Micron FQ4 FY2026 re-scan: one tripwire finally fired, and it is the one that was never allowed to flip the phase on its own

Not a recommendation. No buy, sell, short, or allocation is expressed or implied anywhere below. This is indicator tracking against thresholds the vault pre-registered in June and July 2026.

The question

Verbatim: "Re-run the DRAM/HBM Phase 2→3 tripwire scoreboard against Micron's FQ4 FY2026 results (reported 2026-09-30): DIO, finished-goods days, actual vs ~86% guided gross margin, FY2027 capex level AND composition, plus a fresh Q3/Q4 2026 reading on Samsung's 250k HBM wafers/month target, fab utilization, and DRAM lead times."

Context: this is the time-critical re-score that [[2026-09-25-hbm-dram-phase-2-3-transition-q3-2026]] named as its own decisive next event. It is run one day after the print, on the 8-K and the company's own prepared remarks. The FY2026 10-K was not yet filed when this brief was written (EDGAR submissions index for CIK 0000723125 checked 2026-10-01 01:03 ET: most recent filing is the 2026-09-30 8-K, items 2.02/9.01). Every income-statement and balance-sheet figure below is therefore unaudited, and one requested metric (finished-goods days) is unavailable until the 10-K inventory footnote lands.

What we already know (from the vault)

• The thresholds are pre-registered and this brief only re-scores them. [[2026-06-28-chip-memory-cycle-phase2-phase3-indicators]] set the six indicators; [[2026-07-07-dram-hbm-phase2-phase3-early-signals]] compressed them into five tripwires plus a composite rule: declare Phase 3 only on (#2 inventory rising >2Q) AND (#4 ASP rolling over 4-6 months), or (#3 greenfield bits online / Samsung 250k early) AND (#4 HBM ASP -10% QoQ). No single tripwire flips the state.
• The prior score was 0 of 5 fired, 1 flicker, 2 unverifiable, as of 2026-09-25 ([[2026-09-25-hbm-dram-phase-2-3-transition-q3-2026]]). That brief also recorded a worry it could not resolve: the free-tier evidence hinted TrendForce had lifted its 4Q26 outlook, so the only bearish signal in the set might have reset.
• The Micron DIO series was falling, not rising. [[2026-09-07-memory-supplier-inventory-days-dio-anchor]]: 123.1 (2025-11-27) → 121.9 (2026-02-26) → 120.5 (2026-05-28), finished-goods days at a series-low 8.7. It also flagged that the #2 threshold is stated in third-party weeks of supply, which is not the same metric as filing-derived DIO.
• Capex was already shifting into concrete before this print. [[2026-09-06-vendor-capex-composition-memory-anchor]] read construction-in-progress $10,935M vs $5,518M at FY2025 year-end, against "equipment not placed into service" up only 3.7%.
• The physical Phase-3 clock starts H2 2027 at the earliest. [[2026-09-15-dram-greenfield-fab-online-dates]] dated SK hynix Yongin Y1 tools Q2 2027, Micron ID1 first wafer mid-CY2027, Samsung P5 mass production late 2028. [[2026-09-09-2018-capacity-announce-asp-rollover-lag]] measured a 12-16 month announce→rollover lag as a floor.
• The Phase-3 side of the trade now exists. [[2026-09-29-phase3-memory-cycle-rotation-basket]] closed the parent brief's largest gap by specifying ex-ante universe rules for the rotation, and found every expression pays from the hindsight date and none from the date our own rule fires. That retires the "rotation is unexecutable" finding but sharpens the cost of a false positive.

What the web says

All Micron figures below are from the 8-K Ex-99.1 earnings release filed 2026-09-30 (SEC EDGAR, retrieved 2026-10-01) and Micron's own FQ4 FY2026 prepared remarks PDF dated 2026-09-30 (retrieved 2026-10-01). FQ4 FY2026 ended 2026-09-03 and was a 14-week quarter (FY2026 was a 53-week year).

