Micron FQ4 FY2026: The Contract Book Jumps to $150B, the Guide Beats by 9%, and the Tape Shrugs

Results vs. Consensus

FQ4 FY2026 Scorecard

Year-Over-Year Comparison

Quarter-Over-Quarter Comparison

Fiscal 2026 in Full, and the Four Quarters That Built It

The four quarters sum exactly to the fiscal year on revenue ($133.188B), gross profit, operating expense, operating income and net income. Fiscal-year EPS of $75.52 against a four-quarter sum of $75.51 is share-count rounding, not a reconciling item. For the record: fiscal 2026 revenue was 3.56 times fiscal 2025's, gross margin expanded 4,020 basis points, and EPS rose 811.0%. One calendar note: fiscal Q4 2026 was a 14-week quarter and fiscal 2026 a 53-week year, so both carry one more week than a standard period.

Revenue. Still a price story, but a visibly different one. DRAM revenue of a record $39.8B rose 27% sequentially on mid-single-digit bit growth and high-teens percent price increases. NAND revenue of a record $14.1B rose 42% on roughly 10% bit growth and roughly 30% price increases. Compare that with last quarter, when DRAM prices rose in the low-60s percent and NAND in the mid-80s. The price slope has decisively moderated, exactly as management told us in June it would, and revenue still grew 30.8% sequentially. What makes this quarter different is where the pricing is being set: management has now completed agreements for the vast majority of calendar-2027 HBM bits at significantly higher prices, resetting on January 1, and 26 take-or-pay agreements now cover more than 35% of revenue through 2030. The marginal price is increasingly contracted rather than quoted.

Margins. The 87.0% non-GAAP gross margin is a company record and 210 basis points above FQ3, attributed to higher pricing and execution, partially offset by mix. The more interesting line is operating margin, which rose 110 basis points to 82.3% despite operating expense jumping 69.2% sequentially. That is the arithmetic of operating leverage at an extreme: revenue added $12.8B while cost added $1.05B. Operating expense is now 4.7% of revenue against 3.7% last quarter, which is still a rounding error by any normal standard and is the reason the fiscal 2027 cost build, real as it is, does not move the operating-margin needle much. On the GAAP side a $500M patent license charge is new in the quarter, widening the GAAP-to-non-GAAP operating-expense gap to $728M.

EPS. Clean. The $33.42 non-GAAP figure sits $0.55, or under 2%, above the $32.87 GAAP figure, and the gap is almost entirely stock-based compensation plus the patent charge. The 15% non-GAAP effective tax rate is in line with the run rate. Below the line, interest expense was zero for the quarter against $124M a year ago, while interest income was $575M. Micron now earns more in a quarter from its cash than it used to pay in a year on its debt. Cash conversion was the best in company history: $43.97B of operating cash flow and $33.20B of adjusted free cash flow after $10.77B of net capex, a free-cash-flow margin of 61%.

Segment Performance

Revenue by Technology

The slope is the story. Three months ago DRAM prices rose in the low-60s percent sequentially and NAND in the mid-80s. This quarter it was high-teens and roughly 30%. Both are still extraordinary absolute increases, and both represent a sharp deceleration from a parabolic base. The level keeps rising; the second derivative has turned, on schedule.

Revenue by Business Unit (all four at records)

The four units sum to $54.223B against reported revenue of $54.229B; the $6M residual is unallocated. Margin point-changes are sequential.

Core Data Center (CDBU): the largest unit in the company, at a 90% gross margin

CDBU overtook Cloud Memory this quarter to become Micron's biggest business, on revenue up 56.2% sequentially and more than eleven-fold year-on-year, at a 90% gross margin and an 85% operating margin. This is the conventional server DRAM and data center SSD unit, the part of the portfolio that a year ago was the lowest-margin piece of the data center story at a 41% gross margin. Inside it, data center SSDs are now the standout.

"Data center SSD revenue in fiscal Q4 was nearly $10 billion, more than 10 times the year-ago quarter, and was over two-thirds of total company NAND revenue."
— Sanjay Mehrotra, Chairman and CEO

Assessment: A unit that was earning 41% gross margins a year ago is earning 90% today, and it is now the largest in the company. That is the cleanest single illustration of what the AI build has done to conventional memory, and it is also the piece most exposed if pricing normalizes, because management attributed the sequential margin gain to higher pricing first and favorable mix second, with no technology or cost lever named. The durable part is the share gain: a fifth consecutive year of record data center SSD market share, won on controller, firmware and qualification work that does not unwind when prices fall. We would separate the two when the cycle turns.

Cloud Memory (CMBU): the HBM unit, and the only margin that did not improve

CMBU grew 18.3% sequentially to $16.283B, the second-slowest growth of the four units, with gross margin flat at 83% and operating margin down two points to 76%. The explanation is counterintuitive and important: margin was held back by a higher HBM mix, because HBM still carries a lower gross margin than conventional DRAM at current spot levels.

"CMBU gross margins were 83%, flat sequentially, driven by higher pricing, offset by higher HBM mix."
— Mark Murphy, Chief Financial Officer

Management addressed this directly on the call. HBM pricing for calendar 2027 has been reset substantially higher, and the explicit purpose is to close the gap.

"In HBM, our revenue in fiscal Q4 2026 grew faster than total company revenue in the quarter, as we ramp HBM shipments across a growing number of customers. We have completed agreements for the vast majority of our calendar 2027 HBM bit supply with significant price increases year over year, narrowing the gross margin gap with conventional DRAM."
— Sanjay Mehrotra, Chairman and CEO

Assessment: The HBM margin inversion that we flagged two quarters ago as a late-cycle tell persists, and this quarter it is explicitly costing the Cloud Memory unit two points of operating margin. The January 1 price reset is the fix, and it is already contracted rather than hoped for, which is the material difference from last quarter. The read-through matters for modeling fiscal 2027: a mix shift toward HBM, which the company expects to keep outgrowing conventional DRAM through calendar 2028, is a margin headwind at 2026 pricing and a margin tailwind at 2027 pricing. The reset lands inside FQ2 FY27, which is part of why management can call FQ1 the floor.

Mobile & Client (MCBU): a 90% gross margin on falling bits, for the second quarter running

MCBU posted a 90% gross margin and an 88% operating margin, the highest operating margin in the company, on revenue up 13.8% sequentially. It achieved that while shipping fewer bits for the second consecutive quarter.

