us whether the trough is behind the company before we pay up for the recovery.