Micron $MU FQ3'26 — Ranked Guidance Clues print Tue 2026-06-24 after close · guides FQ4'26

The reported FQ3 quarter is priced; the FQ4 GUIDANCE is the swing factor. These are the data/clues to gather BEFORE the print, merged from two research passes and ranked by impact = predictive weight on the FQ4 guidance call × accessibility before the print. Per the ACH lens (the corpus is ~90% bull and largely circular), leading and disconfirming signals carry the real edge and get a premium — the genuinely-new, decisive datum is whether management extends "sold out" to CY2027.
⚑ Live 6/24 AM: MU −13.2% to $1,051.77 / ~$1.19T (from the $1,211 close); the ±11% band re-strikes to ~$936–$1,168 (spot now ~10% below the ~$1,172 options reference). But the morning leading data is HOT — Korea June 1–20 chip exports nearly tripled and DDR5 spot/contract is still rising — so the selloff is positioning + a fresh June-23 demand wobble, not a fundamentals roll-over (see the two new panels below). Lead with the big picture; expand any row for why-it-predicts-the-guide + where-to-get-it.

25 merged rows 12 leading 6 disconfirming 3 confirming 4 contrarian prepared 2026-06-23 (day before print)
Reported FQ3 (priced)
~$34.5B / ~$19.7
rev / non-GAAP EPS · GM ~81–82%
Micron's own guide: $33.5B ±$750M / 81% / $19.15
FQ4 GUIDANCE BAR (the number that matters)
~$43B / ~$25.4
Street model rev (19 analysts) / EPS (17) · GM inferred low-to-mid 80s
no clean public GM consensus — FLAGGED
Options-implied move
±11% (±$116)
recomputed off live $1,052 → range ~$936–$1,168
options ref was ~$1,172 (front-wk IV 155%); spot now ~10% BELOW ref — band re-struck off the 6/24 AM tape · IV crush likely
Stock / positioning (LIVE 6/24 AM)
$1,051.77 / $1.19T
−13.2% from the $1,211 close (6/24 AM range $1,038.50–$1,125)
now WELL BELOW consensus PT ($1,186 mean / $1,210 median / $1,625 high / $400 low) · froth fully out pre-event · no overnight PT cuts · 50d avg $748.60
BEAT — stock up
rev guide >= $44-45B AND GM >= ~83-85% AND/OR EPS >= ~$26 PLUS explicit CY2027-sold-out / pricing-accelerating language
IN-LINE — flat-to-down (given the run)
~$42-44B rev / ~81-82% GM / ~$25 EPS -- meets the model, risks sell-the-news given the +54%/30d run
MISS — down hard
rev guide < ~$41B OR flat/down GM OR cautious CY2027 pricing-normalization / coordinated-capacity-additions language

Morning-of last-mile signals — ran 6/24 AM (~09:58 UTC) · the highest-frequency leading reads, live

Ran the morning-of last-mile pull (6/24 AM, ~09:58 UTC). The leading data is HOT and one-directional — memory demand/pricing is still accelerating into the print. This is the strongest near-term *confirming* read in the whole corpus, and it lands the same morning as a −13% stock and a fresh June-23 demand wobble. Net: the reported quarter + the FQ4 guide are supported by the highest-frequency leading indicators; the bear is now purely a *forward / 2027 / multiple* story, not a near-term fundamentals roll-over.
① Korea June 1–20 export flash — the single best leading read strongly bullish
Chip exports nearly TRIPLED to $25.5B (June 1&ndash;20) on AI-memory demand; semiconductors = 41.2% of all exports (+18.3pp YoY). Total Korea exports +60.4% YoY to a record $62B (vs $38.6B yr-ago). Korea's 20-day flash is the cleanest real-time proxy for the global memory cycle — this is an *acceleration*, not a peak-rollover signature. Directly corroborates a confident, runway-extending FQ4 guide.
② TrendForce spot + contract (June) bullish pricing
DDR5 spot tight, buyers accepting higher quotes; mainstream DDR4 spot +2.22% WoW ($35.12→$35.90, Jun 3→9). PC DRAM contract prices grew significantly in Q2, extending into Q3 & Q4. Crucially, TrendForce: DDR5 contract profitability in 2026 is expected to SURPASS HBM3e — which independently *confirms the margin-inversion flag* (commodity DDR5 is carrying margin, not just HBM).
③ Live MU tape + PT consensus (6/24 AM) priced-for-perfection
MU $1,051.77, −13.2% from the $1,211.38 close (AM range $1,038.50–$1,125); mkt cap ~$1.19T; 50d avg $748.60. PT consensus UNCHANGED overnight — mean $1,186 / median $1,210 / high $1,625 / low $400 → no analyst PT cuts despite the selloff. So the drop is positioning + the June-23 wobble, *not* sell-side downgrades. Stock now sits ~11% below the mean PT.

The June 23 demand-side crack — SK Hynix −12% on NVIDIA/Rubin + HBM4 reports (verified, contested)

The one genuinely-new DISCONFIRMER, verified (read-only WebSearch). On June 23 SK Hynix fell ~12% in a single session on reports that NVIDIA may cut its Rubin production forecasts and that SK Hynix is slowing HBM4 capacity expansion and reallocating wafers to commodity DRAM — read by the tape as the first sign AI-memory demand could *soften*. MU fell on the same setup (our −13%). This is exactly the 2027-normalization / bullwhip tell the bear case predicted — and it lands the day before the print. It does not touch the *reported* quarter (Korea/TrendForce show pricing still hot), but it is the freshest crack in the *forward* demand story.
What the reports say
SK Hynix slowing HBM4 expansion; TrendForce note (same window) trimmed FY HBM4 outlook ~4.5B→4.0B Gb; April DigiTimes: SK Hynix may cut NVDA HBM4 shipments ~20–30% amid Vera Rubin ramp delays. The reallocation toward commodity DRAM is *rational* given DDR5 contract margins now rival HBM (see morning panel ②) — i.e. it's partly a margin-mix move, not purely a demand verdict.
⊕ The contested / tempering facts
These are reports — NVIDIA has not officially confirmed a Rubin cut. They sit *against* a June 2 report that Vera Rubin had entered "full production" and was set to ship to AWS/Azure/GCP/Oracle that summer (Samsung/SK Hynix/Micron all named HBM4 suppliers). So the signal is genuine but contested — a forecast trim and a margin-driven reallocation, not a confirmed demand collapse.
