Micron FQ3 FY2025: The Margin Inflection Arrives, and We Upgrade

Results vs. Consensus

Quality of Beat

• Revenue: Broad-based and high quality. Growth came across every end market, with DRAM up 15% sequentially on 20%-plus bit-shipment growth and NAND up 16% on mid-20s bit growth. This was a volume-led record (bits up sharply, blended price still modestly down on consumer mix), which is the healthiest kind of memory beat because it shows demand, not just price.
• Margins: The 39.0% gross margin beat the guide by 250 bps, and management was explicit that the defining factor was better-than-expected pricing in both DRAM and NAND, plus good cost performance. This is the cleanest possible margin beat: pricing came in ahead of a deliberately cautious guide, not a one-time benefit.
• EPS: The $0.31 beat versus consensus was helped modestly by a 12.3% tax rate (below guide on discrete items), but the operating story did the heavy lifting: operating income of $2.5B at a 26.8% margin, up 13 points year-on-year. EPS grew 22% sequentially and more than 200% year-on-year.

Segment Performance

Revenue by Technology

Revenue by Business Unit

The whole portfolio fired

Unlike the prior quarter, where CNBU's record masked sharp declines in Mobile, Storage, and Embedded, this quarter every business unit grew sequentially. Mobile rebounded 45% as handset customer inventories cleared and content growth kicked in; Embedded grew 20% on an industrial and consumer recovery; Storage and CNBU both advanced. This is what a cyclical trough turning into an up-cycle looks like in the segment table: the consumer-facing units that were the drag in fiscal Q2 became contributors in fiscal Q3.

"Higher sequential revenue was driven by growth across our end markets, including record data center revenues and strong sequential growth in consumer-oriented markets." — Mark Murphy, CFO

Assessment: The breadth matters more than any single line. A broad-based sequential recovery is far harder to dismiss as a one-product story than a DRAM-only print, and it reduces the risk that a single end-market air pocket derails the trajectory. The cyclical recovery thesis from last quarter is now showing up in the numbers across the board.

Data center: the structural core

Data-center revenue more than doubled year-on-year and set a quarterly record, with data-center DRAM hitting a record for the fourth consecutive quarter. Beyond HBM, the high-capacity DIMM and low-power server DRAM franchise has grown roughly fivefold year-on-year to multiple billions in fiscal 2025, and Micron became the clear #2 brand in data-center SSDs by share in calendar Q1.

Assessment: Micron now has three distinct multi-billion-dollar data-center franchises (HBM, high-capacity/LP server DRAM, and data-center SSD), each gaining share. That diversification within the AI build-out is exactly what de-risks the structural bull case from being a single-product HBM bet.

Key Topics & Management Commentary

Overall Management Tone: Confident and forward-leaning, a clear step up from the prior quarter's defensiveness on margin. Where the fiscal-Q2 call was dominated by management defending the gross-margin trajectory, this call let the numbers carry the message and pivoted to the structural story: HBM scaling, the $200B U.S. investment plan, and a business-unit reorganization around AI. The one place management stayed deliberately non-committal was fiscal-Q1 FY26 guidance, where they would only say margins "can be up."

1. The Gross Margin Inflection

Fiscal-Q3 gross margin of 39.0% beat the 36.5% guide by 250 bps, and management guided fiscal Q4 to 42%. The driver was better-than-expected pricing in both DRAM and NAND, with the sequential step-up to 42% layered on favorable mix (more DRAM than NAND, more data-center than consumer).

"Prices were down, as I said... but we had better-than-expected pricing that drove the margin improvement versus expectations... third quarter to fourth quarter, we have favorable mix effects that is helping drive margins up to the current guide of 42%." — Mark Murphy, CFO

Assessment: This is the pillar that flips our rating. A two-quarter march from 36.5% guided to 39% printed to 42% guided is a genuine inflection, and it is being driven by the two highest-quality sources, pricing and mix, rather than cost timing. The margin question that anchored our Hold has been answered.