"We did see sequential bit decline in our mobile and client business unit, but we did see revenue growth, obviously, with both higher pricing and favorable mix."
— Manish Bhatia, President and Chief Operating Officer

The constructive offset is technology adoption: nearly half of MCBU revenue in the quarter came from 1-gamma products, and Micron is now shipping 6-channel 1-gamma LPDDR5X for flagship phones and has design wins with every Tier 1 OEM for AI workstations.

Assessment: An 88% operating margin in the consumer-facing unit, earned on declining unit volume, is the sharpest evidence in the entire print that consumer memory pricing has overshot end demand. Two consecutive quarters of bit decline is no longer a data point, it is a trend, and management now concedes potential double-digit unit declines in both PCs and smartphones. The mitigant is real and specific: the bits are being deliberately reallocated to contracted data center demand that earns more, and the mix inside MCBU is moving to premium devices where content growth offsets unit loss. But this is where demand destruction shows up first, and it is showing up.

Automotive & Embedded (AEBU): the quarter's biggest surprise versus expectations

AEBU revenue of $6.824B rose 47.3% sequentially and 376% year-on-year at an 84% gross margin, up five points. Pre-print estimates for this unit sat near $4.7B, so it cleared expectations by roughly 45%. Management framed the medium-term driver as physical AI: memory content in Level 4 and higher autonomous vehicles typically exceeds 200GB with storage in multiple terabytes, each more than an order of magnitude above today's Level 2+ and Level 3 systems, and humanoid robots are expected to carry comparable requirements.

Assessment: The near-term AEBU result is a pricing and allocation outcome, not a physical-AI outcome, and it should be read that way. Automotive and industrial buyers are paying up because memory is scarce, not because robots are shipping. The physical-AI framing is a credible long-dated option, with 1-gamma LPDDR6 already sampling to those markets, but it is not what produced a $6.8B quarter. Treat the 47% sequential move as scarcity rent and the content thesis as a 2029-and-beyond call.

Key KPIs

Key Topics & Management Commentary

Overall Management Tone: Settled, and noticeably less preoccupied with defending the cycle than on any call in the past two years. Where June's call argued that the business had structurally changed, this one treated that as established and spent its airtime on contract mechanics, cleanroom construction schedules and fiscal-2027 cost build, the vocabulary of a company planning capacity rather than defending a price. The one place the tone turned careful was gross margin, where management volunteered a sequential decline, named its cause, and committed to a floor, which is a more confident posture than it first appears. Analyst pushback was narrow and concentrated almost entirely on capital return, where management declined to size anything beyond a December start date.

1. The contract book grew by about half, and now runs to 2031

Micron signed ten new strategic customer agreements in the quarter, taking the total from 16 to 26. The structural disclosures all stepped up together: remaining performance obligations rose from roughly $100B to approximately $150B, customer financial commitments from $22B to $32B, and the agreements now extend past the original 2030 horizon.

"To date, we have signed 26 SCAs, which we currently estimate to be over 35% of our revenue through 2030. Three-quarters of this estimated revenue has a defined pricing framework, a majority of which have pricing bands with floor and ceiling prices."
— Sanjay Mehrotra, Chairman and CEO

"Customers want SCA-assured supply beyond 2030, and we have now signed SCAs that extend into 2031, as well as one-year extensions to 2031 for two agreements. … For the 26 signed SCAs and extensions, financial commitments from customers have increased to $32 billion, the vast majority of which are cash deposits."
— Sanjay Mehrotra, Chairman and CEO

The CFO was careful about what the RPO figure does and does not represent, and reaffirmed the floor-margin claim that drove our June upgrade.

"All SCAs have take-or-pay contracted volumes, and RPO reflects the contract value for only SCAs that have a determined pricing framework, which can be either a fixed price or subject to a pricing floor and ceiling. RPO is based on committed volumes and minimum pricing and is inherently conservative. … As mentioned in our last earnings call, even at floor prices, we expect margins meaningfully above any prior cycle peak margins."
— Mark Murphy, Chief Financial Officer

Assessment: This is the pillar we upgraded on in June, and one quarter later it is roughly 50% bigger. A $150B RPO measured at minimum pricing on take-or-pay volumes is a genuinely different asset from a $100B one, and the extension into 2031 matters more than the dollar increase, because it is customers volunteering to lengthen commitments in a market where they already cannot get enough supply. The honest limit is unchanged: the floor is still unquantified, and a floor that merely beats prior-cycle peak margins can sit far below 87%. We now know the shape of the protection with precision and its level not at all.

2. The fiscal Q1 guide: $61.5B and $38.15, roughly 9% above the Street

Guidance put the next quarter above $60B for the first time, and did it on a revenue base that had already quadrupled in a year.

"We expect FQ1 revenue to be a record $61.5 billion, plus or minus $1.5 billion; gross margin to be approximately 86.25%; and operating expenses to be approximately $2.06 billion. … We expect fiscal 2027 to be another record year, with sequential revenue growth each quarter."
— Mark Murphy, Chief Financial Officer

The midpoint is 13.4% above the quarter just reported and roughly 8.8% above the $56.5B the Street carried. On $38.15 of EPS against approximately 1.15 billion diluted shares and a 15.5% tax rate, the guide implies $53.0B of gross profit and $51.0B of operating income, an 82.9% operating margin. Operating margin therefore expands 60 basis points sequentially even as gross margin contracts, because the FQ4 incentive-compensation and community-investment spend does not repeat at the same level in operating expense.

Assessment: A guide 9% above consensus, on a revenue base that has already quadrupled in a year, is not a company managing expectations down. The composition is what should reassure: operating margin expands, the revenue step is 13% rather than the 31% just delivered, and the whole thing rests on a book where more than three-quarters of the year's output is already committed. This is the first guide of this cycle that reads like a plan rather than a price extrapolation.

3. The first guided gross-margin decline of the cycle, and why it is a cost story

Gross margin guides to 86.25% against the 87.0% just printed, the first sequential decline management has guided in this entire up-cycle. It volunteered the cause before anyone asked, and attached a floor commitment to it.

"We anticipate fiscal Q1 to be the floor for gross margins in fiscal 2027. … We expect higher gross margins beyond fiscal Q1 for the remainder of fiscal 2027, with a more moderate rate of price increases."
— Mark Murphy, Chief Financial Officer

The mechanism is the incentive-compensation decision taken in FQ4. Manufacturing-related incentive compensation was absorbed into inventory in FQ4, so most of its margin effect lands when that inventory sells, in FQ1. Days of inventory rose nine to 129 for the same reason, alongside node end-of-life build-ahead.