• The print beat its own gross-margin guide. FQ4 revenue $54,229M (vs $41,456M in FQ3); GAAP gross margin 86.8%, non-GAAP 87.0%, up 210bps sequentially. The guide going in was ~86%. So the #6b "margin peaks" confirm did not arrive early; margin went through the guide.
• FQ1 FY2027 is the first guided sequential gross-margin decline of the cycle, and the company pre-explained it as a cost artifact. Guide: revenue $61.5B ± $1.5B, GAAP GM ~85.95%, non-GAAP ~86.25%. The CFO states FQ1 is "the floor for gross margins in fiscal 2027," attributing the dip to FY2026 manufacturing incentive compensation absorbed into FQ4 inventory and sold through in FQ1, and guides higher margins for the rest of FY2027 "with a more moderate rate of price increases."
• DIO turned up for the first time in four quarters. Company-stated: ending inventory $10.4B, DIO 129 days, an increase of nine days sequentially (prepared remarks, 2026-09-30). Independent cross-check off the 8-K balance sheet ($10,372M inventories, $7,182M COGS): 130 days on the vault's 90-day convention, or 141.5 days on actual quarter length (98 days) versus 121.8 for FQ3's 91 days. The company convention normalises away the extra week and therefore understates the rise. Micron attributes the increase to node end-of-life build-ahead plus the capitalised incentive compensation, says "inventory levels and supply remain extremely tight," and expects DIO to decline in the coming quarters.
• Finished-goods days are not yet public. The 8-K balance sheet carries only total inventories; the raw-materials / WIP / finished-goods split lives in the 10-K inventory footnote, and no FY2026 10-K was on EDGAR as of 2026-10-01. Micron's recent 10-K filing dates (2025-10-03, 2024-10-04, 2023-10-06) put the expected window at roughly 2026-10-05 to 2026-10-08.
• FY2027 capex roughly doubles, and the increase is mostly buildings. FY2026 net capex $27.37B (gross PP&E expenditures $30.71B), above the ~$25B the vault had been carrying. For FY2027: FQ1 ~$11.5B, first-half ~$25B, and "capex to be higher in the second half" - an implied FY2027 run-rate above $50B. Composition is stated twice and unambiguously: "A majority of the increase is for construction capex, most of which is to help accelerate clean room space availability in late calendar 2028 and beyond," and "we expect a meaningfully higher growth rate in construction capex as compared to equipment capex in fiscal 2027," with only "some pull-forward of equipment spending" to optimise existing cleanrooms.
• Micron's own greenfield register confirms the vault's dates at primary and adds two. ID1 wafer output mid-CY2027; ID2 late CY2028; first New York fab initial wafer output CY2030 (concrete pour milestone in FQ4); Japan DRAM expansion groundbreaking in FQ4 with initial output late CY2028; Tongluo meaningful shipments mid-CY2027; Singapore HBM advanced packaging initial output early CY2027 (packaging, not wafers, so out of scope for #3); Singapore NAND output H2 CY2028. The remarks add the ramp caveat explicitly: new fab production "gradually becomes more meaningful starting a few quarters after initial output."
• The 4Q26 DRAM contract-price number came off the paywall, and it decelerates. "AI Server Demand Sustains Memory Contract Price Increases in 4Q26, While Consumer-Side Pressure Persists" (TrendForce, 2026-09-30, free press centre, retrieved 2026-10-01): conventional DRAM contract prices projected +10-15% QoQ in 4Q26, NAND +15-20%, and in TrendForce's own words "the pace of contract price increases is expected to moderate." Mobile DRAM allocations "remain tight"; server DRAM "undersupplied in 4Q26."
• HBM pricing moved the wrong way for the confirm tripwire. "HBM Supply Constraints Persist, 2027 Price Outlook Revised Upward with Blended ASP Forecast to Rise 121% YoY" (TrendForce, 2026-09-29, retrieved 2026-10-01): the 2027 HBM price outlook was revised upward, blended ASP +121% YoY, and 8-Hi stacks are expected to carry a 10-20% per-Gb premium over 12-Hi. Micron corroborates at primary: it has "completed agreements for the vast majority of our calendar 2027 HBM bit supply with significant price increases year over year."
• Samsung's 250k HBM wafers/month target has slipped out of 2026. Samsung plans to raise average monthly HBM wafer input from ~180,000 this year to ~250,000 in 2027 ("Samsung and SK Hynix accelerate HBM production," Sammy Fans, 2026-09-22, retrieved 2026-10-01; Counterpoint Research cited for the HBM revenue-share figures in the same piece). Source tier B - trade press relaying industry/analyst figures, not a Samsung release. This is the first post-Dec-2025 reading the vault has, and it says the target is running roughly a year behind the end-2026 date the #3 alternate leg was written against.
• Fab utilization: no memory reading obtained. Staleness is unchanged from June 2026. The only utilization print surfaced in the window is TrendForce's "Large-Generation Panel Fab Utilization Down to 79.6% in October" (2026-09-21) - LCD panel fabs, a different industry, and it must not be mistaken for a memory-fab reading.
• DRAM lead times: no verified-grade quantitative refresh. Search surfaced only distributor and procurement-blog material (findchips, a2global, utmel, versalogic) restating ">40 weeks" and "allocation controls through at least Q3 2026" without a dated primary basis - tier C, not entered as a reading. The verifiable statements are qualitative and primary: Micron, 2026-09-30, "inventory levels and supply remain extremely tight" and "we do not have line of sight to when supply and demand will return to balance"; TrendForce, 2026-09-30, mobile DRAM allocations remain tight.