⚑ The tell on tonight's call
If Micron echoes any Rubin-demand softness, trims HBM4 allocation language, or sounds hedged on 2027 HBM → the crack is corroborated (bearish forward). If Micron reaffirms HBM "sold out," raises HBM4 allocations, or names incremental hyperscaler SCAs → it refutes the crack (bullish). This is the cleanest single disconfirming/confirming fork in the event.

The big ranked table — 25 clues, impact-ranked 1→25 (each row expands)

leading disconfirming contrarian confirming pre-print ✓
1 Explicit CY2027 HBM 'sold out' / contracted language on the call (the #1 call-watch) Company-specific confirming partial 44 src
impact: THE single highest-impact line. Pre-print everything else is priced; the one genuinely-new, decisive datum is whether management extends sold-out language to CY2027. Explicit 'CY2027 contracted/sold out' = cleanest BEAT tell; 'monitoring 2027 supply additions' = cleanest MISS tell. Not fully pre-print, but it IS the swing variable so it ranks #1.
reading: CY2026 HBM sold out (price+volume, incl HBM4); 'supply tight well beyond 2026'; customers moving to 3-5yr supply agreements. NO explicit 'CY2027 sold out' quote exists as of 2026-06-23 -- FLAGGED, this is the new info that prints 6/24.
Why it predicts the guide
Removes (or re-introduces) downside variance on the largest, highest-margin part of the FQ4/FY27 revenue line. CY2026 is already locked; the marginal pricing-power question is entirely about 2027 contracting status.
Where to get it
Micron FQ3 call verbatim 6/24; pre-print proxies: Mehrotra at Bernstein (late May, '50-66% of demand', 'well beyond 2026'); HBM TAM ~$100B/2028; UBS LTA note.
Merged from
fresh#7, corpus C1 LTA, corpus claim#1
Accessible before print
Partial (the UPDATE comes ON the call 6/24; pre-print only 'tight well beyond 2026' + 3-5yr SCAs)
2 DRAM/NAND SPOT-minus-CONTRACT spread and its DIRECTION (rate-of-change), weekly, split DDR5 / DDR4-legacy / HBM3E Supply-pricing leading pre-print ✓ 48 src
impact: Spot LEADS contract by 1-3 months -> the single most diagnostic LEADING read on what management can credibly guide for FQ4 ASPs. Widening spot premium = shortage intensifying = guide up; spot rolling over while contract catches up = first crack of a cycle top = guide-disappointment risk even on a strong reported quarter. Freshest high-frequency tell, fully pre-print.
reading: DDR4 1Gx8 3200 spot $33.60 -> $34.80 (Jun 3, +3.57%) -> $35.90 (Jun 10, +2.22%); DDR5 'especially brisk', gains every week in June; was -0.25% early May => clear acceleration. Caveat: thin volume as buyers hesitate at higher quotes. June 17 spot NOT yet indexed -- pull AM of 6/24.
Why it predicts the guide
Contract prices ARE Micron's realized ASP; spot is the leading edge of the next contract settlement that sets FQ4 revenue and gross margin.
Where to get it
TrendForce/DRAMeXchange weekly spot + monthly contract by product; module-maker (ADATA/Team Group/Transcend) ask-vs-realized gap; distributor 'no allocation' counts.
Merged from
fresh#3, corpus B1, corpus claim#3
Accessible before print
Yes (next TrendForce spot may publish AM of 6/24)
3 LTA terms from primary filings: FIXED-price take-or-pay vs fixed-VOLUME-at-spot vs cost-plus (8-K Item 1.01 exhibits + 10-Q purchase-commitment footnotes, NOT call paraphrase) Company-specific confirming pre-print ✓ 44 src
impact: THE load-bearing bull claim ('CY2026 sold out at fixed prices'). If LTAs are fixed-VOLUME at re-setting prices, the FQ4 guide is still a spot-cycle bet and 'cycle structurally broken' is wrong. Confirmed fixed-PRICE take-or-pay = FQ4 revenue largely pre-booked = management can guide with unusual confidence. Grade-A primary source, fully pre-print -> top tier.
reading: FQ1 call characterized LTAs as 'much stronger contract structure... not like anything before'; UBS says LTAs cover the 'vast majority of core customers'. 'Six customers' / fixed-PRICE vs fixed-VOLUME is UNVERIFIED and the sibling claims are mutually contradictory -- to pull from 8-K exhibits.
Why it predicts the guide
Determines whether the FQ4 revenue line is contractually locked (low-variance high guide) or still a spot bet (variance stays in the guide).
Where to get it
SEC EDGAR Micron 8-K Item 1.01 LTA exhibits + 10-Q/10-K commitments footnotes; reciprocal purchase-commitment disclosures in NVIDIA/hyperscaler 10-Qs; the FQ3 call verbatim on the 'sold out' horizon.
Merged from
corpus C1, corpus claim#1, fresh#7
Accessible before print
Yes (EDGAR 8-K/10-Q already filed)
4 China supply ramp: CXMT (DRAM/DDR5/LPDDR5) and YMTC (NAND) wafer-starts, yield/node, tier-1 OEM qualification, commodity-tier pricing Supply-pricing disconfirming pre-print ✓ 47 src
impact: The single highest-value DISCONFIRMER and the variable ~90% of the corpus ignores. Any sign CXMT/YMTC commodity DDR4/DDR5/NAND is hitting tier-1 qual caps the very commodity-price upside management would otherwise guide into FQ4. Premium-weighted by the ACH lens; fully pre-print.
reading: To pull. The bull guide rests on a 3-company oligopoly with 'no new supply until 2027-28'. No confirmed tier-1 qual of CXMT/YMTC commodity output sourced as of 2026-06-23.
Why it predicts the guide
Breaks the oligopoly-discipline premise that underwrites a confident price-led raise; commodity DRAM/NAND is ~half the mix.
Where to get it
China customs/equipment-import data; CXMT IPO prospectus + Big-Fund subsidy filings; TrendForce CXMT capacity trackers; retail-module teardowns for CXMT-marked dies; export-control license status.