"The increase in DIO includes the effect of node end-of-life-related build-ahead and manufacturing-related incentive compensation in fiscal Q4 that was absorbed into inventories. Our inventory levels and supply remain extremely tight, and we expect DIO to decline in the coming quarters."
— Mark Murphy, Chief Financial Officer

Assessment: The distinction between a cost-driven and a price-driven margin decline is the whole ballgame here, and on the evidence this is cost. Prices are still rising in FQ1, HBM resets higher on January 1, and management is willing to call FQ1 the floor and guide margin up for the following three quarters. A price-driven decline would not come with a floor commitment attached. The legitimate concern is a different one: the company has voluntarily and permanently raised its cost base at the cycle peak, and roughly $1B of that is described as ongoing. If pricing does eventually roll, it rolls against a higher cost structure than the one that produced these margins.

4. Pricing: the slope moderated on schedule, and calendar-2027 HBM resets higher

DRAM prices rose in the high-teens percent sequentially and NAND roughly 30%, against the low-60s and mid-80s a quarter ago. Management had foreshadowed exactly this in June, and reiterated the framing.

"… in '27, we had indicated for the balance of the year that we would see margin expansion relative to Q1 a function of continued price increases, albeit at a more moderate pace. And we've said that for some time that eventually price increases would moderate."
— Mark Murphy, Chief Financial Officer

The more consequential pricing event is prospective. HBM contract pricing for calendar 2027 has been negotiated substantially higher and resets at the start of the calendar year.

"As you know, for 2026, our prices for HBM were negotiated with our customers last year. We mentioned that now, for 2027, a large part of the volume is already sold out for 2027 for HBM, and the prices are much higher than 2026 prices. Of course, that is helping us narrow the margin gap with the non-HBM memory there."
— Sanjay Mehrotra, Chairman and CEO

Assessment: Pricing power is confirmed again, but its character has changed in a way that should be read as an improvement rather than a deterioration. Spot escalation is slowing; contracted escalation is accelerating. A high-teens percent spot increase plus a large, already-signed HBM reset in January is a more durable combination than a 60% spot increase with nothing locked, because it survives the first soft week in the channel. The second derivative turning negative is the oldest sell signal in memory, and the reason we are not treating it as one here is that the book behind it is now 26 agreements deep.

5. Supply: no line of sight to balance, and bit growth decelerates anyway

Management extended its tightness call by a full year and, for the first time, said flatly that it cannot see the other side.

"The structural gap between DRAM supply and demand growth rates is resulting in ongoing supply tightness, and cleanroom additions are required to augment node-transition supply growth and help narrow the gap. Even with additional industry DRAM cleanroom space plans, with robust demand trends including new upside requests from customers, we do not have line of sight to when supply and demand will return to balance."
— Sanjay Mehrotra, Chairman and CEO

The supporting industry forecast is the part worth modeling. In calendar 2026 industry DRAM bit shipments grow in the mid-20s percent. In calendar 2027 and 2028 they grow in the low-20s, with HBM bits growing faster than conventional DRAM throughout. NAND industry bits grow in the low-20s percent in calendar 2026 and roughly the mid-20s in 2027 and 2028, supply constrained in both years.

Assessment: Industry DRAM bit supply growth decelerating in the teeth of the largest capex expansion in memory history is the single most thesis-relevant disclosure of the quarter, and it is not intuitive. The explanation is mechanical: HBM consumes more wafer per bit than conventional DRAM, the HBM mix is rising, the trade ratio worsens with each generation, node transitions yield less incremental bit per wafer than they used to, and new cleanrooms take years. Put together, the industry can spend enormous sums and still deliver fewer incremental bits in 2027 than in 2026. That is the structural argument for why this cycle breaks the historical pattern, and it is the most defensible version of the bull case we have heard management make.

6. Capex: half a year of fiscal 2027 nearly equals all of fiscal 2026

The capital bill stepped up again, and the increase is deliberately weighted away from anything that produces bits soon.

"In fiscal Q1, we project capex of around $11.5 billion and anticipate first-half fiscal 2027 capex to be approximately $25 billion. We project capex to be higher in the second half of fiscal 2027. We expect a meaningfully higher growth rate in construction capex as compared to equipment capex in fiscal 2027."
— Mark Murphy, Chief Financial Officer

Approximately $25B in the first half compares with $27.37B for all of fiscal 2026, with the second half guided higher still. The composition question was pressed directly in the analyst session, where a line of questioning put it to management that construction spending produces no bit output until calendar 2029 at the earliest. Management agreed, and treated that as the point rather than the objection.

"… the majority of the construction CapEx increase for fiscal '27 versus our prior plans is for clean rooms that will come online in late '28 and beyond, which shows both how long it takes to build these clean rooms and why we need to get started with the construction investments now. … we will equip the clean rooms and build capacity to the demand trends that we see."
— Manish Bhatia, President and Chief Operating Officer

The fab map behind the spend: Idaho ID1 first wafer output mid-calendar 2027 and ID2 late calendar 2028; Taiwan's Tongluo facility reaching meaningful shipments mid-calendar 2027; a Singapore HBM advanced-packaging facility ahead of plan with initial output in early calendar 2027; a new Singapore NAND facility beginning output in the second half of calendar 2028; a Japan DRAM expansion that broke ground this quarter with initial output in late calendar 2028; and the first New York fab, concrete poured this quarter, with initial wafer output expected in calendar 2030.

Assessment: This remains the dominant residual risk and it got larger, but its shape changed in a way that genuinely matters. A construction-led build is an option on future capacity, not a commitment to future bits: shells can sit unequipped, and management said explicitly it will equip to demand. That decoupling is why industry supply growth can decelerate in 2027 while capex doubles. The risk has not been removed, it has been pushed out and made reversible, and the window where the aggregate industry response lands as actual supply now looks like calendar 2029 rather than 2028. We will track the equipment-versus-construction split every quarter from here, because the quarter construction converts to tools is the quarter the clock starts.

7. The balance sheet, and a December 9 capital-return switch with a $2.2B authorization behind it

Micron ended fiscal 2026 with $73.48B of cash, investments and restricted cash against $5.179B of total debt, and zero interest expense in the quarter.

"We closed the quarter with $5.2 billion of debt and a net cash balance of $68.3 billion. … As noted previously, we intend to increase our capital return from Dec. 9, 2026, the second anniversary of the signature of our definitive CHIPS agreements. Over time, we expect to return 100% of our excess cash to shareholders."
— Mark Murphy, Chief Financial Officer

The CFO added two concrete markers in the Q&A that were not in the prepared remarks: the company expects to reach its target cash level by the end of FQ1, and free cash flow in FQ1 will be "significantly higher than the $33 billion we reported in fiscal Q4." The standing repurchase authorization is $2.2B, with an intent to seek more in the near term. The quarterly dividend stays at $0.15, payable October 29.