2. HBM Ahead of Plan, Share Parity Pulled Forward

HBM grew nearly 50% sequentially to a $6B-plus annualized run rate, with HBM3E 12-high yields ramping faster than 8-high did and shipment crossover to 12-high expected in fiscal Q4. Management now expects to reach HBM share in line with overall DRAM share in the second half of calendar 2025, potentially earlier than the prior "exiting the year" framing.

"We are already at... more than $6 billion run rate here with our HBM... it could be that we do end up getting to our industry share for HBM share earlier than that... our 12-high yield ramp is actually going faster than our 8-high yield ramp." — Sanjay Mehrotra, CEO

Assessment: Reaching DRAM-parity HBM share, possibly ahead of schedule, would cement Micron as a genuine co-leader in the highest-value memory product. The faster 12-high yield ramp is the operational proof that Micron's HBM execution is not a one-generation fluke, which is what underwrites the HBM4 ramp.

3. HBM4 Sampled, 2026 De-Risking

Micron has delivered HBM4 samples to multiple customers, with volume production planned for calendar 2026. HBM4 uses the proven, cost-effective 1-beta node plus an internally developed advanced CMOS logic base die, delivering over 2 TB/s per stack (60%-plus higher performance than HBM3E) and 20% lower power.

"Micron has delivered samples of HBM4 to multiple customers and expect to ramp volume production in calendar 2026... HBM4 leverages our well-established 1-beta DRAM technology along with an internally developed and manufactured advanced CMOS logic base die." — Sanjay Mehrotra, CEO

Assessment: Building HBM4 on the mature 1-beta node, rather than a bleeding-edge node, is a smart risk/cost choice that should support both yield and margin at launch. With samples already out, the most important 2026 product is de-risked a year ahead, which lengthens the visibility on the structural growth story.

4. The Rising Trade Ratio Is a Structural Tailwind

Management quantified the HBM trade ratio at approximately 3:1 for HBM3E, greater than 3 for HBM4, and rising toward 4 for HBM4E. Because each HBM bit consumes roughly three-plus bits of equivalent standard-DRAM wafer capacity, HBM growth structurally tightens non-HBM leading-edge DRAM supply.

"HBM4's trade ratio is greater than 3... as we are going from HBM3E to 4 to 4E, the trade ratio is going from about 3 towards 4... this puts pressure on the non-HBM supply in the industry as well." — Sanjay Mehrotra, CEO

Assessment: The trade ratio is the underappreciated mechanism that makes this cycle different. As long as HBM grows, it siphons wafer capacity away from standard DRAM, keeping the broader DRAM market tight even without supply discipline. It is a self-reinforcing tailwind for blended pricing.

5. NAND Heals From the Bottom Up

NAND grew 16% sequentially with bit shipments up mid-20s percent, set a record data-center SSD share (clear #2 brand), and reached a record QLC mix. Management noted the NAND leading edge is now fully utilized even as part of the trailing-node base remains under-loaded, and the structural 10% wafer-capacity reduction is largely complete.

"As that capacity has come down, of course, our underutilization has come down as well, although part of NAND continues to remain underutilized. I must note here that, of course, leading edge of NAND is fully utilized." — Sanjay Mehrotra, CEO

Assessment: NAND is moving from drag toward contributor faster than expected. A fully utilized leading edge, record SSD share, and rising QLC mix are exactly the signals that the supply-discipline strategy is working. This is the bear pillar from last quarter actively weakening.

6. The $200B U.S. Investment Plan

Micron announced plans to invest approximately $200B in the U.S. ($150B manufacturing, $50B R&D) over 20-plus years, including roughly $30B beyond prior plans: a second Boise fab, a Manassas, Virginia expansion, and future U.S. advanced packaging. First Idaho fab (ID1) wafer output is expected in the second half of calendar 2027.