The scoreboard, re-scored

Score: 1 of 5 fired on DRAM/HBM (was 0 of 5). The composite rule still cannot fire, because both composite branches require #4, and #4 has zero months of its 4-6-month clock on the board. Even a rollover beginning today puts confirmation in Q1-Q2 2027.

Convergences and contradictions

• Convergence - the parent brief's open worry resolved in the parent brief's favour, not against it. [[2026-09-25-hbm-dram-phase-2-3-transition-q3-2026]] flagged that it could not tell whether tripwire #1 had reset, and that the free-tier hints pointed toward a lifted 4Q26 outlook. The now-free TrendForce release says the opposite: the pace moderates again, to +10-15%. The flicker did not reset; it completed. Worth naming because the bear signal was the paywalled one, and when it came off the paywall it was bearish-at-the-margin rather than bullish.
• Contradiction - the only tripwire that fired and the only one that confirms are pointing in opposite directions, from the same two documents. TrendForce's 2026-09-30 release decelerates conventional DRAM pricing (fires #1) while its 2026-09-29 release raises 2027 HBM pricing (pushes #4 away). Micron does both in one PDF: "a more moderate rate of price increases" alongside CY2027 HBM bits contracted at "significant price increases year over year." The coherent reading is second derivative turning, first derivative nowhere near zero - which is exactly the state the tripwire set was designed to distinguish, and exactly why #1 was never allowed to flip the phase alone.
• Contradiction - capex is up 2x and that is evidence for Phase 2 lasting longer, not for Phase 3 arriving. The parent brief's falsification point (3) asked whether a guided capex cut would appear. The opposite printed: FY2027 implied >$50B against FY2026's $27.37B. Critically, the increase is composition-shifted toward construction aimed at late CY2028 cleanrooms, which is the same concrete-not-tools signature [[2026-09-06-vendor-capex-composition-memory-anchor]] read in the 10-Q, now confirmed in guidance. Dollars at this scale, pointed at 2028-2030 shells, do not produce 2027 bits. The announce→rollover lag from [[2026-09-09-2018-capacity-announce-asp-rollover-lag]] applies from the tool date, not the announcement.
• Contradiction - the one tripwire whose refresh the backlog item specifically bundled came back worse for Phase 3, not better. Samsung's 250k wpm was the #3 alternate leg precisely because it was the one way #3 could fire before greenfield concrete. It slipped a year. That removes the fast path.

Synthesis for RDCO

The headline is narrow and it should stay narrow: one of five tripwires fired, it is the earliest and noisiest one in the set, and the composite rule that governs the phase call is arithmetically unable to fire. The Phase 2→3 transition evidence did move this run - it moved for the first time since the set was written - but it moved on the indicator the vault itself classified as early warning rather than confirmation. Tripwire #1 is now satisfied on its literal text: 58-63% → 13-18% → 10-15% is two consecutive decelerations, TrendForce's own prose calls the pace moderating, and Micron's guidance independently says "a more moderate rate of price increases." Everything else either held its July verdict or moved away from Phase 3. If the discipline of a pre-registered scoreboard means anything, it means scoring #1 as fired without letting that score leak into the phase call.