Assessment: A $2.2B authorization against a business about to generate more than $33B of free cash flow in a single quarter is the gap that defines the next two months. The December 9 date is now roughly ten weeks away and the target cash level is reached at the end of FQ1, which means the mechanical constraints clear at almost exactly the same moment. What is missing is a number. Management would not size the program, would not name a target cash level, and would not commit to a fiscal-2027 return figure, which is why this was the only place on the call where analysts pushed twice. We treat an authorization increase as the single most likely near-term catalyst in the name.

8. NAND stops being the drag and becomes a $14B business

NAND revenue of $14.1B rose 526% year-on-year, with data center SSDs alone at nearly $10B and over two-thirds of total NAND revenue. Micron is on track for a fifth consecutive year of record data center SSD market share in calendar 2026, with the demand drivers now AI-specific: KV-cache offload for inference and hard-disk displacement.

"Certainly as you noted, the memory hierarchy from HBM to DRAM to SSD, it continues to be leveraged as context windows grow and as AI advances. … You are seeing the results of that, five consecutive years now of shared growth in the data center market."
— Sanjay Mehrotra, Chairman and CEO

The supply side is deliberately restrained. Micron's NAND bit supply grew less than the industry in calendar 2026, partly because Singapore cleanroom space was redirected to an advanced NAND R&D line and to pilot HBM packaging operations. The new Singapore NAND facility comes online in the second half of calendar 2028.

Assessment: Two years ago NAND was the reason to avoid this stock, dragging blended margin down with under-absorption costs. It is now a $14B quarterly business at prices up 30% sequentially, and two-thirds of it sells into the data center on performance characteristics rather than cost per bit. The restraint on bit supply is the right call and is itself a signal: a company that genuinely expected pricing to roll would be maximizing shipments now, not reallocating cleanroom space to R&D.

9. Demand destruction, escalated rather than contained

Two disclosures moved this risk in the wrong direction. Consumer unit declines are now described as potentially double-digit, and server memory content growth was trimmed.

"PC and mobile industry revenue remain on track to grow this calendar year, driven by strength at the premium end of the market, despite potential double-digit overall unit declines in both markets."
— Sanjay Mehrotra, Chairman and CEO

"We expect server unit growth in the high-teens percentage range in both calendar 2026 and 2027. This strong server unit growth is supported by a modestly lower rate of content growth than prior expectations, amid tight memory supply."
— Sanjay Mehrotra, Chairman and CEO

Management's counter is that the trimming is a customer choice to maximize unit shipments against scarce memory, and that it leaves demand latent rather than destroyed.

"… that latent demand, I think, is aligned with your concept that were more memory to become available, it would have — it would easily get put into use in higher content growth in AI workloads, whether attached to accelerators or CPUs."
— Manish Bhatia, President and Chief Operating Officer

Assessment: This is the one pillar that deteriorated this quarter, and we are marking it accordingly. Content growth being revised down because memory is too scarce or too expensive is demand destruction by its proper name, even when the units still ship. The latent-demand argument is logically sound and empirically untestable until supply loosens, which is precisely when it stops mattering. What keeps this from being a rating issue is that server units were held at high-teens growth for both calendar 2026 and 2027, and that a new demand vector appeared this quarter which no one was modelling.

10. The CPU surprise: agentic workloads became a memory demand vector

A line of questioning on the call put it to management that CPU strength had been the biggest incremental surprise since the start of the year. Management agreed, and named the mechanism.

"I think definitely, the realization that agentic workloads are executed across CPU has been a big driver. There is a large attach rate of both LP as well as DDR memory and SSD to enable those agentic workflows. … And that's been one of the reasons that server units are growing so strongly, as we noted, in the high teens."
— Manish Bhatia, President and Chief Operating Officer

The structural point management drew from it is that this is the vector without an HBM trade-ratio penalty, and that it reframes which component is actually scarce: "creating more of a constraint on DRAM and making clear that DRAM is the principal constraint versus logic or power to the data center."

Assessment: This is the most genuinely new piece of information on the call and it is being underweighted. Agentic inference running on general-purpose CPUs attaches conventional DDR and LPDRAM and SSDs, not HBM, which means incremental demand lands on the product lines with the best trade ratio and the highest current margins. It also explains the otherwise odd combination of high-teens server unit growth with trimmed per-server content: more servers, each slightly less loaded, is exactly what you get when memory is the binding constraint and customers optimize around it. If this vector is durable, it widens the demand base beyond the accelerator cycle, which is the single best answer to the peak-earnings bear case.

11. Cost inflation, self-inflicted, becomes a model line item

Operating expense jumped 69.2% sequentially to $2.568B, $918M above the guide.

"Operating expenses in fiscal Q4 were $2.6 billion, up $1.1 billion quarter over quarter. The sequential change was primarily due to the increase in incentive compensation for every global team member, along with our decisions to contribute $300 million to community investments."
— Mark Murphy, Chief Financial Officer

The fiscal-2027 build is larger and permanent in character.

"We project operating expenses to increase by approximately $2.5 billion in fiscal 2027, primarily from higher R&D to support an unprecedented set of opportunities in memory and storage and from higher incentive compensation plans."
— Mark Murphy, Chief Financial Officer

In the analyst session the CFO went further, raising the research-and-development component above what had been signalled a quarter earlier: "excluding the incentive comp effects, the year-over-year R&D increase is going to be more than the $1 billion that we said last quarter, will be over $1 billion in '27 as we have added additional R&D activities."

Assessment: Paying the workforce at the top of a cycle you created is defensible, and $300M of community investment at a 61% free-cash-flow margin is noise. The item to watch is the research-and-development step-up, because it is the one piece of the $2.5B that will not come back out if conditions soften, and management raised it twice in two quarters. At 4.7% of revenue, operating expense is not the risk today. It becomes a risk in the scenario where revenue halves and this cost base does not, which is the scenario the SCA floor is supposed to cover and which nobody has yet put a number on.

12. A new management structure, and a second call run by the new presidents

Micron announced leadership appointments last month that took effect this quarter. Manish Bhatia became president and chief operating officer with accountability for the operating profit and loss; Dr. Scott DeBoer became president and chief technology and products officer.