Assessment: The headline number is a two-decade figure, not a near-term capex shock, so it does not change the FY25 spend (held at ~$14B). Strategically it secures domestic leading-edge capacity, deepens the policy relationship, and supports the long-dated HBM packaging roadmap. A long-term positive that is easy to over-read as a near-term cash drain; it is not.

7. Cash Generation Inflects

Operating cash flow was $4.6B and free cash flow was $1.9B, the highest in over six years, even with $2.7B of capex. Inventory fell to 139 days (down 19), net debt dropped to about $3B, and management reiterated returning capital via the dividend and opportunistic buyback.

"Free cash flows in the quarter were over $1.9 billion, the highest quarterly amount in over 6 years... net debt now down to $3 billion." — Mark Murphy, CFO

Assessment: The cash inflection removes the last reservation from the prior quarter. Micron can fund the HBM build, de-lever, and return capital simultaneously, which means the heavy capex cycle is no longer a free-cash-flow concern. This is the financial corroboration of the operating turn.

8. Business-Unit Reorganization Around AI

Micron completed a reorganization of its business units around market segments to focus resources on AI opportunities, with new-segment revenue and margin reporting beginning in fiscal Q4. This is a structural signal of where the company sees its growth concentrating.

Assessment: Re-orienting the org chart around AI end markets is the right move for capital and talent allocation, though it will create a one-quarter discontinuity in segment comparability when the new structure debuts. Watch fiscal Q4 for the first look at the new segment economics.

Guidance & Outlook

The fiscal-Q4 guide is a clean raise on every axis. Revenue at $10.7B would grow 15% sequentially to another record, driven by DRAM, with the margin guide to 42% reflecting favorable mix (more DRAM than NAND, more data center than consumer) on top of continued pricing strength. EPS at $2.50 would be up 31% sequentially.

Implied trajectory: Management would not guide fiscal Q1 FY26 but said gross margins "can be up," citing very tight inventories (leading-edge DRAM and pockets of legacy) that will create bit constraints. The framing is one of supply tightness limiting volume, not demand softness, which is the favorable kind of constraint.

Street at: The $10.7B revenue guide is roughly $0.8B above the ~$9.9B consensus, and the $2.50 EPS guide is well ahead of where the Street sat. This is a genuine upside surprise, not an in-line guide.

Watch item: The FY26 tax rate steps up to the high-teens (from ~13%) on Singapore's adoption of the global minimum tax, a real but quantifiable headwind to FY26 EPS that the bull case has to absorb.

Analyst Q&A Highlights

Is 42% the new baseline for gross margin?

The dominant line of questioning probed whether the margin step-up is durable or pricing-dependent. Management framed fiscal Q3 as a rebaseline driven by better pricing and cost, with fiscal Q4's move to 42% layered on favorable DRAM-over-NAND and data-center-over-consumer mix, and declined to call a fiscal-Q1 number while expressing positive bias.

Q: "I wanted to talk about gross margins. So first, what is driving the upside sequentially? And then is this the new baseline for gross margins?"
— Vivek Arya, Bank of America Securities

A: "The defining factor was we had... better-than-expected pricing that drove the margin improvement versus expectations... third quarter to fourth quarter, we have favorable mix effects... So yes, it's a continued good market backdrop. We're going to continue to focus on pricing."
— Mark Murphy, CFO

Assessment: Management was careful not to declare a permanent baseline, which is appropriate for a cyclical, but the combination of pricing plus mix plus tight inventories points to margins holding or rising near-term. The refusal to over-promise is a feature, not a hedge.

HBM TAM scaling with the accelerator market

A question explored whether HBM demand keeps pace with the rapidly rising accelerator TAM, and whether there is a ceiling on HBM attach. Management framed HBM growing from roughly $18B in 2024 to ~$35B in 2025, with 2026 HBM bit-demand growth significantly exceeding overall DRAM demand growth as the industry transitions to 12-high and HBM4.