Two readings earn their keep beyond the scoreboard. First, the DIO turn is real but one quarter old and heavily explained. 120.5 → 129 days is the first rise in four quarters, and on actual quarter length (98 days versus 91) the rise is nearly 20 days rather than 9, because Micron's own ~90-day convention normalises away the 53rd week. That is the single most Phase-3-shaped datapoint in the print. Against it: the company names two non-cyclical causes (node end-of-life build-ahead, manufacturing incentive compensation capitalised into inventory), says supply remains "extremely tight," and guides DIO down. The threshold needs two consecutive quarters and is written in third-party weeks of supply, not DIO. So the correct posture is that #2's clock started on 2026-09-03 and FQ1 FY2027 (reporting roughly late December 2026) is the quarter that either confirms it or kills it. That is now the decisive next event, the way this print was.

Second, the capex composition finding is the one that should actually change how the thesis is held. FY2027 capex implied above $50B, against $27.37B in FY2026, with the majority of the increase in construction aimed at cleanroom availability in late calendar 2028 and beyond. This is the single clearest statement yet that the supply response to this cycle is being built as buildings first and tools later, by choice and with Strategic Customer Agreement visibility behind it. For the capital-cycle frame that is a duration finding, not a direction finding: it pushes the plausible bits-online window later and widens it, which lengthens the interval over which Phase 2 can persist while simultaneously enlarging the eventual overshoot. It also makes the "HBM broke the cycle" plateau variant in [[2026-09-09-hbm-plateau-hypothesis-backtest-variant]] harder to dismiss, because contracted-forward HBM pricing plus SCA-backed capex is a structurally different supply-response mechanism from 2017's.

Calibration. High confidence (~90%) on the mechanical claim that one of five tripwires fired, the composite cannot fire, and #4 has zero months on its clock - that is threshold-matching against primary filings and dated releases read this run. Moderate confidence (~55%) that the turn is still three or more quarters out, which is weaker than last run's 60% because #1 completing and DIO turning are both genuine, if thin, movement. I am explicitly not converting "one early tripwire fired" into "the turn is near," and I am not revising the 2026-09-25 position posture: nothing in this print licenses a rotation under the pre-registered rules, and [[2026-09-29-phase3-memory-cycle-rotation-basket]] independently found that acting early on these rules has historically cost money even when the eventual direction was right. Two of eight scoreboard lines remain UNVERIFIED (fab utilization, lead-time weeks), and both have been dark since June 2026, so any claim about confirming features is being made on a partial board.

What resolves next, in date order. The FY2026 10-K, expected roughly 2026-10-05 to 2026-10-08, delivers finished-goods days (#2b, off the 8.7 floor or not) and the construction-in-progress versus equipment-not-placed-in-service split that turns the capex composition claim from guidance language into a balance-sheet reading. FQ1 FY2027 (late December 2026) is the second-quarter test for #2's DIO clock and the first test of whether FQ1 really was the gross-margin floor. The 4Q26 conventional-DRAM print, when it lands as actuals rather than forecast, removes the one caveat on #1's firing.

Why this is in the vault

This is the pre-registered re-score the 2026-09-25 brief named as its own decisive next event, and it is the first run in which the tripwire set produced a non-zero score - so it is the row the memory-cycle-v1.1 thesis and the Markov pipeline's Layer-1 phase tracker will diff the next three quarters against. It also fixes the exact date and metric of the next decision point (FQ1 FY2027 DIO, late December 2026) and records which two indicators the vault still cannot read at all.

