The Peak Moved Out. The Bill Moved Up.
A record quarter and a guide above the Street; capital spending above $50 billion; prices rising a third as fast. Five of eight frozen questions cleared, one missed, two not given.
Download PDFDownload modelThe original valuation →
How to read it. Each bar is a fiscal quarter’s revenue divided by its number of weeks, in billions of dollars per week; the figure above each bar is that rate and the percentage is its change on the quarter before. The black bar is the quarter just reported, which had fourteen weeks; the dashed bar is the company’s guide for the next quarter, at its midpoint. Reported revenue and the week count sit under each bar. Dividing by weeks removes the extra week from the comparison.
What the print changed
Our valuation of Micron, Are We Past the Peak?, asked its question of margins, of the stock, and of the cycle. The print answered the first. Of margins: no. Gross margin was 87.0% against an 86% guide, and the company guided 86.25% for the quarter now under way while calling it "the floor for gross margins in fiscal 2027." It also answered a question the title did not ask. Of the rate of price increase: yes, and sharply. DRAM prices rose in the high-teens percent in the quarter, against the low-60s three months earlier; NAND rose about 30%, against the mid-80s. Prices are still rising. They are rising a third as fast; for DRAM, about the pace of nine months ago. Of the stock's own high and of the cycle, the print could not say, and did not.
The quarter beat its guide on revenue, margin and earnings. Revenue was $54.2 billion, 6% above the top of the range, up 31% on the prior quarter. Earnings per share were $33.42 against a $31.59 consensus. Free cash flow was $33.2 billion. Operating expenses ran the other way, $2.6 billion against about $1.65 billion guided, on a company-wide incentive payout and a $300 million community contribution. Fiscal 2026 closed at $133.2 billion of revenue, 3.6 times the year before.
The quarter had fourteen weeks, and calendar was 30% of the headline growth. Strip the extra week and the quarter was worth $50.4 billion, against the $46.4 billion our scorecard required, and revenue per week rose 21%. The guide for the next quarter, a thirteen-week one, is $61.5 billion: revenue per week up another 22%. The weekly rate has not slowed. Its composition has: DRAM bits grew mid-single digits in a quarter one-thirteenth longer, which is two to three percent fewer bits per week, so price did all of the work and then some. Supply is the constraint, by the company's account: conditions will be "much tighter" in 2027 and 2028 than in 2026, it has no "line of sight" to balance, and "more than 75% of our output is already committed for 2027."
How to read it. Each point is the change in Micron’s average selling price from the quarter before, as the company states it in its prepared remarks: black for DRAM, gold for NAND, the five fiscal quarters to the one just reported. The company gives ranges in words (“high-teens,” “low-60s”); the label shows the range those words denote and the point sits at its middle. Every point is above zero, so prices rose in every quarter shown. The line’s height is the speed of the rise, not the level of prices: the last point is lower than the two before it because the increase slowed; for DRAM, to about the pace of the second point.
The contracts are the largest change in the file. In June Micron had sixteen strategic customer agreements carrying about $100 billion of minimum revenue. It now has twenty-six, with remaining performance obligations (RPO, the contracted revenue at minimum prices) of about $150 billion, which is about $35 billion a year if spread over the terms to 2030, a quarter of fiscal 2026's revenue. Customer financial commitments rose to $32 billion from $22 billion, and $12.3 billion of deposits arrived in the quarter. Three-quarters of the contracted revenue has a defined pricing framework, most of it a band with a floor and a ceiling; the rest is repriced periodically at market, and new agreements are struck at higher prices. No floor price was disclosed, and the stated ambition is unchanged at about half of revenue.
One of the eight questions we froze resolved against the company, and it is the one about supply. Micron did not guide fiscal 2027 capital spending as a number. It guided about $11.5 billion for the first quarter and about $25 billion for the first half, and said the second half would be higher. That puts the year above $50 billion, and at the second quarter's implied $13.5 billion rate, at $52 billion or more; an analyst's "higher than $50 billion" went uncorrected on the call. Our bar was $48 billion; fiscal 2026 was $27.4 billion. Most of the increase is construction, "to help accelerate clean room space availability in late calendar 2028 and beyond."