Q: "How do you see the HBM TAM scaling with that market? And then is there some sort of limit or like asymptote that we begin to reach in terms of HBM attached to these GPUs and these custom ASICs?"
— Timothy Arcuri, UBS

A: "HBM is growing from last year about $18 billion in revenue to approximately $35 billion in calendar year '25... in calendar year '26, HBM bit demand growth will significantly exceed the overall DRAM industry demand growth... the trajectory of HBM in terms of the value proposition in these accelerators continues to be strong."
— Sanjay Mehrotra, CEO

Assessment: The roadmap from 12-high to HBM4 to HBM4E to in-2028 customization gives HBM a multi-year value-and-content growth runway, with no near-term asymptote in sight. This is the core of why the business deserves a structural re-rating, not just a cyclical bounce.

HBM4 trade ratio and GPU-versus-ASIC economics

A technical question sought the HBM4 trade ratio and whether HBM margins differ between GPU and ASIC customers. Management quantified the trade ratio above 3 for HBM4 and rising toward 4 for HBM4E, and characterized HBM value as strong across both GPU and ASIC platforms without differentiating.

Q: "When you look into HBM4, given the double I/O count through silicon via, what kind of trade ratio should we expect for HBM4? And also... is there a difference in HBM bits and margin profile between selling to GPU versus ASIC customers?"
— Krish Sankar, TD Cowen

A: "HBM4's trade ratio is greater than 3... as we are going from HBM3E to 4 to 4E, the trade ratio is going from about 3 towards 4... HBM commands high value, whether it's in GPU accelerators or ASIC-based accelerators. And that value proposition is only growing."
— Sanjay Mehrotra, CEO

Assessment: That HBM economics are attractive across both GPU and ASIC paths matters for diversification: Micron benefits whether the accelerator market tilts toward merchant GPUs or custom silicon, which insulates the franchise from any single customer's roadmap.

NAND utilization and the path back to full loading

Questions pressed on where NAND utilization stands and whether Micron must bring capacity back on to avoid a margin headwind. Management indicated the structural 10% capacity reduction is largely complete, under-utilization has fallen as a result, the leading edge is fully utilized, and node conversions (G9) will be managed at a measured pace tied to demand.

Q: "On the NAND side, where does utilization stand today?... as you move into next year, what is your plan for utilization? Is that going to be market dependent? Or at some point, do you need to start bringing on utilization just from the fact that there's a gross margin headwind?"
— Thomas O'Malley, Barclays

A: "Toward the end of fiscal '25, our NAND overall capacity would be down versus end of fiscal '24 by about 10%... as that capacity has come down, our underutilization has come down as well... leading edge of NAND is fully utilized."
— Sanjay Mehrotra, CEO

Assessment: The under-absorption drag that loomed over the prior quarter is shrinking by design. With the structural cut largely done and the leading edge full, NAND's margin headwind is converting into a tailwind, which is central to the upgrade.

2026 HBM supply and pricing commitments

A question asked where 2026 HBM supply and pricing negotiations stand and whether customer demand exceeds Micron's planned supply. Management reaffirmed HBM sold out for 2025, said it is mid-negotiation for 2026 as customers finalize their 12-high versus HBM4 platform mix, and reiterated 2026 HBM bit-demand growth significantly outpacing DRAM.

Q: "One year later, where are you on your negotiations for your calendar '26 HBM supply and pricing discussions? Is the team's supply outlook for next year fully committed, too?... is it exceeding your forecasted supply capability?"
— Harlan Sur, JPMorgan

A: "Our HBM is sold out for 2025... we are working closely with our customers as their AI accelerated platforms are continuing to evolve... we see, with respect to bit demand growth, strong trend in 2026, bit demand growth for HBM significantly exceeding DRAM demand growth."
— Sanjay Mehrotra, CEO

Assessment: Management did not yet commit 2026 supply as "sold out," because customers are still finalizing their 12-high versus HBM4 mix. That is a timing nuance, not a demand concern, but the 2026 pricing outcome remains the most important unquantified variable for FY26 estimates.