Open follow-ups

• Read the FY2026 10-K when it files (expected ~2026-10-05 to 2026-10-08) for the three figures the 8-K cannot give: finished-goods days against the 8.7-day floor, construction-in-progress versus equipment-not-placed-into-service at FY2026 year-end, and the audited restatement of any figure used above.
• Re-score tripwire #2 on Micron FQ1 FY2027 (reports late December 2026). A second consecutive DIO rise is the single highest-information event on the board; a decline per company guidance kills the clock. Pre-commit the verdict now so it is not re-argued after the fact.
• Instrument fab utilization and DRAM lead times, or formally retire them from the scoreboard. Both have been UNVERIFIED for two consecutive runs and four months. Either find a repeatable free source with a dated series, or amend the tripwire set to stop scoring features the vault cannot read - carrying permanently-dark indicators inflates the apparent board size.
• Confirm Samsung's 250k HBM wpm slippage against a tier-A source (Samsung IR, earnings call, or a TrendForce release rather than trade press relaying it). The current reading is tier B and it is doing real work: it removes the only fast path to firing #3.
• Feed the 4Q26 TrendForce conventional-DRAM print into anchors/dram-spot/. The newest row is 2026-04-24, so the series is five months stale even though the number that resolved tripwire #1 is now on the free tier - the gap is ingestion, not access.
• Decide whether the server-DRAM → conventional-DRAM basket switch invalidates the #1 firing. The 58-63% and 13-18% readings are server DRAM; the 10-15% is conventional DRAM. Either find a single consistent series across all three quarters or write down explicitly that the tripwire is scored on mixed baskets.
• Re-test the HBM plateau hypothesis against SCA-contracted forward pricing. Micron having locked the vast majority of CY2027 HBM bits at significant YoY increases is a mechanism the plateau variant did not model, and it bears directly on whether #4 can ever fire on the HBM leg inside this cycle.

Related

• [[2026-09-25-hbm-dram-phase-2-3-transition-q3-2026]] - the parent brief; the 0-of-5 baseline this run diffs from, and the source of the five tripwires re-scored here
• [[2026-07-07-dram-hbm-phase2-phase3-early-signals]] - the five-row tripwire scoreboard and the composite rule
• [[2026-06-28-chip-memory-cycle-phase2-phase3-indicators]] - the original six-indicator threshold set
• [[2026-09-07-memory-supplier-inventory-days-dio-anchor]] - the Micron DIO series this run extends to 129 days
• [[2026-09-06-vendor-capex-composition-memory-anchor]] - the capex-composition method; its CIP finding is now confirmed in FY2027 guidance
• [[2026-09-15-dram-greenfield-fab-online-dates]] - the per-fab online register, confirmed at primary and extended with ID2 and Japan
• [[2026-09-08-cxmt-ymtc-bit-share-phase3-threshold]] - the 5a/5b split and the falsifiability table whose capex-cut prediction this print contradicts
• [[2026-09-09-2018-capacity-announce-asp-rollover-lag]] - the 12-16 month announce→rollover lag used as a floor
• [[2026-09-09-hbm-plateau-hypothesis-backtest-variant]] - the plateau variant that forward-contracted HBM pricing strengthens
• [[2026-09-29-phase3-memory-cycle-rotation-basket]] - the now-specified Phase-3 universe rules, and the finding that early action on them has cost money
• [[2026-09-26-financing-led-capital-cycle-phase3-precedent]] - the financing-led transition precedent, relevant to SCA-backed capex
• [[2026-05-27-markov-equities-pipeline-spec]] - the Layer-1 phase tracker this scoreboard feeds
• [[2026-05-18-memory-cycle-v1.1]] - the thesis and its anchor-based exit logic
• [[2026-05-18-hbm3e-hbm4-capacity-timeline-phase-b-end]] - where the Samsung 250k wafer milestone was first dated

Sources

Primary (read this run, 2026-10-01)

• Micron 8-K, filed 2026-09-30, Ex-99.1 FQ4 and FY2026 earnings release - SEC EDGAR - unaudited; quarter ended 2026-09-03
• Micron FQ4 FY2026 Earnings Call Prepared Remarks, 2026-09-30 (PDF) - DIO 129 (+9), FY2027 capex level and composition, fab online dates, FQ1 guidance rationale
• SEC EDGAR submissions index, CIK 0000723125 - checked 2026-10-01 01:03 ET: no FY2026 10-K on file; prior 10-K filing dates 2025-10-03 / 2024-10-04 / 2023-10-06

Third party (dated, retrieved 2026-10-01)