"Manish Bhatia has been appointed to president and chief operating officer, and Dr. Scott DeBoer has been appointed to president and chief technology and products officer."
— Sanjay Mehrotra, Chairman and CEO

The practical consequence was visible in the mechanics of the day. Mehrotra and the CFO ran the main call; the separate post-earnings analyst session was handled by Bhatia, DeBoer and the CFO, with Mehrotra absent. On technology, DeBoer used that session to set the competitive frame directly: at least two nodes of leadership over Chinese competitors, 1-gamma already the majority of Micron's bits and on track to be the largest node in company history, 1-delta ramping in the second half of calendar 2027, and EUV expertise as the differentiator. China exposure is expected in the single-digit percent range in fiscal 2027.

Assessment: A chief operating officer who owns the operating profit and loss, and a chief technology officer who owns the roadmap, is the right structure for a company whose next three years are an execution problem rather than a strategy problem. The succession read is obvious and we would not over-interpret it: giving two internal candidates visible platforms at a peak is good governance, not a signal. The substantive takeaway from the second call is that the technology gap to Chinese DRAM is being held at two nodes while Micron's China revenue exposure falls to single digits, which removes the most-cited structural bear argument from the next two years of the debate.

Guidance & Outlook

The guide is a 13.4% sequential revenue step on top of a 30.8% one, which is a deceleration in rate and an acceleration in absolute dollars: $7.3B of incremental quarterly revenue against $12.8B last quarter. Annualizing the $38.15 EPS guide yields a run rate of $152.60, roughly double the $75.52 that fiscal 2026 just produced in full.

Implied fiscal-2027 shape: management committed to sequential revenue growth in every quarter of fiscal 2027, to FQ1 being the gross-margin floor, to roughly $2.5B of additional operating expense for the year, and to a 15.5% tax rate. Hold the FQ1 guide and apply decelerating sequential growth of 6%, 4% and 3% across the remaining quarters with gross margin drifting to 88% by FQ4, and fiscal 2027 lands near $264B of revenue, an 87% blended gross margin, and roughly $165 of EPS. That is a framework, not a forecast, and its sensitivity is almost entirely to the sequential price path rather than to anything in the cost base.

Street at: consensus carried roughly $56.5B for FQ1, so the guide lands about 8.8% above it. Estimates have chased this company in every one of the past five quarters and will again.

Guidance style: historically conservative, and still conservative in substance despite the headline aggression. Management guided a gross-margin decline it could plausibly have smoothed, named the cause precisely, attached a floor commitment, and gave an unusually detailed cost walk for a company with no obligation to do so. The one piece of housekeeping it volunteered at the end of the analyst session is the modeller's tell: "Our first quarter guidance factors in a single-digit sequential bit growth and double-digit for cost for both DRAM and NAND." Bit growth in single digits with cost in double digits is a company telling you that FQ1 margin rests entirely on price, which is exactly the walk it had just described.

Analyst Q&A Highlights

What counts as "enough cash" before the buyback starts

The opening question of the main call went straight to the gap between a $68.3B net cash balance and a $2.2B repurchase authorization, pressing for any quantitative marker on the target cash level. Management supplied dates and a mechanism while declining to name a number, and disclosed that the current authorization will be expanded.

Q: "Mark, I wanted to ask about capital return. I know you do not want to front-run yourself, but can you give us any mileposts here? What are you thinking of minimum cash balance? You have pretty much double the cash that Apple or NVIDIA have. So, what is enough cash? Do you want to keep $100 billion in cash and you will return everything beyond that?"
— Timothy Arcuri, UBS Securities

A: "On target cash, we expect to be around a target cash level by the end of fiscal Q1. Over time, with the rate and pace determined by various factors, we plan to return excess, primarily done through share repurchase. We intend to increase capital return, as we have talked about before and in the script today, from December 9th. Finally, I will just add that our current authorization, that you can see from our previous filings, for share repurchase stands at $2.2 billion. You can assume that we will seek additional authorization in the near term on more authorization."
— Mark Murphy, Chief Financial Officer

Assessment: The most actionable exchange on the call, and the only one where management added material disclosure under questioning. Three facts came out that were not in the prepared remarks: the target cash level is reached by the end of FQ1, free cash flow in FQ1 will exceed the $33B just delivered, and the existing authorization is inadequate and will be increased. Put together, the constraint is procedural rather than financial, and it clears inside the next quarter. The refusal to name a target cash level is the standard pre-announcement posture and we read nothing into it.

Quantifying the fiscal Q1 gross-margin step-down

The guided sequential gross-margin decline drew the most detailed answer of the call, and the number that emerged is the one that determines whether the step-down is a pricing event or an accounting one.

Q: "Within the guide for November, can you quantify the impact from the higher ASP inventory? Are there other sort of mix shifts that we should be thinking about that are impacting the sequential?"
— C.J. Muse, Cantor Fitzgerald

A: "We also have some startup costs, which I have talked about previously, and some other costs, but incentive comp is the big driver to that gross margin outlook. In total, these factors in the first quarter are roughly $1 billion of higher cost in Q1. So that gives you a sense of the margin impact. … Beyond the normal cost increases from higher volume and depreciation, you have this roughly $1 billion of ongoing costs from incentive comp, higher startup, and some other costs in fiscal 2027."
— Mark Murphy, Chief Financial Officer

Assessment: Roughly $1B of cost on $61.5B of revenue is about 160 basis points of gross margin, against a guided decline of 75. The arithmetic says the cost headwind is larger than the margin decline, which means price and mix are still contributing positively underneath it. That resolves the question the guide raised: this is not the price roll. The sting in the tail is the word "ongoing" applied to the same $1B for the full fiscal year, which converts a one-quarter absorption into a permanent step in the cost base.

Whether calendar 2027 is the peak the market is pricing

A recurring line of questioning put the valuation puzzle to management directly: memory multiples imply the market believes next year is the peak for pricing and earnings, whether from incremental capacity or from customers trimming content. The answer was the most forceful supply-demand statement on the call.

Q: "When we look, Sanjay, at just the valuation of memory stocks, very depressed, which suggests that people feel that next year might be a peak for pricing, might be a peak for earnings in this cycle, either because the industry is bringing on incremental capacity, or there is a narrative that maybe some of your customers, even in the data center, might de-spec products, either because of shortages of memory or just because it's such a bigger part of the material."
— Vivek Arya, Bank of America Securities

A: "As we noted, in calendar 2027 as well as 2028, we see demand exceeding supply. In fact, we see greater tightness in the industry in 2027 and in 2028 versus 2026. Overall, supply-demand environment is only getting tighter. … I'll tell you that 2027, more than 75% of our output is already committed for 2027. A majority of discussions with our customers today are already around 2028."
— Sanjay Mehrotra, Chairman and CEO

Assessment: The question is the right one and the answer is the strongest available, but it is still management's view of its own market. What elevates it above an assertion is the committed-output figure. More than 75% of fiscal-2027 output sold, with allocation conversations already running on 2028, is a verifiable commercial fact rather than a forecast, and it is hard to reconcile with a 2027 peak. The remaining bear path is not that demand disappears; it is that the un-contracted quarter of the book clears at much lower prices once any slack appears.

