Built by Grok 4.6 High

Micron Technology reports fiscal Q4 FY2026 after the close on Wednesday, September 30, 2026. Prepared September 24, 2026. The quarter ended in late August and has not been reported. The call is 2:30 p.m. Mountain / 4:30 p.m. Eastern the same day.

MU | Earnings Preview

FY2026Q4 has not printed. The baseline is FY2026Q3, reported June 24, 2026: revenue $41,456 million (+345.7% YoY versus $9,301 million), DRAM $31,328 million (+343.0%), NAND $9,943 million (+361.4%), non-GAAP gross margin 84.9% ($35,199 million on that revenue, +4,595 bps versus 39.0% a year earlier), and non-GAAP EPS $25.11 (+1,215% versus $1.91). Bits were a sideshow: DRAM shipments up a low-single-digit percent and NAND a mid-single-digit percent, while prices rose a low-60s percent in DRAM and a mid-80s percent in NAND. Those price comments are sequential call ranges, not a Daloopa series. The comparable growth rate is the YoY above.

The guide still in force, spoken that afternoon and never revised, is revenue $50.0 billion ± $1 billion, gross margin 86%, and non-GAAP EPS $31.00 ± $1.00. Versus FY2025Q4 that is +342% revenue, +4,032 bps of gross margin, and +923% EPS. Hitting the midpoint would stop the acceleration in the YoY rate without reversing the level: last print was +345.7% revenue and +1,215% EPS. Revenue growth would be about 390 bps slower. EPS growth would decelerate hard because the year-ago quarter was already $3.03, not $1.91. Margin would still expand a lot versus 45.7% a year ago, but the step versus the 84.9% just printed is only about 110 bps, and Mark Murphy said the outlook already reflects a meaningful moderation in the rate of price increases.

Why the rate did what it did. Revenue YoY fell from +93.3% in FY2024Q4 to a trough of +36.6% in FY2025Q3, then re-accelerated: +46.0%, +56.7%, +196.3%, +345.7%. The re-acceleration is price in a tight market, not a volume or restocking surprise. Days of inventory were 120 in FQ3, flat with FQ2. NAND flipped from −4.8% YoY in FY2025Q4 to +361% in FY2026Q3 and is now growing faster than DRAM. That mix is unguided inside a consolidated box that already assumes slower price increases.

Watch items into Wednesday

• Guidance is the bar, and the street has already adopted it. FMP’s analyst average is $50.82 billion and $31.43 (22 / 22), only +1.6% / +1.4% above the midpoint and still inside $49–51 billion and $30–32. The earnings-calendar row is $50.92 billion and $31.49. On the night of June 24 the same guide was about +16% / +22% above a Bloomberg / LSEG street near $43 billion and $25. The positive surprise versus consensus was taken in June. A print at the midpoint would miss today’s FMP mark. Five straight revenue quarters cleared the high end, and seven of the last eight EPS prints did too, but the beat rate cooled: revenue versus the midpoint went +27.6% then +23.7%; EPS +44.9% then +31.1%; gross-margin beat +692 bps then +391 bps. The hurdle that extends the streak is above $51 billion and $32, not $50 billion and $31. The street is not underwriting another 20% sandbag.
• Tariffs and trade are off the model. Murphy excluded trade or geopolitical outcomes from the guide, the same exclusion as the prior-year FQ4 call on potential new tariffs. Do not treat 86% as a tariff-stress case. The consumer crack, if allocation eases, shows up first in PC and smartphone units, which management already assumes are constrained by price.
• The stock-moving sentence is the quarter they have refused to guide. FY2027Q1 street is $56.70 billion and $35.15, +316% / +635% versus FY2026Q1 $13,643 million and $4.78 — a slower YoY rate than this quarter’s guide. S&P’s DRAM consensus near $38.3 billion is about +326% versus FY2025Q4 DRAM of $8,984 million, and the next-quarter DRAM figure of $42.8 billion is about +296% versus FY2026Q1 DRAM of $10,812 million. That is a slower year-over-year rate. Opex is the line that has not been conservative: FQ3 non-GAAP opex $1,518 million ran through a $1,400 million guide, and FQ4 steps up again to $1,650 million, the start of an about $1 billion FY2027 opex increase weighted to the second half.
• Peers already printed the cycle, and a high bar can still be sold. Samsung, SK hynix, and SanDisk all reported very large year-over-year profit increases. SK hynix revenue was +257% YoY and the shares fell 9.6% that session because revenue missed a stretched bar (Reuters). NVIDIA revenue was +105.9% YoY and Broadcom +85.5%. There is no demand-side excuse on the tape. Micron’s own last print, a +15.4% revenue beat versus FMP, was paid +15.74% the next session (Daloopa closes, June 24 to June 25) and partially given back the day after.