• AI Server Demand Sustains Memory Contract Price Increases in 4Q26, While Consumer-Side Pressure Persists - TrendForce, 2026-09-30 - conventional DRAM +10-15% QoQ 4Q26; NAND +15-20%; pace "expected to moderate". Now free tier, previously the paywalled gap in the 2026-09-25 run
• HBM Supply Constraints Persist, 2027 Price Outlook Revised Upward with Blended ASP Forecast to Rise 121% YoY - TrendForce, 2026-09-29
• Samsung and SK Hynix accelerate HBM production - Sammy Fans, 2026-09-22 - tier B; Samsung ~180k HBM wpm in 2026 → ~250k in 2027; Counterpoint HBM revenue share 21% Q1 → 33% Q2 2026
• Large-Generation Panel Fab Utilization Down to 79.6% in October - TrendForce, 2026-09-21 - LCD panel fabs, not memory; listed only to record that it is not a #6a reading

Not retrieved / flagged

• 403 FORBIDDEN, skipped not retried: investors.micron.com (403 on all paths, 2026-10-01 - the prepared remarks were obtained from the q4cdn mirror instead); news.alphastreet.com Micron Q4 FY2026 transcript (403); sammyfans.com via WebFetch (403, obtained via Playwright)
• NOT OBTAINED: any dated memory-fab utilization reading or primary DRAM lead-time series for Q3/Q4 2026. Distributor and procurement-blog sources (findchips, a2globalelectronics, utmel, versalogic) were surfaced and rejected as tier C - undated or rolling claims with no primary basis
• NOT YET AVAILABLE: Micron FY2026 10-K inventory footnote (finished-goods days) and PP&E footnote (construction-in-progress) - expected ~2026-10-05 to 2026-10-08

Vault

• ~/rdco-vault/06-reference/research/2026-09-25-hbm-dram-phase-2-3-transition-q3-2026.md
• ~/rdco-vault/06-reference/research/2026-07-07-dram-hbm-phase2-phase3-early-signals.md
• ~/rdco-vault/06-reference/research/2026-06-28-chip-memory-cycle-phase2-phase3-indicators.md
• ~/rdco-vault/06-reference/research/2026-09-07-memory-supplier-inventory-days-dio-anchor.md
• ~/rdco-vault/06-reference/research/2026-09-06-vendor-capex-composition-memory-anchor.md
• ~/rdco-vault/06-reference/research/2026-09-15-dram-greenfield-fab-online-dates.md
• ~/rdco-vault/06-reference/research/2026-09-08-cxmt-ymtc-bit-share-phase3-threshold.md
• ~/rdco-vault/06-reference/research/2026-09-09-2018-capacity-announce-asp-rollover-lag.md
• ~/rdco-vault/06-reference/research/2026-09-09-hbm-plateau-hypothesis-backtest-variant.md
• ~/rdco-vault/06-reference/research/2026-09-29-phase3-memory-cycle-rotation-basket.md
• ~/rdco-vault/06-reference/research/2026-09-26-financing-led-capital-cycle-phase3-precedent.md
• ~/rdco-vault/01-projects/investing/2026-05-27-markov-equities-pipeline-spec.md
• ~/rdco-vault/01-projects/investing/theses/2026-05-18-memory-cycle-v1.1.md
• Stale price anchor - ~/rdco-vault/01-projects/investing/anchors/dram-spot/trendforce-historical-2017-2026.csv (newest row 2026-04-24, verified 2026-10-01)

Correction 2026-10-03. This rescan's Tripwire #1 caveat - "the basket changes from server DRAM to conventional DRAM," verdict "FIRED (qualified)" - rested on a mislabel inherited from [[2026-09-25-hbm-dram-phase-2-3-transition-q3-2026]], not on the data. All three readings (2Q26 +58-63%, 3Q26 +13-18%, 4Q26 +10-15%) are conventional DRAM contract prices in the TrendForce releases that originated them, verified against primary sources on 2026-10-03. There was no basket switch, so the qualification on the firing does not apply and the second derivative is measuring a price turn rather than a product-mix change. The reasoning in this doc was sound; its input label was not. NOT APPLIED HERE: the basket-naming rewrite of row 1 is written up as a labelled proposal awaiting the founder's read, not a change to the scoreboard. See [[2026-10-03-dram-asp-series-basket-consistency-tripwire-1]].