Merged from
corpus B2, corpus claim#8
Accessible before print
Yes (customs/equipment data + TrendForce trackers)
5 Korea + Taiwan monthly memory-IC EXPORT data (Korea MOTIE 20-day + full-month; Taiwan MOEA), decomposed VALUE vs VOLUME (implied ASP), incl June flash Supply-pricing leading pre-print ✓ 18 src
impact: Government-grade-A, near-real-time, un-spinnable leading proxy for the whole DRAM/HBM complex weeks before the company prints. The value-vs-volume split is the killer diagnostic: rising value + flat/falling volume = pure price (sustainable, supports a price-led raise); rising volume = supply ramping (bearish forward pricing). The June 20-day flash is the freshest pre-print read.
reading: April: semis $31.9B +173.5% YoY (record). May: semis $37.16B +169.4% YoY (record); first-20-days chips +202% YoY. 3 straight months total exports >$80B; accelerating. June full-month NOT yet released; June 1-20 flash to pull AM of 6/24.
Why it predicts the guide
Korean chip exports are ~Samsung+Hynix memory and cannot be guided/spun; they coincident-confirm the price+volume surge feeding Micron's quarter.
Where to get it
Korea MOTIE/KITA 20-day export releases (HS 8542.32); Taiwan MOEA export-orders; KITA portal; Korea Times / Bloomberg early-trade.
Merged from
fresh#4, fresh#13, corpus B3, corpus claim (Korea)
Accessible before print
Yes (latest 20-day flash lands ~Jun 21-22 KST, before 6/24)
6 Hyperscaler CY2026/27 capex DELTAS (the QoQ revision) + the verbatim language shift to 'memory pre-paid / supply-secured / long-term-committed' (MSFT/AMZN/GOOGL/META/ORCL) Demand leading pre-print ✓ 43 src
impact: The demand engine read one hop up the chain. Capex guided UP plus management explicitly blaming MEMORY cost/scarcity means demand is price-inelastic and supply-locked -- the ideal backdrop for Micron to guide pricing power higher. A capex air-pocket or 'digesting' language is the disconfirmer that would cap the guide. Strong leading signal, fully pre-print, but ~2 months old.
reading: MSFT CY26 capex ~$190B incl '~$25B from higher component pricing', 'constrained at least through 2026'. AMZN ~$200B, Jassy: 'cost of components, particularly memory, has skyrocketed... not enough capacity.' GOOGL RAISED to $180-190B 'compute constrained.' META RAISED to $125-145B citing 'higher component costs.' Big-four ~$710B 2026; MS sees ~$805B(26)/$1.1T(27). Memory ~30% of hyperscaler AI capex (~4x vs 2023).
Why it predicts the guide
Hyperscaler capex funds the AI servers that consume Micron DRAM/HBM; the QoQ revision + memory-language migration front-runs Micron's LTA disclosures.
Where to get it
MSFT/AMZN/GOOGL/META/ORCL most-recent 10-Qs + earnings-call transcripts (capex line + memory mention-count/tone); Morgan Stanley/Moody's aggregate capex notes.
Merged from
fresh#5, corpus A1, corpus claim#5 (allocation)
Accessible before print
Yes (all calendar-quarter calls already reported)
7 Gross-margin trajectory + DECOMPOSITION: FQ2 actual 75% -> FQ3 guided ~81% -> implied FQ4 mid-80s; HBM-mix (structural) vs commodity-spike + LCM write-down REVERSALS (cyclical) Company-specific disconfirming partial 17 src
impact: GM is THE most-watched line and the cleanest read on pricing-power durability. A FQ4 GM guide stepping clearly above 81% toward mid-80s = beat; flat/down = 'pricing power peaking' = the primary miss trigger. The decomposition (structural HBM mix vs cyclical commodity + write-down reversal) is the disconfirmer that tells you if the GM guide is durable. High impact; partly pre-print.
reading: FQ2 non-GAAP GM 75% (record); FQ3 guided ~81%; FMP FQ4 model ($43.14B/$25.39 EPS) implies mid-80s. No clean published FQ4 GM consensus -- FLAGGED as inferred. Decomposition (whether 81% needs a commodity spike + one-time reversals to hold) is the swing-down test.
Why it predicts the guide
With contract prices still rising on a largely fixed cost base, GM should keep climbing; whether it's mix-driven or spike-driven decides durability of the whole guide.
Where to get it
Micron FQ2 press release; segment margins (CMBU/CDBU/MCBU/AEBU) + MD&A LCM disclosures in the 10-Q; peer op-margins (Hynix 72%, Samsung) as credibility check; back-solve the bridge.
Merged from
fresh#8, corpus C3
Accessible before print
Partial (FQ4 GM set on the call; decomposition needs the FQ3 10-Q; leading contract-price clues live now)
8 HBM4 qualification + Micron's NVIDIA (Rubin/VR200) allocation SHARE vs SK Hynix/Samsung; per-GPU content step (Blackwell 192GB->Rubin 288GB HBM4); HBM4 ramping '2x faster than HBM3E' Demand leading pre-print ✓ 48 src
impact: HBM4 is Micron's highest-margin/highest-ASP product and the next leg of the guide. Confirmed Rubin volume slot removes the 'did Micron get into Rubin?' risk and the per-GPU content jump (+50% capacity, ~2.75x bandwidth) is structural content growth. Faster-than-HBM3E ramp = upside to revenue + margin. Contrarian flip: 3-way qual hands NVIDIA pricing leverage = an HBM-margin risk that could temper the GM guide. Strong leading, fully pre-print.
reading: Jun 2-5 Huang: all three vendors qualified & in HBM4 mass production for Vera Rubin, 'competing fiercely.' Micron HBM4 36GB 12-Hi volume shipments began CQ1'26 (>11Gb/s, >2.8TB/s/stack); 48GB 16-Hi samples out. Rubin R100 = 288GB HBM4/~22TB/s vs Blackwell 192GB HBM3E/~8TB/s. Bhatia (May 20): HBM4 ramping 2x faster than HBM3E. Allocation est: Hynix ~60-70%, Samsung ~25-30%, Micron the balance (fluid). The Jan-2026 'Micron out of initial Rubin' bear thesis is STALE/superseded.
Why it predicts the guide
Converting HBM4 qual into a real volume Rubin slot is where the FQ4 HBM revenue-trajectory upside lives; content-per-GPU decides whether the story is 'more chips' (cyclical) or 'more memory per chip' (structural raise).
Where to get it
Jensen GTC keynotes (Jun 1-5); Micron IR HBM4 PR (Mar 16, volume CQ1'26 for Vera Rubin); Bhatia at JPMorgan (May 20); SemiAnalysis/TrendForce HBM allocation trackers; NVIDIA platform BOM; the FQ3 call (HBM4 share + VR200 vs CPX split).