Micron’s fiscal 2027 capital spending, by the company’s own halves: about $25 billion in the first half and more in the second. No full-year figure was given. Fiscal 2026 was $27.4 billion and the year before $13.8 billion. The largest in the company’s history, committed at an 87% gross margin — the skeptic’s strongest number, and the one question on our scorecard the print failed.
Two questions got no answer. HBM, the stacked memory bonded to AI accelerators, was again described in adjectives: revenue "grew faster than total company revenue," with most of calendar 2027's supply sold at prices "much higher than 2026 prices." Asked on the call for a percentage, the chief executive gave none. And server demand was given as units, up high-teens percent in 2026 and 2027, without the CPU split our last question asked for.
The case against reading this as a higher value (steelman — stated in its own terms, before we test it)
A skeptic of the memory cycle reads this release as the pre-note's sequence arriving on schedule. In 2018 and 2022 the stock peaked first, the record margin came next, and the company's language turned last. The high close is now fourteen weeks old. The record margin has been reported and a slightly lower one guided, and the release carries every ingredient of the turn except the word. The rate of price increase fell by two-thirds in a quarter. Customers are trimming: server growth is "supported by a modestly lower rate of content growth than prior expectations, amid tight memory supply," and PC and phone units face "potential double-digit" declines. Inventory rose nine days to 129, which the company attributes to a build-ahead and expects to reverse. Costs are rising, with operating expenses up $2.5 billion in fiscal 2027 and about $1 billion of new cost in the first quarter alone. And the supply response has begun: capital spending that doubled in a year and will nearly double again, the largest in Micron's history, committed at its highest margin, with new cleanroom space arriving from mid-2027 and in size from late 2028. A questioner on the call put the market's view to the chief executive directly: memory valuations are "very depressed, which suggests that people feel that next year might be a peak for pricing."
On this view a higher peak is only a higher peak. A $61.5 billion quarter is cash in fiscal 2027; it says nothing about what a DRAM bit sells for in 2030, and the one number that would, the contract floor, is still withheld. The skeptic can quote our own pre-note here: we refused to fade growth from fiscal 2026 because a fade from a peak bakes the peak into the terminal value. Holding growth rates on a higher peak does the same thing.
The test. We can enter that case in the model exactly. Which of its claims the rerun tests: the capital spending, directly; the rising costs, directly; the higher peak being temporary, as a run in which every dollar of the guided price step is given back from fiscal 2028. Which it cannot: whether the 2018 and 2022 sequence is repeating, whether customers' trimming becomes a fall in demand, and what the floors are. Both runs are in the next section, side by side.
What Micron is worth after this
How to read it. Each row is one run of the same cash-flow model, in dollars per share on a common scale. The three ticks are the downside, base and upside cases; the dot is their weighted value at 25/50/25, labelled above it. The top row is the valuation as published before the print. The middle row applies the release with growth rates after fiscal 2027 left as published, so a higher guided year lifts every later year. The bottom row applies the same release but returns revenue from fiscal 2028 onward to the published dollar path, so the higher guided prices count as one year’s cash and nothing more. The shaded band is the range the central value was allowed to move on a single print; the dashed line is the closing price on October 1, the session after the report. The gap between the lower two dots is the value of one assumption.
Our published fair value was $1,114: downside $554, base $1,013, upside $1,875, weighted 25/50/25. We re-ran that model on the release. Some of what changed is the company's numbers: the fiscal 2026 base, net cash excluding customer deposits of $55.6 billion (we had estimated $46.0 billion), and the fiscal 2027 guidance for operating expenses and tax. The base-case margin is the guide less four points of in-year fade, 82.25%, the rule we froze, though the company now guides the opposite of a fade.