Long-term capex intensity

A question asked whether the historical ~35%-of-revenue capex level is the right FY26 guide or whether Micron must invest ahead of it. Management pointed to a generational tech-transition opportunity requiring greenfield capacity (driven by the HBM trade ratio), held FY25 capex at ~$14B with lumpy timing, and emphasized continued free-cash-flow generation and balance-sheet improvement.

Q: "I wonder if you could talk in terms of your long-term CapEx plans and the sort of 35% of revenue levels. Is that kind of a reasonable guideline to think about for fiscal '26?"
— Joseph Moore, Morgan Stanley

A: "We have a generational tech transition opportunity in front of us that Micron is exceptionally well positioned for... with the silicon requirements of HBM and with trade ratio, we need to build greenfield capacity... we anticipate generating free cash flow in the fourth quarter... net debt now down to $3 billion."
— Mark Murphy, CFO

Assessment: The capex will stay elevated to fund the HBM-driven greenfield build, which is the right use of capital at a generational inflection. With FCF positive and net debt at $3B, the balance sheet can carry it. Capex intensity is a watch item, not a worry, as long as the cash keeps coming.

What They're NOT Saying

1. A fiscal-Q1 FY26 gross-margin number: Management would only say margins "can be up," declining to extend the strong trajectory into an explicit number. The reluctance is reasonable for a cyclical, but it leaves the durability of 42%-plus unconfirmed.
2. 2026 HBM pricing direction: Negotiations are "in the middle"; no pricing direction or firm 2026 sold-out commitment was given. The single most important FY26 estimate variable remains qualitative.
3. The magnitude of the FY26 tax headwind: Management flagged a high-teens FY26 tax rate from the Singapore global minimum tax but did not size the EPS impact, which is a real offset to the margin tailwind.
4. How much of the demand strength is durable versus tariff pull-in: Management characterized tariff pull-ins as "relatively modest," but did not quantify them, leaving a small risk that a sliver of the back-half strength is borrowed.
5. New-segment economics: The reorganized business units will not report revenue and margin until fiscal Q4, so the market is one quarter away from seeing the AI-segment profitability the reorg is meant to highlight.

Market Reaction

• Pre-print setup: MU closed at $127.25 on June 25, up 51.2% year-to-date and 36.3% over the trailing 30 days, near the high end of its 52-week range. A great deal of recovery optimism was already in the stock heading into the print.
• After-hours move: Shares initially popped in extended trading on the beat-and-raise before paring most of the gains as the print was digested against the run-up.
• Next-day session: The stock closed -1.0% on June 26 at $126.00, a modest fade, on 62.7M shares (3.1x the 30-day average). The S&P 500 was up 0.8%, so MU lagged the tape slightly.

The muted reaction is a story about positioning, not results. A clean beat-and-raise that confirmed the margin inflection met a stock that had already rallied more than 50% year-to-date and 36% in a month, so fast money took profits into the print. We read the lack of a sell-off on a stock this extended as constructive: a "priced in" tape that holds flat after a strong run is healthier than one that gaps and reverses. The fundamental news was unambiguously good even if the one-day price action was not dramatic.

Street Perspective

Debate: Is the margin inflection durable or a pricing-cycle peak?

Bull view: 39% printed, 42% guided, driven by pricing and mix, with tight inventories and the HBM trade ratio structurally supporting DRAM pricing into FY26. The margin is rebaselining higher and HBM mix keeps lifting it.

Bear view: Memory margins are mean-reverting; 42% is near a cyclical peak, supply will catch up, and the Singapore tax plus elevated capex cap the EPS upside. Pay a trough multiple on peak earnings.

Our take: The bull has the better evidence right now. The trade-ratio mechanic genuinely differentiates this cycle, and the margin is being driven by mix shift toward HBM and data center that is structural, not just spot pricing. We do not assume 42% is permanent, but the path of least resistance for the next several quarters is up or flat, not a cliff.

Debate: How much should HBM re-rate the multiple?