Merged from
fresh#6, corpus A2, corpus A3, corpus claim#4
Accessible before print
Yes (Jensen GTC confirmation + Micron PR + JPM conf)
9 SK Hynix / Samsung forward commentary: 2027 supply gap WIDER than 2026, HBM/2026 sold out, refusing long-term fixed-price deals; AND Hynix as the LEADING peer read-through (reports BEFORE Micron, server-DRAM margin >80%) Supply-pricing leading pre-print ✓ 8 src
impact: The #1 and #2 players lead/coincide with Micron's tone; if they say demand outstrips supply through 2027 and won't lock fixed prices (to keep ratcheting), Micron almost certainly echoes and guides FQ4+FY27 strong. Hynix prints first, so its margin trajectory + post-earnings reaction front-run Micron's guide by a quarter and set the sympathetic-fade risk. Highest-signal peer tell, fully pre-print.
reading: SK Hynix: customer HBM requests 'already exceed planned production capacity for the next three years'; Q1'26 record rev W52.58T (~$35.5B, +198% YoY), 72% op margin. Samsung's Kim Jaejune: based on orders in hand, 2027 gap WIDER than 2026; HBM 2026 entirely sold out; chip op profit W53.7T (~49x YoY). Both refusing long-term fixed-price deals. The only moderating note: SK Hynix 'rate of increase may slow after this quarter' -- a pace, not a reversal.
Why it predicts the guide
Peer forward statements are the cleanest external corroboration of the supply/demand balance management is implicitly forecasting; Hynix's print sets the prior and the reaction template for MU.
Where to get it
SK Hynix Q1'26 call (2026-04-23) + 000660.KS price/reaction; Samsung Q1'26 call (2026-04-30); Korea-broker estimates; MU-vs-Hynix beta on event days.
Merged from
fresh#1, corpus B5, corpus claim#6
Accessible before print
Yes (Q1'26 calls done; Hynix prints first)
10 Insider Form-4 selling cadence + 10b5-1-vs-discretionary split for Mehrotra/CFO/Section-16 officers across the run ('63 trades Feb-May, ~100% sells') Positioning disconfirming pre-print ✓ 39 src
impact: Cheap, primary, one-hop DISCONFIRMER in a ~90%-bull crowd. Clustered DISCRETIONARY (non-10b5-1) selling into the parabola right before the print is the classic late-cycle tell that insiders don't believe the forward guide sustains the price; pre-scheduled 10b5-1 is noise. ACH-premium signal, fully pre-print, grade-A source -- but it tips conviction, not the guide level itself.
reading: To pull/parse. 'Everything Money' flags 63 trades Feb-May, ~100% sells -- but the 10b5-1-vs-discretionary split is the diagnostic and is NOT yet separated as of 2026-06-23.
Why it predicts the guide
Insider conviction in the outlook is revealed by whether sells are discretionary (bearish) or mechanical 10b5-1 (noise) into the print.
Where to get it
SEC EDGAR Form 4 (parse transaction codes: open-market 'S' vs 10b5-1 footnote); dollar-weight and % of holdings; benchmark vs prior 'beat' quarters / prior cycle tops.
Merged from
corpus D1, corpus claim#9
Accessible before print
Yes (EDGAR Form 4)
11 DRAM contract price QoQ settlements: Q1'26 +93-98%, Q2'26 +58-63%, Q3'26 (=Micron FQ4) increases CONTINUING (TrendForce June 22) Supply-pricing confirming pre-print ✓ 12 src
impact: Contract prices ARE Micron's realized ASP -- the direct driver of revenue and gross margin. Record QoQ jumps STILL rising in Q3 (Micron's FQ4) on sold-out volume = margin expansion = confident higher guide. TrendForce reversing its Q3-relief call the day before print is a strong confirming tell. Slightly below spot because it confirms (lags) rather than leads, and the exact server-DDR5 % is paywalled.
reading: Q1'26 conventional DRAM contract +93-98% QoQ (record; drove industry rev +81% QoQ to $97B). Q2'26 +58-63%. June 22 TrendForce: consumer DRAM shortage cascading to DDR2, contract prices to CONTINUE RISING in 3Q26 (DDR2 +35-40% QoQ); TrendForce WALKED BACK its earlier 'relief in Q3' thesis -- a fresh bullish tell. Could NOT source exact Q3 server-DDR5 contract % (paywalled DataTrack) -- FLAGGED.
Why it predicts the guide
The Q3'26 (FQ4) contract round being negotiated now sets the ASPs the guide rests on.
Where to get it
TrendForce press releases (Jun 1 +81% 1Q industry rev; Mar 31 Q2 forecast; Jun 22 3Q26 continue-rising); DataTrack server-DDR5 settlement (paywalled).
Merged from
fresh#2
Accessible before print
Yes (TrendForce June 22 note is 2 days old)
12 Peer supply-discipline / forward bit-supply scoreboard: big-3 + CXMT wafer-start & capex calendar vs hyperscaler-implied bit-demand; watch the FIRST capacity-add that breaks 'discipline' Supply-pricing disconfirming pre-print ✓ 24 src
impact: The master balance that sets cycle DURATION -- exactly what management implicitly forecasts when it frames 'sold out into 2027.' Continued discipline underwrites a confident raise; any capex re-acceleration is the disconfirmer. WFE order flow leads capacity online by 12-18 months and confirms/breaks the 'no new supply' premise. ACH-premium disconfirmer, fully pre-print.
reading: Samsung refused 2026 annual guidance, keeps DRAM-first / no large NAND expansion. ASML EUV reportedly sold out through year-end. Micron flagged 'upside pressure' to capital-investment guidance at a competitor conference. To convert into incremental industry bits vs demand.
Why it predicts the guide
Cycle duration = the fulcrum of how durable the guided pricing is; discipline vs re-acceleration is the structural swing.
Where to get it
Big-3 + CXMT capex/wafer-start guidance; SK Hynix earnings; Samsung capex commentary; ASML/AMAT/LRCX/TEL memory book-to-bill + lead times; BESI/ASMPT hybrid-bonding orders; fab-online calendar.