How high the contracted share of revenue can ultimately go

With the agreement count up from 16 to 26 in a single quarter, a natural question was whether the ultimate coverage target has moved with it. Management held the prior figure, and separated two numbers that are easy to conflate.

Q: "Last earnings with 16 SCAs secured, your view was the potential for 50%+ kind of SCA coverage on forward revenues, given the pipeline of SCA signed and under negotiation. Fast-forward to this quarter, 26 SCAs now secured on the coverage. … So what do you think your forward revenue coverage will be if all of your current SCA negotiations are completed?"
— Harlan Sur, JPMorgan

A: "Regarding your question on where could SCAs ultimately reach to, that basically is still around 50% of our revenue through the 2030 timeframe that we had previously commented on. That comment still remains about the same, that we could reach around 50% of our revenue to be covered by SCA."
— Sanjay Mehrotra, Chairman and CEO

Assessment: The ceiling did not move, and that is deliberate rather than disappointing. Management wants to keep roughly half the book un-contracted to retain mix flexibility across customers and end markets, which is a defensible operating choice and a real limit on how far the de-cyclicalization argument can be pushed. The more useful disclosure in this exchange is the distinction management drew: more than 75% of fiscal-2027 output is committed across both agreement and non-agreement customers, while the long-dated contracted share tops out near half. Near-term visibility is much better than the structural contracted share implies.

Customer de-speccing of HBM, and where the wafers would go

Press reports of at least one large customer trimming HBM content per platform were put to management, together with the second-order question of what happens to supply if high-trade-ratio HBM wafers get reallocated to conventional DRAM.

Q: "… there's been talk of maybe one large customer de-speccing HBM. I'm kind of curious what your view on that is, and also implications given the fact that HBM has higher trade ratio. If those wafers get reallocated to DDR, would that increase DDR supply quite a bit?"
— Krish Sankar, TD Cowen

A: "When customers make optimizations, they make these optimizations mainly to be able to ship more units to drive their own more growth, as well as to address the end market opportunity of growing, scaling up AI. These optimizations, when they occur, they do not take away from the latent nature of a need for more memory in the systems. These optimizations also have diminishing return for any further optimizations."
— Sanjay Mehrotra, Chairman and CEO

Assessment: Management did not deny the de-speccing, which is the first thing to notice. It reframed it as customers maximizing unit shipments against a memory constraint, and added the useful observation that optimization has diminishing returns, meaning the second round of de-speccing saves less than the first. The reallocation question went unanswered, and it is the one that matters arithmetically: moving wafers from HBM to conventional DRAM would release more bits than it removes, which is a supply event management described elsewhere as structurally impossible because HBM keeps outgrowing conventional DRAM. Those two claims sit in tension and the tension was not resolved.

Whether the new agreements still carry price floors and ceilings

With ten agreements signed into a tighter market than the ones before them, the analyst session probed whether the pricing architecture had changed and whether Micron was now less willing to cap its own upside.

Q: "Obviously, they want supply, and they want longer-dated supply. But is there any kind of change in the pricing construct either you or they are asking for? … Are you maybe a little bit less inclined to call for the fixed ceiling and floor pricing if you think you can capture a little bit more upside over the next couple of years?"
— James Schneider, Goldman Sachs

A: "I would say that the framework of the SCAs that we have is similar. But what's different is that the negotiations reflect current market conditions and outlook for market pricing, right? And so the direction of travel has been for higher pricing. … The majority of the pricing frameworks have floor and ceiling bands, but the newer ones are negotiated with an eye towards the current market conditions and the future market tightness that we see."
— Manish Bhatia, President and Chief Operating Officer

Assessment: This is a quietly important answer. The architecture is unchanged, floors and ceilings on the majority, but the bands themselves reset upward with each signing cohort. The June agreements were struck at calendar-Q2 market conditions; the September cohort was struck against a materially higher spot market and an expectation of tightness through 2028. That means the blended floor under the whole book rises every quarter that new agreements are added, which is a mechanism for the floor to catch up toward current margins over time without management ever disclosing its level. It is the best available answer to our standing complaint that the floor is unquantified, and it argues for the floor being a moving target in the right direction.

Why industry supply growth decelerates despite record capital spending

The most analytically loaded exchange of the day questioned the apparent contradiction between the largest capex expansion in memory history and a forecast of slower industry bit growth in calendar 2027.

Q: "In terms of supply growth decelerating next year, I guess that's a little surprising in the context of the CapEx. And I know you talked about some of the dampening effects of HBM, but I don't think that delta should be changing that much. So I just wonder if you could just kind of explain what the puts and takes are that supply would decelerate given the CapEx that you see."
— Joseph Moore, Morgan Stanley

A: "I mean I think that we gave the color that HBM is growing faster than conventional DRAM. And the trade ratio as you move more of the industry is shifting from HBM3 to HBM4 and then towards later in the year in '27 HBM4E and these are increasing trade ratio. So if you just look at both of those two things happening together, that has a dampening effect on the ability for bit growth to be growing."
— Manish Bhatia, President and Chief Operating Officer

Assessment: The answer is mechanically correct and it is the load-bearing beam of the entire bull case. If each HBM generation consumes more wafer per delivered bit, and HBM keeps taking share of total output, then industry bit supply can fall even as wafer capacity and spending rise. That is a structural reason this cycle could behave unlike its predecessors, and it is testable: the forecast is low-20s percent DRAM bit growth in calendar 2027 against mid-20s in 2026. If industry bit growth comes in above that, the trade-ratio argument is weaker than claimed and the 2028 supply risk arrives early.