The management point the street may still be slow to treat as the base case is duration, not the August dollar. On August 10 Sumit Sadana said 2027 now looks tighter than 2026. He did not raise the $50 billion. RPO of about $100 billion is a contractual floor on a slice of the book, not the run-rate. Straight-lining it is less than half of one guided quarter.

Issued June 24 on the FY2026Q3 call. No company release since has revised, withdrawn, or reaffirmed the quarterly revenue, gross-margin, or EPS numbers. Non-GAAP unless noted. Classification: conservative versus Micron’s own record, and slightly below to in line with today’s street. It is not an aggressive guide. It is also no longer a gift versus consensus. Opex is the exception: aggressive in the unfavorable sense, because the last two prints ran hot and the new point estimate steps up again. Capex is a raised floor, not a sandbag — FY2026 went from $20 billion to $25 billion to $27 billion.

The revenue half-range is ±2.0% of the midpoint, in line with the ±2.0–2.8% band of the last two years. They widened the dollar box. They did not lose visibility. Over the last seven issued guides, the next quarter’s revenue and non-GAAP EPS both came in above the midpoint every time. Gross margin beat six of seven (the miss was FY2025Q2, −59 bps). Use the high end, not the midpoint, as the number they have actually been managing to on revenue and EPS. Do not assume the same cushion on opex or capex.

Above the high end on revenue starts in FY2025Q3 and has not broken: five in a row. EPS cleared the high end in every one of these eight except FY2025Q1, which was inside the range. Percent surprises peaked in FY2026Q2 and eased in FY2026Q3. Dollars versus the company’s own midpoint were still making highs through FQ3 (+$615 million, +$1,143 million, +$5,160 million, +$7,956 million on revenue over the last four). Into FY2026Q2 and FY2026Q3 the street already sat about 7% above the revenue midpoint and Micron still beat FMP by 19.5% and 15.4%. Into this print the street sits on the high end of a ±$1 billion range. Tagging $51 billion / $32 would be only about a +0.2% revenue surprise and about a +1.6% EPS surprise versus the FMP earnings endpoint.

Micron guided none of the technology or business-unit lines for FQ4. Only the consolidated box has a company number. Internal models are unavailable, so Internal is n/a. Consensus at the driver level is not in FMP. One S&P Global Market Intelligence summary (September 22) puts DRAM consensus near $38.3 billion; that is not an FMP field and is not a company guide.

Current quarter — FY2026Q4

% Diff is (consensus − guide mid) / guide mid. Positive means the street is above the guide. Last reported mix, not a guide: DRAM was 75.6% of FQ3 sales and NAND 24.0%. Cloud plus core data center was $25.3 billion, the “data center exceeded $25 billion” line. The year-ago base this print is judged against is FY2025Q4 DRAM $8,984 million (+68.7% YoY that quarter) and NAND $2,252 million (−4.8% YoY). Neither is guided now.

Eight reported quarters. The last column is the open guide, not a result. YoY is the same fiscal quarter versus the prior year.

Revenue dollars are in millions. Gross margin is non-GAAP gross profit divided by net sales in the same column. GM YoY bps use that unrounded ratio versus the same fiscal quarter a year earlier, and are blank before FY2025Q4 because the year-ago gross profit is outside this table. EPS YoY is n.m. while the prior year is a loss. DRAM and NAND YoY before FY2025Q4 were not in the Phase 2 pull and are left blank rather than estimated. Full five-year annual totals were not re-pulled in this consolidation; the quarterly stack from FY2023Q4 is in the beat record below.

Business units, same-quarter YoY only on the FQ3 compare. FQ4 is not guided. Margins are the company segment rates.