Merged from
corpus B4, corpus B12 (WFE)
Accessible before print
Yes (capex guidance + WFE book-to-bill)
13 Micron inventory days (DIO/DSO), finished-goods-vs-WIP mix, customer prepayments / deferred-revenue / advance-payment & LTA-liability balances QoQ; bit-ship-minus-consumption gap Company-specific disconfirming partial 16 src
impact: The balance-sheet tell of mid- vs late-cycle that disciplines the guide. RISING customer prepayments/deferred revenue = hard proof of genuine forward lock (supports a confident raise). RISING finished-goods/DIO while management says 'sold out' = channel pull-forward/double-ordering = the contradiction that makes the guide fragile. ACH-premium disconfirmer; the trend is pre-print, the level lands with the print.
reading: To pull from prior 10-Q for the trend; FQ3 10-Q lands with the print. Rising prepayments = hard forward lock; rising finished-goods/DIO while 'sold out' = pull-forward/double-ordering contradiction.
Why it predicts the guide
Prepayments prove demand is paid-for forward; inventory bloat under a 'sold out' claim would expose double-ordering that reverses the guide.
Where to get it
Micron prior 10-Q balance sheet + MD&A (inventory days, DSO, deferred revenue, advance payments) on EDGAR now; FQ3 10-Q at the print.
Merged from
corpus C2
Accessible before print
Partial (prior-quarter 10-Q on EDGAR now for trend; FQ3 10-Q drops with the print)
14 Prior beat-and-raise pattern + the gap between Micron's own prior guide and consensus -- defines the bar (FQ3 guided $33.5B/81%/$19.15; Street FQ4 model ~$43B/$25.4) Positioning confirming pre-print ✓ 12 src
impact: Anchors the bar the 6/24 guide must clear. The market is positioned for another beat-and-raise; a guide merely in-line with the steep ~$43B ramp can disappoint despite a huge reported quarter (the beat-and-drop setup). Predicts the REACTION function and defines beat/in-line/miss thresholds more than the guide itself. Fully pre-print.
reading: Micron has beaten EPS each of last 4 quarters (~21.7% avg surprise). FQ3 was itself a record raise that blew past prior $24.29B consensus. Street FQ4 ~$43B rev (19 analysts) / ~$25.4 EPS (17 analysts) -- a steep sequential step up. FQ3 EPS consensus +3.1%/30d, +68%/90d (estimates chasing price). Rosenblatt explicitly expects a beat-and-raise.
Why it predicts the guide
You cannot judge the guide 'beat or miss' without the elevated bar it is measured against.
Where to get it
Micron FQ2 press release + FQ3 guidance range (public); FactSet/Bloomberg/FMP FQ4 consensus distribution + revision velocity; Rosenblatt note.
Merged from
corpus C4, corpus claim#2, fresh#12 (revisions)
Accessible before print
Yes (guidance + consensus public)
15 Channel/OEM price pass-through & lead-time extensions (>30 weeks; ~5-10% server price hikes Apr-Sep 2026; Winbond exiting DDR2); demand-elasticity tripwire (OEM device hikes, quiet base-RAM spec cuts, KV-cache-to-NAND) Demand disconfirming pre-print ✓ 12 src
impact: Sticky pass-through (lead times blowing out, OEMs already raising end-product prices) = price increases are STICKING through the channel = durable ASPs = confident multi-quarter guide. The flip side -- quiet base-RAM spec cuts / memory-efficient architectures / Qualcomm build cuts -- is the leading edge of demand DESTRUCTION that would make the FQ4 demand assumptions fragile. Dual-natured (confirming near-term, disconfirming at the margin), fully pre-print.
reading: Lead times >30 weeks across segments (POs now arrive Q4'26+). OEMs raising prices, cutting low-cost SKUs; ~5-10% server price hikes Apr-Sep 2026; smartphone/notebook brands raising prices & downgrading specs. Winbond exiting DDR2. Tim Cook '100-year flood'. Qualcomm device-build cuts flagged.
Why it predicts the guide
Whether price hikes stick downstream decides if ASPs are durable; spec cuts/efficiency are the first sign of elasticity biting.
Where to get it
TrendForce/trade press (Cinesys, Astute, VersaLogic); Apple/Dell/Samsung 10-Qs + calls (BOM/component-cost language); Qualcomm build-outlook; teardowns; published model-architecture papers.
Merged from
fresh#10, corpus A8 (elasticity)
Accessible before print
Yes (channel checks + OEM calls)
16 eSSD / NAND demand (the second leg): high-cap QLC (122TB/245TB) qual->PO cadence, eSSD lead times, HDD-to-NAND substitution; AND NAND contract vs spot bifurcation Demand leading pre-print ✓ 12 src
impact: The under-watched second leg (~25-30% of revenue). Strong enterprise-SSD contract + HDD substitution lets management guide NAND/storage up independently of HBM -- a corroborating breadth signal. But the weak NAND spot wafer is the one soft spot that could explain a guide landing slightly below the bull case. Moderate impact; fully pre-print.
reading: NAND contract +70-75% QoQ Q2'26 (AI enterprise SSD); NOR +100-120% / SLC +130-150% cumulative 1H26; 2H26 SLC +70-75%, NOR +60-65%. BUT 512Gb TLC spot wafer ~$20.6, flat/weak, 'high prices shrinking volumes' -- a real bifurcation. Mehrotra flags data-center SSDs as a second leg; HDD-to-NAND substitution (Seagate/WD HAMR nearline shortage).
Why it predicts the guide
NAND is an independent leg of the guide; its contract strength vs spot weakness decides whether storage adds to or subtracts from the FQ4 outlook.
Where to get it
Micron QLC qual/PO press + call; Pure Storage/Dell/HPE/Supermicro commentary; Seagate/WD nearline lead-times & book-to-bill; TrendForce Jun 16 NAND/NOR/SLC contract; 512Gb TLC spot wafer.
Merged from
corpus A7, fresh#9
Accessible before print
Yes (TrendForce + OEM commentary)
17 Options skew (25-delta P/C), IV term-structure / implied move, dealer gamma (GEX) walls, large single-name blocks into the 6/24-25 print Positioning contrarian pre-print ✓ 25 src
impact: Positioning/crowdedness gauge that mechanically determines the post-guide reaction. If the ~$43B path is already priced and GEX is call-wall-capped / put-wall-distant, even a strong FQ4 guide can fade (the prior beat-and-drop). Tells you how high the guide bar effectively is and the IV-crush setup. Fully pre-print; contrarian (reaction, not fundamentals).
reading: Implied 1-day move +/-11.03% (+/-$132.22) off ~$1,172; front-weekly IV 155.3% vs 108.5% next monthly (~47pt event premium); implied range ~$1,066-$1,331. 26-Jun weekly P/C volume 0.89 (more calls), but P/C OPEN INTEREST 2.38 (~2.4 puts/call held -- event hedging); longer-dated strongly call-biased. IV crush likely post-print. Flagged $19.4M put / $23M bull block.