The rest is our construction, and the frozen rules do not dictate it. The revenue rule we froze was the exit run-rate times 1.18, with the first-half price set from the guide. Read alone, its first clause gives a base year of $238 billion, less than the guided quarter held flat; one reading of the two clauses together gives $279 billion. We start each case from the guided quarter instead: the base takes the $61.5 billion midpoint and grows on bits alone with prices flat, $262 billion, from $219 billion; the downside takes the low end and no growth, $240 billion; the upside the high end and modest further price, $280 billion, with first-year margins of 78% and 87% beside the base. Capital spending is our estimate, $55 billion in fiscal 2027 ($57 billion in the upside case) and $8 billion more than published in fiscal 2028. The cost of equity, the growth rates after fiscal 2027, the recovered margins and the weights are unchanged.
The result: downside $679, base $1,203, upside $2,131, a weighted fair value of $1,304, from $1,114. Those two readings of the frozen rule give $1,243 and $1,345; the band we froze before the print was $959–1,332, and the higher reading would breach it. At the extremes, every operating claim granted for a decade is worth $2,983, from $2,639, and the downside with the contract floors failing is worth $613, from about $500.
Nearly all of that move is one assumption, and it is the skeptic's target. Rerun with the new balance sheet, costs and capital spending but fiscal 2027 revenue left where we published it, the value is $1,085. The higher starting year adds $219 a share: $27 is fiscal 2027's own extra cash, and $192 exists only because the model's down-leg is a percentage of the peak, so a higher peak carries a higher recovery. Give the guided price step back from fiscal 2028, every later year on the published dollar path, and the value is $1,112: unchanged. At a 14% cost of equity, the skeptic's rate, the rate-preserving run is worth $1,028.
The evidence between the two runs is the contract book. Minimum contracted revenue rose 50% in a quarter, at floors the company says exceed any prior peak margin, with new agreements priced higher. That is a rising floor under the recovery, which is what the rate-preserving run assumes, and it is unverifiable for the same reason as before: the floors are not disclosed. We hold the weights at 25/50/25. The upside conditions we published are not met: coverage is "over 35%" of revenue against the half we required, with no floor disclosed. The downside condition nearest to firing was capital spending above $50 billion "without a matching revenue path." We read matching as capital intensity: about $55 billion against a $246 billion annual rate is 22% of revenue, against the 21% fiscal 2026 spent. The skeptic's objection stands: the revenue is this year's and the cleanrooms are 2028's. So the condition is neither met nor retired, and the re-armed bar is $60 billion, a quarter of the guided run-rate, without a higher revenue outlook.
The stock closed at $1,065.11 into the print and at $1,097.39 the session after, up 3.0%. At that price it is 16% below the weighted value and 9% below the base case. Reverse-engineered on the rate-preserving path, $1,097.39 requires a recovered mid-cycle gross margin of 53.3% (46% to 61% across a two-point range of the cost of equity), against 63.5% before the print and a prior-cycle peak of 61%; or $91 billion a year of free cash flow forever, against about $123 billion at the quarter's rate on a thirteen-week basis. On the given-back path the price is within 2% of value.
The scorecard, frozen five days before the print — 5 cleared, 1 missed, 2 not given
Eight questions were frozen before the print, each with a rule for what silence would mean. All eight are scored here, the miss first. Three misses of our own belong beside the company's: the base case's whole first year, $219 billion, was below the single quarter the company then guided, annualized, and that annualized $246 billion is above the $205–235 billion range our frozen addendum gave for base-case fiscal 2027 revenue; our estimate of the quarter's free cash flow was $11 billion too low; and the setup we flagged, a guide that cleared our bar and missed the Street, did not occur.
missednot givencleared
5Is fiscal 2027 capital spending inside the modeled range, $48bn or less?MISSED
No. No full-year figure was given. But $25bn in the first half with a second half the company calls "higher" is above $50bn, the downside tell we froze. Read literally, our silence rule (no fiscal 2027 number means not given) would have let this pass as unanswered; a quarter figure, a half-year figure and a direction are a number, and we score the miss. The rerun carries $55bn, our estimate, where the rule's default was $45bn.
6Is an HBM number disclosed?NOT GIVEN
No — adjectives only; asked for a percentage, the chief executive gave none. By the frozen rule silence is the answer: the share of the lead accelerator platform stays a secondary-sourced range.