The sharpest margin catch-up is mobile and client, to 87%, level with core data center and above cloud memory. Murphy attributed that to price, not an HBM mix shift. Auto is 11% of the quarter. It cannot move a $50 billion print by itself. HBM is not a separate Daloopa series. The only quantified HBM4 figure is management’s: more than $1 billion of HBM4 12-high already shipped, with the ramp tracking twice the HBM3E 12-high ramp.

FQ+1 guide — FY2027Q1

Not guided. Management refused any revenue, EPS, or gross-margin number past FQ4. The column is the street’s standing estimate for the period ending about November 28, 2026. It is the bar a new guide will be compared with. It is not a company number and not an internal forecast. % Diff is n/a because there is no guide mid.

The only forward color that touches the quarter after FQ4 is annual: FY2027 opex up about $1 billion versus FY2026, weighted to the second half, so not an FQ1 opex step-up, and FY2027 quarterly capex above the ~$10 billion FQ4 level, with more than half of the year-over-year increase from construction.

FY+1 guide — FY2027

There is no old revenue, EPS, or margin guide for FY2027. Micron does not guide the year on those lines. Street versus the FY2026 build — three reported quarters plus the FQ4 guide mid, $128.96 billion revenue and $73.09 EPS — is +95.4% revenue and +115.8% EPS. That is consensus versus an implied year, not versus a company target. FMP’s own FY2026 averages ($129.93 billion / $73.65) are only +0.8% above that build, so the annual file is the quarterly guide, not a second opinion.

The EPS consensus range is $131.36 to $220.27. Do not anchor a model on $157.73 as if it were a guided midpoint. Revenue low / high in that file are $226.1 billion / $284.6 billion. FY2026 capex of $27 billion and the ~15% tax rate are current-year guides. They do not apply to FY2027.

The numbers still in force are the June 24 outlook. What changed after the guide is the duration language, not the quarterly model. Confidence on the shortage went up. Willingness to put a number past one quarter did not. The beat rate stopped accelerating, and the company is explicitly guiding a slower pace of price increases.

Assumptions under the box. Tight DRAM and NAND, but price growth slowing, not another step-function. Industry calendar-2026 DRAM bits were described as low-to-mid 20s, slightly above the prior outlook; NAND bits about 20%, unchanged. Micron DRAM supply about in line with the industry; NAND supply somewhat below. Share count about 1.15 billion. Tax about 15%. Opex about $1.65 billion. Ceiling prices on the large strategic customer agreements are set off calendar Q2 market prices, which management said are already in the FQ3 results and in the FQ4 guide. The guide is not a bet on a new leg of price above those ceilings for the contracted slice. About another $10 billion of customer deposits is expected in FQ4. That is financing cash flow. It does not lift free cash flow, and Murphy still expects FQ4 free cash flow to rise substantially versus FQ3’s $18.3 billion (June 24 call figure; no Daloopa id in the task file). Cash deposits are a poor base for a buyback before the December 9, 2026 CHIPS-agreement anniversary.

What they would not say. Any revenue, EPS, or margin past FQ4. A normalized 2027–2028 margin. Dollars per gigabyte on the SCA floor. How much of demand they can fill through 2028. A FY2027 capex total beyond “above” the FQ4 run-rate. Those refusals are consistent across the December, March, and June calls.

That arc is more confident on duration and less confident on incremental margin. August, from the chief business officer rather than the CEO or CFO, pushed only the duration line. It is the opposite of a guide-down. It is also not a raise of the August quarter.

Tone versus the record. They have not missed a revenue or EPS high end in this window. They have also been systematically low, and the size of the upside stopped getting larger in FQ3. A team that keeps sandbagging by 20%+ is not giving a point estimate the model should hug. A team that then says price increases are moderating, with the street already at the midpoint, is less likely to deliver another FQ2-style surprise. The credible combination is another above-high print while margin expansion slows toward the guided ~110 bps versus 84.9%. Those are not the same signal.