What They're NOT Saying

1. The floor margin, for a second consecutive quarter: management reaffirmed that floor-price margins are "meaningfully above any prior cycle peak margins" and again declined to quantify them. This is now the longest-standing unanswered question in the name and it is the one that sets through-cycle earnings power. A floor that merely beats any prior-cycle peak still sits a long way below an 87% print, and management has never said how far.
2. Any fiscal-2027 revenue or full-year capex figure: the company guided one quarter of revenue, a half-year of capex and a direction for the second half. An analyst floated fiscal-2027 capex above $50B on the call and management neither confirmed nor corrected it. For a company committing to construction that will not produce bits until calendar 2029, the absence of a full-year capital plan is a conspicuous gap.
3. The target cash level and the size of the buyback: management said it reaches a target cash level by the end of FQ1 and will seek a larger authorization, without naming either number, ten weeks before the December 9 start date.
4. How much of the FQ1 guide is contracted versus spot: we know more than 75% of fiscal-2027 output is committed and that agreements cover more than 35% of revenue through 2030, but not what share of the $61.5B guide is priced under an agreement. That is the number that would tell you how protected the next print actually is.
5. An updated HBM total addressable market, for the second consecutive quarter: asked directly to refresh the prior ~$100B-by-2028 figure in light of a transformed pricing environment, management declined. In a quarter where almost every other forward number was revised higher, declining to raise the HBM market estimate is the one place the company chose not to press its advantage.
6. The DRAM-versus-NAND split of the agreement book: management confirmed the agreements cover both and offered only that DRAM bit volume is a little under the roughly 35% headline and NAND a little over. The revenue split, which is what matters given the margin differential, was withheld.
7. What happens to conventional DRAM supply if HBM is de-specced: the reallocation question was put directly and went unaddressed, leaving the asymmetry between the de-speccing answer and the trade-ratio argument unresolved.

Market Reaction

• Pre-print setup: MU closed at $1,065.11 on September 30, up 273.2% year-to-date and 536.6% over the trailing twelve months, having added 11.1% over the prior 30 days. The stock entered the print 12.2% below its 52-week closing high of $1,213.56, recovering toward that high rather than sitting on it. The S&P 500 was up 11.8% year-to-date.
• After-hours move: shares rose roughly 1.7% immediately after the release, gave the gain back during the call, and settled extended trading up about 0.4% near $1,069.
• Reaction session: the regular session opened at $1,054.08, 1.0% below the pre-print close, traded in a range of $1,022.90 to $1,098.90 (down 4.0% to up 3.2%), and closed at $1,097.39, up 3.0% (+$32.28), near the session high. Volume of 44.6M shares was 1.8 times the 30-day average of 25.0M. The S&P 500 rose 0.2% on the day.

Zoom out one quarter and the picture sharpens. MU closed at $1,048.51 the day it reported fiscal Q3 on June 24 and at $1,065.11 the day it reported fiscal Q4, a gain of 1.6% across a quarter in which revenue grew 30.8%, the contract book went from 16 agreements to 26, and the next fiscal year's guide arrived 9% above consensus. The stock reached $1,213.56 on a closing basis in between and gave it back.

A 3.0% gain on a beat of this size, with a guide 9% above consensus, is a muted reaction and the day's most important non-financial signal, and it is a sharp contrast with last quarter. In June the stock jumped 13% to 14% after hours on a comparable beat, and we wrote that the market had begun paying up for the structural story rather than fading the news. One quarter later it is paying up far less, without the fundamentals giving it a reason.

The attributions circulating are three, and they are of unequal quality. The first is positioning: a stock up 536% in twelve months discounts a great deal of good news, and 11.1% of that came in the 30 days immediately before the print, so a strong quarter was already in the price. That is legitimate. The second is the gross-margin guide: 86.25% against 87.0% printed, with the word "floor" attached. That one does not survive contact with the cost walk, which showed roughly $1B of incentive-compensation and start-up cost driving a 75 basis-point decline, implying price and mix were still positive underneath. The third is capex, and it is the serious objection: approximately $25B in the first half of fiscal 2027 against $27.37B for all of fiscal 2026, with the second half higher, is the kind of number that makes investors reach for the memory-cycle playbook regardless of what the money is buying.

Our read is that the market is marking the capex and the content-growth trim, and discounting the contract book. That is a defensible trade on a twelve-month view and a poor one on a three-year view, because the thing that changed this quarter was not the margin trajectory but the degree to which it is contracted. The stock sitting nearly 10% below its high while remaining performance obligations rose 50% and more than three-quarters of fiscal-2027 output is committed is a better entry than June's, not a worse one.

Street Perspective

Debate: Is calendar 2027 the earnings peak the multiple implies?

Bull view: at roughly 7 times the annualized fiscal Q1 guide, the market is explicitly pricing mean reversion that the order book contradicts. More than 75% of fiscal-2027 output is committed, agreements run to 2031, industry bit growth decelerates in 2027 and 2028, and management says it cannot see when supply and demand balance. A peak you can see twelve months out does not usually have customers already negotiating 2028 supply.

Bear view: every memory cycle has had a quarter where management explained why this one was different, and the tells are all present: the first guided margin decline, capex nearly doubling, content growth trimmed, and consumer units potentially falling double digits. Three-quarters of the long-dated agreement book has a defined pricing framework, mostly floor-and-ceiling bands nobody has quantified, and the other quarter reprices periodically at market. A single-digit multiple on a guided peak is the market's forecast, not its mistake.

Our take: the bear is describing the correct checklist and the bull has the better specific evidence this quarter. What breaks the tie is the trade-ratio argument, because it supplies a mechanical reason for industry supply to decelerate while capex accelerates, which no prior memory cycle had. The bear case now requires either that argument to be wrong or demand to crack, and the de-speccing evidence is the thing to watch on the demand side. We would own the stock and size for the possibility that we are a year early in the other direction.

Debate: Does the construction-weighted capex make the 2028 glut better or worse?

Bull view: a construction-led build is an option, not an obligation. Shells can sit unequipped, management said explicitly it will equip to demand, and the cleanrooms in question come online in late calendar 2028 and beyond, with meaningful output a few quarters later still. Funding that from a $68.3B net cash position while generating more than $33B of quarterly free cash flow costs shareholders almost nothing.

Bear view: capital intensity is rising at the top of a cycle, which is the single most reliable leading indicator in the industry's history, and the discipline argument assumes a restraint no memory producer has ever demonstrated when a competitor is building. Shells get equipped. The only question is whether 2029 or 2030 is the year the bill arrives.

Our take: the composition genuinely changes the timing and not the existence of the risk, which is a smaller concession than the bulls want and a larger one than the bears allow. We moved our expected window for the aggregate supply response from calendar 2028 to calendar 2029 on this call, and the specific metric we will track is the equipment-versus-construction split. The quarter construction converts to tool orders across the industry is the quarter to reassess, and it will be visible in semicap bookings before it is visible here.