Why it predicts the guide
Determines whether a given guide produces an up or down move given how much is priced and how dealers are positioned.
Where to get it
CBOE/OPRA tape (not Twitter screenshots); OCC OI by strike; dealer GEX models; short-interest/cost-to-borrow/days-to-cover; Saxo implied-move note.
Merged from
corpus D2, market_expectations sec.5
Accessible before print
Yes (OPRA/OCC tape)
18 HBM wafer-conversion '~3x trade ratio' + net commodity-DRAM bits REMOVED from market each quarter as the big-3 convert DDR5 wafers to HBM Supply-pricing leading pre-print ✓ 14 src
impact: Mechanistic core of the tightening thesis: quantifies whether the 'shortage' management guides into is real physics or narrative, and raises the capex-just-to-stand-still the guide must fund. Leading and explanatory; fully pre-print but harder to source cleanly, so mid-table.
reading: To verify the ~3x figure via die-size/yield teardowns. Every HBM wafer converted starves commodity DRAM, tightening DDR5 price -- the engine behind a price-led FQ4 raise.
Why it predicts the guide
It is the physical mechanism that makes the commodity-DRAM price leg of the guide self-reinforcing as HBM demand grows.
Where to get it
Die-size/yield teardowns (TechInsights/SemiAnalysis HBM die analyses) to verify 3x; big-3 wafer-start mix disclosures; WFE memory-segment mix.
Merged from
corpus B7
Accessible before print
Yes (teardowns + wafer-mix disclosures)
19 Datacenter buildout pace as upstream demand pull: BTM-gas funnel (101->57->7 GW), SemiAnalysis H100/B200 1-yr CONTRACT rental DIRECTION (rose ~40% Oct'25->Mar'26), memory ~30% of AI spend / ~70% of 2026 DRAM to AI DCs Demand leading pre-print ✓ 5 src
impact: Compute landing sooner pulls HBM/DRAM demand forward into the FQ4 window; the rental-curve direction is a leading tell on whether demand is outrunning the buildout. Sizes the FQ4 demand denominator directly. Leading but lower source-count and one more hop removed, so lower-mid.
reading: To pull. GPU contract-rental curve rose ~40% Oct'25->Mar'26 (demand outrunning buildout = bullish); a roll-over would be the first crack. Memory now ~30% of hyperscaler AI spend; ~70% of 2026 global DRAM to AI DCs.
Why it predicts the guide
The pace of datacenter power/compute coming online gates how much memory demand actually lands in the FQ4 window.
Where to get it
SemiAnalysis GPU cloud rental index; RBC/SemiAnalysis BTM trackers; turbine-OEM order PRs; SemiAnalysis memory model; consumer DDR5 spot (PCPartPicker) as ungameable spillover read.
Merged from
corpus A5, corpus A3 (attach)
Accessible before print
Yes (SemiAnalysis trackers + spot DDR5 spillover)
20 HBM3E ~20% 2026 price hike + HBM4 ~20-30% premium; BUT HBM-to-DDR5 premium COMPRESSING (4-5x -> 1-2x by end-2026) as commodity DDR5 surges Supply-pricing disconfirming pre-print ✓ 6 src
impact: HBM pricing is the high-margin engine of the guide; the ~20% hike + HBM4 premium support the margin guide. The subtle bear nuance -- as commodity DDR5 surges, the HBM premium OVER DDR5 compresses, diluting HBM's RELATIVE margin advantage even as absolute ASPs rise -- could temper an otherwise-bullish margin guide. Niche disconfirmer; fully pre-print.
reading: HBM3E +~20% for 2026 (Samsung & SK Hynix, Dec 2025). HBM4 ~20% premium at launch, some reports ~30% over 12-Hi HBM3E (~$500/unit). HBM4 ~55% of 2026 HBM revenue mix. HBM-vs-server-DDR5 premium compressing from 4-5x toward 1-2x by end-2026 -- a relative-margin headwind nuance.
Why it predicts the guide
Relative HBM margin advantage feeds the GM mix; its compression is a quiet headwind to the structural-margin story.
Where to get it
TrendForce Dec 24 2025 (HBM3E +20%); HBM4 premium reporting (~$500/unit); HBM/DDR5 gap-compression commentary.
Merged from
fresh#11
Accessible before print
Yes (TrendForce HBM pricing)
21 Advanced-packaging bottleneck: TSMC CoWoS monthly wpm + Micron Singapore HBM-packaging ramp; TSV/hybrid-bonding tool orders (Hanmi/BESI/ASMPT); CoWoS-L yield on Rubin Supply-pricing leading pre-print ✓ 5 src
impact: A packaging ceiling caps the HBM line in the guide even with demand sold out -- so it's a leading tell on whether 'sold out' converts to guided REVENUE or stays supply-gated. Lower source-count and second-order, so lower-mid.
reading: To pull. HBM is gated by packaging, not just wafers. CoWoS adds and hybrid-bonding orders lead how fast HBM bits can SHIP into FQ4.
Why it predicts the guide
Determines the deliverable HBM bit-supply that backs the revenue the guide can promise.
Where to get it
TSMC CoWoS capacity commentary & NVIDIA share; Hanmi/BESI/ASMPT/Disco order books & lead times; Micron Singapore packaging milestones.
Merged from
corpus B8
Accessible before print
Yes (TSMC commentary + tool order books)
22 Capex guidance trajectory (FY26 +$5B to >$25B, rising into 2027; Idaho/Taiwan fabs; 'upside pressure' flagged) Company-specific confirming pre-print ✓ 8 src
impact: A capex raise alongside the revenue guide is a tell on management's own confidence in durable demand (you don't add ~$200B of capacity for a cyclical blip). But a steep capex step also pressures FQ4 FCF and is the 'capex absorbs the record earnings' cyclical-peak signature. Confirming with a contrarian edge; fully pre-print.
reading: Micron raised FY26 capex by $5B to >$25B and signaled further rises into 2027; flagged 'upside pressure' to capital-investment guidance at a competitor conference. Updated number lands on the FQ3 call.