Post-guide updates that are not a new model. July 15 notice and the August 10 KeyBanc appearance are the only sell-side forum in the gap. August 20: Mehrotra announced Micron Research Labs, a planned $10 billion over the next decade, groundbreaking in calendar 2027. That is not a change to the $27 billion FY2026 capex guide, and it is not the FQ4 capex of $10 billion. August 26: the call date was set and held, and Manish Bhatia became president and COO with pricing, delivery, manufacturing, and capital spending. Scott DeBoer took technology and products, including Research Labs. Sumit Sadana moved to senior advisor. Mehrotra stays chairman and CEO. Murphy was not part of the announcement. That is not a CEO or CFO departure. It does put pricing under a new operator six weeks before a print whose guide already assumed slower price increases. September 23: Citi’s own model at $51 billion and $31.45, just above the FMP average, with pricing momentum described as cooling over the next several quarters. That is Citi’s model, not Micron’s, and it is not shown here as a quarter-to-quarter growth rate.

No guide was withdrawn. The release date did not move. Insider sales after the June print were under a Rule 10b5-1 plan adopted January 30, 2026. Pre-earnings red flags of that type are absent. The red flag is the bar.

New fabs are not an FQ4 volume catalyst. Idaho-1 first wafers are mid-calendar 2027. Idaho-2 is late calendar 2028. Singapore HBM packaging is first-half calendar 2027. Tongluo shipments are mid-calendar 2027. The print tests moderation, HBM4, and whether the contract book shows up in the filing.

Listen in this order: gross margin versus 86%, revenue versus the $51 billion high end, any HBM4 dollar, DRAM versus the roughly $38 billion S&P figure, the 10-K’s next-12-month RPO and the cash-deposit line, and the first FY2027Q1 guide versus $56.7 billion and $35.15. Tightness “beyond calendar 2027” is already a slogan. It only matters if the new guide, the capex number, and the RPO schedule agree with it.

Window is June 25 through September 24, after the FQ3 report and before this unreported print. Nothing in the flow revises the $50 billion / 86% / $31 guide. New wafers from the projects below do not land inside this quarter.

What changed is the shape around an unchanged print. Demand visibility is being institutionalized off-stage, mostly in auto and edge, while the four very large agreements stay unnamed. Supply response is back-end loaded. Cash and capex trajectory beyond FY2026 is up, not down, without a change to the FQ4 or FY2026 capex guides. Management continuity is intact at the top and thinner in commercial. The competitive crack is mobile DRAM, not AI memory.

Outside the window, for context only: the Anthropic supply agreement and Series H investment were announced June 22, two days before the last report. That is the named frontier-lab deal. It is not post-report news.

Micron is a consistent beater. Over the last 12 reported quarters (FY2023Q4–FY2026Q3) non-GAAP EPS beat FMP 12/12 and revenue beat 10/12. Over the last four quarters both lines are 4/4, and all four cleared the company’s own high end. There is a 12-quarter EPS beat streak and a 6-quarter double-beat streak. Dollar surprise is still widening. Percent surprise peaked in FY2026Q2 and eased in FY2026Q3. FY2026Q4 is not in the rate.

The two revenue misses never broke the company’s own guide. FY2024Q1 (−1.7%) cleared the company’s high end and missed a street number that sat above that box. FY2025Q1 (−0.06%, −$5 million) was inside the company’s range. When Micron misses the street, the street has been sitting on or above the company’s box. When Micron beats, the calls in hand attribute it to pricing running ahead of the guide, with bits a sideshow.

Green is a beat versus FMP. Red is a miss. Darker is a larger absolute surprise. Yellow is the FY2024Q2 EPS sign-flip, a beat from a loss estimate to a profit, excluded from percent averages.

Last four versus the prior four: average revenue surprise +10.4% and about +$2.6 billion, versus +2.0% and +$172 million. Average EPS surprise +19.8% and +$2.03, versus +9.0% and +$0.14. Inside the last four, dollars are still making highs on both lines through FY2026Q3 (+$4.13 of EPS and +$5.5 billion of revenue versus FMP in that quarter). Percent stepped down about 410 bps on revenue and about 1,310 bps on EPS from the FY2026Q2 peak. That is a smaller percent of a much larger base, not a return to the old +1% to +3% revenue regime.

No quarter in the twelve printed below the low end of the company’s own revenue or EPS range. Versus the prior-call midpoint, the last four revenue beats were +5.7%, +9.1%, +27.6%, +23.7%. The FQ4 street bar now sits on the guide high end. The hit-rate record supports another beat of the company’s box. The percent record is the part that can compress, because the street is no longer sitting in the middle of the box.