Frozen silence rule: Adjectives only -> not-given by construction; the gauge grades disclosure posture, so silence IS the negative branch and the Vera Rubin allocation stays a secondary-sourced range
8Does management tie server memory demand to CPU servers, with a number?NOT GIVEN
Server units up high-teens in 2026 and 2027, unattributed. A context question with no model consequence.
Frozen silence rule: Not mentioned -> not-given, and the house CPU-vs-accelerator call is unaffected by this print
1Did the quarter beat on a thirteen-week basis?CLEARED
Yes — $54.2bn reported, $50.4bn on thirteen weeks, against bars of $50.0bn and $46.4bn. Calendar was 30% of the sequential growth.
2Does the next-quarter guide hold the weekly rate?CLEARED
Yes — $61.5bn, above the $46.4bn bar and the $50bn upside tell; revenue per week up 22%.
3Is the rate of price increase still positive?CLEARED
Yes — a gross-margin guide of 86.25% against an 84% bar, after about $1bn of new cost in the quarter; the company calls it the year's floor.
4Did DRAM prices rise in the quarter?CLEARED
Yes — up high-teens percent, short of the 20% upside tell and down from low-60s. Direction cleared; the deceleration is the fact.
7Did the contract program advance?CLEARED
Yes on three of four tests: 26 agreements, about $150bn of minimum revenue, $12.3bn of deposits received. The fourth, a disclosed floor price, remains unmet.
The market reads, resolved or pending. The quarter beat the consensus frozen on September 25 by 5.8% on revenue and on earnings, and the guide sits 7% above the Street's figures for the next quarter. The one prediction-market contract, a beat of $32.22 priced at 96.6%, had its threshold cleared. The options market paid ±8.2% for the event against a ±9.1% median realized move. The October 1 session delivered +3.0%: the stock opened 1.1% lower, traded 4.0% lower within the first hour, and closed at $1,097.39, well inside what was priced. It rose, where it had fallen after five of the last eight prints. The rest of the memory and storage group rose with it (SanDisk +2.7%, Seagate +2.5%, Western Digital +1.8%). Short interest and insider filings resolve over the next thirty days. The call flipped no gauge.
What would change our view (reversal)
Re-armed as dated triggers, in both directions. Toward the upside case: a gross margin at or above 86.25% in both of the next two reports with DRAM prices still rising, which is the company's own "floor" statement made good; an HBM revenue or share figure at or above the DRAM share; contract coverage reaching half of revenue with floor prices disclosed; or a durable fall of about 150 basis points in the ten-year Treasury, in which case our discount rate, not the company, was the error. Toward the downside case: a quarter-on-quarter decline in DRAM prices before the September 2027 quarter; next-quarter revenue below the $60.0 billion low end of the guide, or any guide that breaks the promised "sequential revenue growth each quarter"; a gross margin guided below the floor the company has now named; inventory above 140 days rather than declining as guided, or receivables above 80% of a quarter's revenue; fiscal 2027 capital spending guided above $60 billion without a higher revenue path; a customer publicly disputing or renegotiating an agreement; or the Chinese producer CXMT shipping DDR5 at volume into one of Micron's ten largest customers. A disclosed floor cuts either way: the recovered margin is re-derived from it, and a floor that implies less than the 60% our base case assumes, or the first decline in DRAM prices, makes the given-back run our central case.
Recommendation
Fair value $1,304 (from $1,114), range $679–$2,131, against a $1,097.39 close after the reaction session; $1,112 if the guided price step is given back. The print moved the peak out and up: a record quarter, a guide above the Street and above ours, and a contract book half again as large. It also moved the bill: capital spending above $50 billion, committed at a record margin. On the rate-preserving run the stock is below the base case; on the given-back run it is fairly priced, and the difference is a floor price the company has not disclosed. This print could not settle which run holds, and the size of any position belongs to that one unknown. Nothing here is a trade recommendation. The next test is the December report, where the company's own floor, 86.25%, is the number to hold.