Debate: Is a $68.3B net cash pile with a $2.2B authorization a catalyst or a credibility test?

Bull view: the constraints clear simultaneously inside the next ten weeks. Target cash is reached at the end of fiscal Q1, the contractual restriction lifts on December 9, free cash flow in fiscal Q1 alone exceeds $33B, and management has pre-announced that it will seek a larger authorization. A buyback at 7 times forward earnings is the highest-return use of capital available to this company.

Bear view: a company that returns 100% of excess cash does not simultaneously guide capex to approximately $25B in a half year and raise operating expense by $2.5B. "Excess" is doing a great deal of work in that sentence, and on a capital plan this large the residual could be far smaller than the cash balance suggests.

Our take: the bear has identified the right ambiguity and the arithmetic still favors the bull. Annualize the fiscal Q1 free-cash-flow guidance and the full-year figure comfortably exceeds any plausible reading of the capital plan, even at $50B-plus of capex, which leaves a very large residual. The risk is not that the cash is not there; it is that the pace disappoints relative to a balance sheet this liquid. We regard an authorization increase as the most likely near-term catalyst and would treat a small one as a genuine negative surprise.

Model Update & Valuation Framework

Valuation framework. At the $1,097.39 close on October 1 and roughly 1.15 billion diluted shares, Micron carries a market capitalization near $1.26 trillion and, net of $68.3B of net cash, an enterprise value near $1.19 trillion. Against fiscal 2026's realized $75.52 of EPS that is 14.5 times trailing earnings. Against the annualized fiscal Q1 guide of $152.60 it is 7.2 times, and against the implied annualized fiscal Q1 operating income of roughly $204B the enterprise value is 5.9 times. On the fiscal-2027 framework above it is about 6.7 times.

Where we land. The valuation question has not changed in substance since June: a single-digit multiple on guided peak earnings is the market pricing mean reversion, and the argument turns on what normalized earnings actually are. What changed this quarter is the weight of evidence on that question. The contracted book grew 50% at minimum pricing, fiscal-2027 output is more than three-quarters committed, HBM pricing for 2027 is already signed substantially higher, and the industry's own bit supply decelerates even as it spends record sums. Each of those raises the floor under normalized earnings without telling us where it sits. The multiple has not re-rated at all, which means the market has given back the entire structural re-rating it granted in June while the structure itself got stronger. That is the asymmetry we are paid to take. We would revisit if spot pricing turns negative before the contracted share of revenue reaches half, or if the equipment share of capex starts rising faster than construction.

Thesis Scorecard Post-Earnings

Overall: the thesis strengthened. Of the eight commitments we set in June, five were met or resolved favorably (the fiscal Q4 guide delivered and beaten, no pricing rollover, no cancellations, the HBM4 ramp on track, the capital-return pivot confirmed with dates), one made material progress without arriving (the contracted share of revenue at more than 35% against a 50% target), one was only partly disclosed (fiscal-2027 capex, given as a half year plus a direction), and one was declined for a second consecutive quarter (the floor-margin number). Neither downgrade trigger we named fired, the Hold trigger in fact moved further away, and the structural pillar we upgraded on in June grew 50% in a single quarter. Two things got worse: demand destruction escalated from contained to emerging on trimmed content growth and a second quarter of falling consumer bits, and the cost base stepped up permanently. Neither is a rating issue at 7 times forward earnings with more than three-quarters of fiscal-2027 output committed. The decisive open question is unchanged from June and is now a quarter older: where the contractual floor actually sits.

Action: Maintain Outperform; raise conviction from 7 to 8. Add on weakness. Move to higher conviction still if management quantifies the floor margin, if the contracted share of revenue reaches half, or if the December authorization is large relative to the cash balance. Move back to Hold if spot DRAM or NAND pricing turns outright negative before the contracted book can cushion it, or if industry bit growth comes in above the low-20s percent in calendar 2027, which would falsify the trade-ratio argument the bull case now rests on. Move to Underperform only on order cancellations or a breach of the take-or-pay commitments.

Bottom Line

Micron closed fiscal 2026 with $54.2B of revenue at an 87.0% gross margin and $33.42 of EPS, beating a guide it had already set at a record while absorbing nearly a billion dollars of unbudgeted cost. The full year came in at $133.2B of revenue and $75.52 of EPS against $37.4B and $8.29 a year earlier, and it then guided the next quarter to $61.5B and $38.15, roughly 9% above where the Street sat. The stock traded down as much as 4.0% the next day before closing up a muted 3.0%.

We are maintaining Outperform and raising conviction, and the reason is the contract book rather than the quarter. Three months ago we upgraded on a single disclosure: that take-or-pay agreements floored gross margins above any prior-cycle peak. This quarter that book went from 16 agreements to 26, from roughly $100B of remaining performance obligations to approximately $150B measured at minimum pricing, from $22B of customer commitments to $32B, and from a 2030 horizon to 2031. More than three-quarters of fiscal-2027 output is already committed, and a majority of the conversations management is now having concern 2028. That is not the shape of a business twelve months from a peak.

We will be plain about what got worse. Content growth per server was trimmed because memory is scarce, consumer units could now fall by double digits, mobile bits declined for a second consecutive quarter, and the company declined for a second consecutive quarter to say where its margin floor sits. Capex for the first half of fiscal 2027 alone nearly matches all of fiscal 2026, and operating expense steps up $2.5B permanently. Any of those is a legitimate reason to own less of this stock than the headline numbers would suggest.

What tips the balance is a single mechanical argument that no previous memory cycle could make. Industry DRAM bit supply growth decelerates in calendar 2027 and 2028, to the low-20s percent from the mid-20s, despite the largest capital expansion in the industry's history, because HBM consumes more wafer per bit with every generation and keeps taking share of total output. If that is right, the record capex does not produce the record glut, and management's statement that it has no line of sight to supply and demand balancing is a description rather than a boast. The capex that worries the market is construction, buying cleanrooms that come online in late calendar 2028 at the earliest, and the company has said it will equip them to demand. The supply reckoning has moved out, not away.

At the $1,097.39 close the stock trades at 7.2 times the annualized fiscal Q1 guide, 9.6% below its 52-week closing high, with $68.3B of net cash, zero interest expense, a capital-return switch that flips on December 9, and a free-cash-flow quarter ahead that management says will exceed the $33.2B just delivered. The market gave back most of the structural re-rating it granted in June while the structure itself got stronger. We think that is the trade, we think the floor is the number that will eventually settle the argument, and we would rather be early again on the right side of a contracted book than late to a cycle call. Outperform, conviction 8.