Why it predicts the guide
Management's own capital commitment reveals its conviction in the demand the guide projects -- and the FCF pressure that comes with it.
Where to get it
Micron FQ2 capex guidance (public); FQ3 call for the updated number; CHIPS/state-subsidy offset terms.
Merged from
corpus C5
Accessible before print
Yes (FQ2 number public; FQ3 update on call)
23 SOCAMM2 / LPDDR5X module content per NVIDIA Vera/Rubin rack (was per-rack content CUT 192->96GB?) + Micron's SOCAMM allocation share Demand leading pre-print ✓ 2 src
impact: A specific, non-obvious bit-demand swing for the non-HBM DRAM line, plus a distinct Micron design-win leg. Leading and pre-print, but only 2 sources and a narrow line item, so lower-mid.
reading: To pull. A permanent per-rack LPDDR5X cut would trim FQ4 non-HBM DRAM demand; a 'board-first, redistribute across 2x racks' read is neutral-to-positive. Micron's SOCAMM share is a distinct design-win leg beyond HBM.
Why it predicts the guide
Per-rack LPDDR5X content x rack volume is a measurable input to the non-HBM DRAM portion of the guide.
Where to get it
NVIDIA Vera/Rubin platform spec & module BOM; Micron/Samsung/Hynix SOCAMM qual; hyperscaler memory-upgrade PO cadence.
Merged from
corpus A6
Accessible before print
Yes (NVIDIA platform spec + qual)
24 13F institutional flow + forced thematic-ETF flow (Tepper/Appaloosa, Citadel hedged-vs-directional; Roundhill DRAM ETF ~26% MU weight; days-to-cover); the ~90%-bull crowding + circular-convergence audit Positioning contrarian pre-print ✓ 17 src
impact: Smart-money/flow-fragility gauge plus the circular-convergence audit that identifies which single line in the guide actually moves the stock. Useful context but 13F is lagged 45 days and it tips the reaction, not the guide; lower.
reading: To pull. Marquee holders trimming into the print, or heavy passive/thematic crowding, configures a violent post-guide unwind. The crowd's 'convergence' is largely circular (one Goldman/MS/Aletheia note + TrendForce prints echoed); the real disagreement is on ONE variable (peak ASP / HBM share / exit P/E).
Why it predicts the guide
Crowding/flow shape the post-guide unwind, and the convergence audit isolates the one variable the guide must move.
Where to get it
SEC 13F (Appaloosa, Citadel) QoQ deltas (option vs share lines); Roundhill DRAM ETF creation/redemption & AUM; FINRA short-interest/utilization/borrow fee; trace each 'corroborating' figure to its ONE originating note.
Merged from
corpus D4, corpus D3
Accessible before print
Yes (13F filings + FINRA short interest, but 13F is lagged 45 days)
25 March-2026 analog: the 'biggest beat in 2 years' that fell ~30% over ~8 sessions -- exact path, IV-crush timing, cross-name sympathetic-selloff map Positioning contrarian pre-print ✓ 4 src
impact: The base-rate for how THIS stock trades a strong guide when positioning is stretched -- calibrates 'how good must the guide be' and the fade risk. Pure event-mechanics context (doesn't predict the guide itself), so lowest, but a valuable sanity check on the reaction.
reading: To pull the exact path. The prior beat-and-drop is the cleanest evidence that a good FQ4 guide is necessary-but-not-sufficient for the stock.
Why it predicts the guide
It calibrates the reaction function: it shows that even a strong guide can fade from stretched positioning.
Where to get it
Price/IV history around the prior print; cross-name beta (Hynix/Samsung/SanDisk/WDC/DRAM-ETF) on event days; days-to-cover.
Merged from
corpus D5
Accessible before print
Yes (price/IV history)

What would tip the call — the three surfaced panels

★ #1 call-watch — explicit "CY2027 sold out" language
The single most important, genuinely-new datum that prints 6/24.
As of 2026-06-23 NO explicit 'CY2027 sold out' statement exists -- only 'supply tight well beyond 2026' + customers moving to 3-5yr supply agreements.
BEATAn explicit CY2027-contracted/sold-out quote on the call = the clearest BEAT signal.
MISSHedging or 'monitoring 2027 supply additions' = the clearest MISS signal.
CORRECTION — "sold out" was MEDIA framing, not Micron's word. It appears zero times in the FQ2 (Mar 18) call transcript; management actually said "tight beyond calendar 2026" + signed its first multi-year SCAs (incl. a 5-year deal). That *resets the bar* for tonight: an explicit "CY2027 sold out / pricing locked" = a genuine UPGRADE (new information); a mere repeat of "tight beyond 2026" = just a HOLD (no new info, already priced).
NEW yellow flag — is the GM step-up HBM-mix-driven (durable) or commodity-price-driven (fragile)? Micron's CEO said on the FQ2 call (Mar 18) that *"the margins for non-HBM today are HIGHER than HBM margins"* — the HBM/DDR5 premium compressed from ~4–5x toward ~1–2x. If the FQ4 GM step-up is being carried by commodity DRAM pricing (the tier most exposed to a 2027 China-led reset + PC −11% / phones −14% unit contraction) rather than HBM mix, the margin durability the bull case assumes is weaker than it looks. If management leans on HBM as the margin story while commodity is actually carrying it, that's the durability flag.
Supporting bull context on record: HBM TAM raised to ~$100B by 2028 (from ~$35B 2025, ~40% CAGR, pulled forward ~2yr); HBM now >30% of total DRAM revenue (vs <5% in 2022); FQ2 Cloud Memory BU (incl. HBM) ran ~$7.7B, +163% YoY (cleanest public HBM proxy — no clean HBM-only line disclosed).
Panel A — the "what would tip the call" trio
The three reads with the most leverage to flip beat ↔ miss.
LTA fixed-price take-or-pay ranked #3
BEATConfirmed fixed-PRICE take-or-pay in 8-K/10-Q => FQ4 revenue largely pre-booked => management guides unusually high with low risk.
MISSFixed-VOLUME at re-setting prices => the guide is still a spot-cycle bet and 'cycle structurally broken' framing is wrong.
Spot DRAM still widening vs rolling over ranked #2
BEATSpot-over-contract premium still widening straddling 6/24 => next contract settles higher => management guides FQ4 pricing up with confidence.