Sources & Method
Print and guide facts: Micron 8-K EX-99.1 filed 2026-09-30 (archived data/raw/2026-09-30_a2026q4ex991-pressrelease.htm.txt) and the prepared remarks posted with it on the company's investor CDN (data/raw/mu_fq4_2026_prepared_remarks.pdf): segment revenue, bit and price changes, inventory days, the contract count, RPO, deposits, the next-quarter guide, the capital-spending figures and every quoted sentence not attributed to the question-and-answer session. The prior guide (operating expenses of about $1.65 billion) is the June 24 release. Call quotes are attributed to the speaker and used only where two independent transcriptions agree (AlphaStreet and MarketBeat, fetched 2026-09-30; table at data/tells/transcript.md). Derived figures (the thirteen-week equivalent, revenue per week, the calendar share of growth, bits per week on a 4–6% reading of "mid-single-digit," RPO per year over 4.25 years, capital intensity) are computed in data/score_release.py → data/release_facts.json. Street figures are the consensus frozen on September 25 (../mu-q4-fy26/data/tells/revisions.json). Prices via Robinhood (data/reaction.py): the official September 30 close, $1,065.11; the October 1 reaction session's open and low from thirty-minute regular-session bars and its official close, $1,097.39; peer moves are official close to official close. Valuation: data/model_rerun.py, the pre-note's archived engine imported unchanged, every input change documented line by line and marked as company figure, frozen rule or author estimate; results at data/model_results.json, including both literal readings of the frozen revenue rule, the given-back run and the floors-fail downside. The engine reproduces the published $554 / $1,013 / $1,875 before any input moves. The reverse-engineered cash flow uses market capitalization on shares outstanding, as in the pre-note. The scorecard: research/mu-q4-fy26/data/rubric.json (frozen 2026-09-25T18:06:55), all eight gauges scored, none added, none dropped; the capital-spending grade departs from the literal silence branch and says so in its row. Scenario weights, the 4.5% equity risk premium and the beta adjustment are house judgment inputs, unchanged from the pre-note. Standard disclosure block attaches at publication.
Disclosures
Information only. TON618 Capital. This report is for information purposes only. Nothing here is an offer to sell or a solicitation of an offer to buy any security, fund interest, or digital asset, and nothing here is personalized investment advice or a recommendation regarding any instrument.
Publisher’s exclusion. All research is published solely as general, impersonal information of regular circulation. It is not tailored to the objectives or circumstances of any individual and is not issued in connection with compensation from any client. The Fund has no clients and distributes all research free of charge. On that basis it publishes in reliance on the publisher’s exclusion from the definition of “investment adviser” under the Investment Advisers Act of 1940 (§202(a)(11)(D); cf. Lowe v. SEC, 472 U.S. 181 (1985)).
Registration & conflicts. TON618 Capital is not registered as an investment adviser or broker-dealer in any capacity. The Fund is a Bitcoin fund and may hold or transact in the securities or digital assets it discusses; material conflicts are disclosed where they exist. Ownership: the Fund holds no position in MU as of the report date, and MU is not a Fund holding or a BTC-correlated instrument. Compensation: the Fund received no compensation from any party in connection with this report and charges nothing for it.
Use of AI. Artificial intelligence is used in the creation of this research. All methodology and data integrity are reviewed and approved before publication by TON618 Capital’s Chief Investment Officer, Keyth Beck; errors may nonetheless occur, and readers should verify independently.
CFA. This report was prepared to align with CFA Institute analytical standards (methodology only). CFA® and Chartered Financial Analyst® are registered trademarks owned by CFA Institute. That reference describes the analytical framework applied; it does not imply the report was prepared, reviewed, or authored by a CFA charterholder, and the report is not issued, reviewed, endorsed, certified, or approved by — nor affiliated with — CFA Institute.
Risk & feedback. Past performance is not indicative of future results. Digital assets and equities are volatile and may result in total loss of capital. Corrections and feedback are welcome — please direct them to CIO Keyth Beck at keyth@ton618capital.com. Version 1.0.