Bull view: HBM converts memory into a visibility business: annual agreements, sold-out capacity, a rising trade ratio, and a TAM doubling year-on-year. Micron at DRAM-parity HBM share deserves a materially higher multiple than historical memory.

Bear view: HBM is still DRAM, the multiple should not de-cyclicalize, and after a 50%-plus year-to-date run the easy re-rating is done. Future returns track earnings, and earnings are cyclical.

Our take: Both are partly right. The multiple should expand somewhat to reflect better demand visibility and customer stickiness, but the business is not a secular grower end to end. The key is that forward earnings power is rising fast enough that even a flat multiple produces upside, which is the crux of our Outperform.

Debate: Has the stock already priced the good news?

Bull view: Forward estimates are still chasing the guide higher; at a $2.50 fiscal-Q4 EPS run rate the stock trades at a low-teens forward multiple, undemanding for a company compounding earnings and free cash flow.

Bear view: Up 51% year-to-date with a flat reaction to a beat-and-raise signals exhaustion; the marginal buyer is gone and the risk is a pause or pullback.

Our take: The flat reaction reflects positioning, not valuation. With EPS guided to $2.50 and rising into FY26 on HBM4 and pricing, the multiple is compressing even as the stock rises. That is the setup we want to own, and it is why we are willing to upgrade despite the year-to-date move.

Model Update & Valuation Framework

Valuation framework: At the $126.00 post-print close, the fiscal-Q4 guide of $2.50 non-GAAP EPS annualizes to a run rate approaching $10, putting the stock in the low-teens on forward earnings before accounting for further FY26 growth from HBM4 and pricing. That is an undemanding multiple for a company at a margin inflection with three multi-billion-dollar data-center franchises gaining share and a six-year free-cash-flow high. The bull case is that FY26 earnings power is meaningfully above current consensus as HBM scales and NAND heals, in which case the stock is cheap even after the year-to-date run. The bear case is cyclical mean reversion plus the FY26 tax headwind, which would cap EPS and the multiple.

Where we land: The variable that held our rating (gross margin) has inflected up, NAND is healing, and the forward earnings power is re-rating faster than the price. We upgrade to Outperform. We would revisit the rating if the margin trajectory stalls into fiscal Q1 FY26, if 2026 HBM pricing negotiations disappoint, or if the stock runs well ahead of the earnings revisions.

Thesis Scorecard Post-Earnings

Overall: Thesis materially strengthened. All three bull pillars are confirmed and the two near-term bear pillars (NAND and blended margin) have moved from headwind toward tailwind. The capital-intensity pillar remains the only open question, and it is well-covered by the cash inflection.

Action: Upgrade to Outperform from Hold. Add on the margin inflection. The catalysts to watch are fiscal-Q4 delivery of the 42% margin, the first new-segment reporting, and the 2026 HBM pricing outcome.

Bottom Line

Last quarter we initiated at Hold and said plainly what would change our mind: evidence that the blended gross margin had inflected. Micron just delivered exactly that, and then some. It beat on every line, printed a 39% gross margin against a feared 36.5% guide, raised fiscal Q4 to a record $10.7B at a 42% margin and $2.50 of EPS, and generated the most free cash flow in over six years. The consumer-facing units that dragged the prior quarter all rebounded, NAND is healing from the leading edge down, and HBM is running ahead of plan with HBM4 already sampled for 2026.

We are upgrading to Outperform. The stock has run more than 50% year-to-date and the one-day reaction was a shrug, but the rating is a twelve-month view, and on that horizon the forward earnings power is re-rating faster than the price. The margin question that anchored our caution has been answered by the numbers, the structural HBM story keeps compounding, and the valuation is undemanding on a guide that consensus is still chasing. The risk/reward has shifted to the upside, and we are positioned for it. The two things we will watch closely are whether the margin holds into fiscal Q1 FY26 and where 2026 HBM pricing lands; for now, the trajectory is firmly Micron's to lose.