MISSSpot rolling over while contract catches up => first crack of a cycle top => guide-disappointment risk even on a strong reported quarter.
HBM4 confirmed in Rubin volume + share ranked #8
BEATConfirmed HBM4 volume slot at NVIDIA Rubin/VR200 with a real share => management raises the HBM revenue trajectory and frames the share gain structural.
MISSMulti-source qual handing NVIDIA pricing leverage, or Micron stuck at token share => HBM-margin risk that tempers the GM guide.
Panel B — tomorrow-AM fresh-pull list
Not out as of 6/23 — re-check the morning of 2026-06-24 before the print.
  • June Korea 20-day export FLASH (June 1-20 chips YoY) -- not out as of 6/23; lands ~Jun 21-22 KST, check AM of 6/24 (clue rank 5).
  • Next TrendForce weekly spot update (June 17/24) -- June 10 is the latest indexed; check whether DDR4/DDR5 spot is still climbing AM of 6/24 (clue rank 2).
  • Latest TrendForce contract note on Q3'26 (=FQ4) server-DDR5 settlement % -- exact figure was paywalled (DataTrack) as of 6/23 (clue rank 11).
  • Any 6/23-6/24 Micron 8-K LTA exhibit or insider Form 4 hitting EDGAR (clues rank 3 / rank 10).

2028–2030 supply-wall timeline — the cycle-turn risk is 2028+, NOT a 2026–2027 guidance risk

The cycle-turn risk is a 2028+ story — NOT a 2026–2027 guidance risk. Through 2027 nearly every credible model still shows DRAM in *deficit* (supply ~+16% vs demand +20–35%; Goldman even DEEPENED its 2027 shortage to −5.9%). The genuinely large *greenfield* wall lands 2028–2030. So the FQ4-guide horizon rests on the committed / sold-out tranche; the oversupply risk is a *timing-and-multiple* trade, not a near-term guide risk.
2026–2027 — in-flight tranche committed / sold-out
Samsung P4 + 1c ramp + NAND→DRAM conversions; SK Hynix M15X; Micron Idaho ID1 first wafer mid-CY2027. Largely committed; HBM cannibalization is *tightening* commodity bits here → supports a confident, runway-extending guide.
2028–2030 — the real greenfield wall cycle-turn risk
SK Hynix Yongin 360k wpm (full ramp ~H1 2030); Samsung P5 Fab1/Fab2 (~H2 2028); Micron ID2 (beyond 2028, before NY) + NY megafab. This is the supply wave the long-dated $1,625 PTs may be mispricing.
Micron's own disclosed dates (FQ2 FY26 call)
"New fabs not delivering meaningful output until fiscal 2028." ID1 first wafer mid-CY2027 · ID2 production beyond 2028 (before NY) · Tongluo/Taiwan meaningful shipments fiscal 2028 · Hiroshima HBM ~2028 · Singapore HBM packaging meaningful CY2027. FY26 capex >$25B; FY27 construction capex +$10B+ YoY.
Two factors that SUPPRESS effective commodity supply
HBM ~3:1 wafer cannibalization — HBM consumes ~3× the wafer area per bit vs commodity DRAM, so every HBM wafer *starves* commodity DRAM (by 2027 ~30% of wafers → ~13% of bits). Node migration now yields only ~10–15% density gain (diminishing returns) — process shrinks add far fewer bits than in prior cycles. Both cut *against* the simple 'oversupply wall' thesis.
⚠ The long-end swing variable — NY Fab1 timing CONFLICT
Micron says ~2029 "on track"; TrendForce (Nov-2025) reports a 2–3yr delay to late-2030. Track *both* — this dated variable most moves the 2028–2030 oversupply picture. Lead times make upside surprises slow: EUV tool 12–24mo, cleanroom-to-volume ~1.5–2yr, greenfield groundbreaking-to-first-wafer 3–5yr (Micron ID1 = Sep-2022 announce → mid-2027 output).

The bear case — valuation + margin inversion (the adversarial counterweight to the $1,625 PT)

The bull case (~$1,625 PT) is mechanically a bet that FY2027 *peak* earnings are durable and deserve a non-trough multiple. The adversarial read: the bear is not a 2027 supply glut — it's a timing-and-multiple trade. At ~$1.22T on peak book, there is essentially no margin of safety.
① The peak trap — valuation
MU trades at ~17–22× book (equity ~$54.2B vs ~$1.22T cap) on *peak* earnings — an all-time extreme. Memory has historically been *bought near book* at ~1.3× P/B (FY20/FY22 troughs). MU's ~10–11× forward P/E is the *lowest* in the SOX — a contrarian SELL (memory looks 'cheapest' right before peak EPS collapses). 'This time is different' is the verbatim 2018/2022 peak signal.
② The margin inversion (the key insight)
Micron's CEO Mehrotra (FQ2 call, Mar 18): "the margins for non-HBM today are HIGHER than HBM margins." HBM3E's premium over server-DDR5 compressed from ~4–5× toward ~1–2× — *because DDR5 is RISING, not HBM falling* (DDR5 64GB RDIMM per-wafer revenue overtook HBM in Q1'26). So the GM step-up is being carried by commodity DRAM — the tier most exposed to a 2027 China-led reset. The bull premise that HBM is a permanently premium-margin product is *already false at the margin*.
③ Demand cracks
Lenovo CFO (Nov 2025): component inventory ~50% above normal (enough for all of 2026) → double-ordering risk. End-markets in record contraction: PC −11.3% units, smartphones −13.9% (2026, IDC/Counterpoint) — the commodity tier is a *victim* of memory inflation. Capex exceeds operating cash flow ~Q3'26 (Epoch) — hyperscaler FCF goes negative first time in ~35yrs; if MSFT/META/AMZN signal AI-server slowing into H2-2026, HBM-under-construction swings to commodity → *double glut*.
⊕ The two genuine bull rebuttals (still standing)
1. Channel inventory is at cycle *lows* — not a pre-glut buffer. 2. HBM is contractually *sold out* through 2026. These are the facts that genuinely cut *against* the bear and are why the trade is *timing-and-multiple*, not an imminent fundamental rollover. The bear doesn't need a 2027 glut — it needs the *market to stop believing* in perpetual peak earnings, which historically happens 1–2 quarters BEFORE the rollover.