Track record
Every issue ends with three watch items. Scored in public, 7–30 days later. Misses stay here.
Every issue ends with 3 Things to Watch. Each is scored HIT / MISS / PENDING 7–30 days later, in public. Misses stay on this page.
| Date | Watch item | Verdict by | Status | Result |
| 2026-11-03 | Tonight's triple overlap — midterm results x Iran's 48-hour deadline x Brent's second settle: if Brent settles below $108 again while the elections produce no restrictive DC-legislation wave (most local ballot measures failing or stalling), AI duration's macro and policy headwinds unlock together (December-hike repricing + contained DC risk) and the duration complex leads the next leg; if talks collapse into the election (strikes escalating, futures back above $110) AND anti-DC election results resonate, oil and power-politics risk tighten simultaneously. | 2026-11-10 | PENDING | Pending. |
| 2026-11-03 | AMD's DC guidance tonight is the first read on whether SOX 10,328 is the bottom: after -17.5% in two sessions the bar is low — strong data-center growth guidance validates the 'Intel DC +78% mirror' (AI demand intact, price/mix pressure is the story) and the new low reads as oversold; PC softness or memory-cost squeeze showing up in guidance confirms the demand-top narrative and points SOX to 10,000. | 2026-11-10 | PENDING | Pending. |
| 2026-11-03 | The memory mispricing adjudication: Apple's filing-grade confirmation (GM guide 45-46% on memory costs, inventory +154% pre-buying, 'costs rise beyond December') printed against new lows across MU/SNDK/WDC/STX — if MU/SKH stabilize this week and November Korean exports hold the record run, this week's lows get labeled 'profitability doubts mispriced as demand doubts' and re-rate into MU's December call; if November exports roll over MoM or MU dodges 2027 LTA language, the deceleration narrative wins and memory makes another leg down. | 2026-12-01 | PENDING | Pending. |
| 2026-11-02 | Tonight's SNDK/WDC session x AAPL's gross-margin language is the twin adjudication of the memory bill: if SNDK holds Friday's low ($1,150.89) while AAPL absorbs memory costs and holds margin guidance, the 'price top' converges into a Kioxia-specific event and Friday's hardware selloff reads as overdiscounted; if SNDK rolls to a second leg while AAPL guides margin down on memory BOM, demand destruction extends to all endpoints and the deceleration narrative takes over the memory complex into MU's December call. | 2026-11-16 | PENDING | Pending — verdict by 2026-11-16. Interim (11/03): Both negative legs hit: SNDK broke Friday's low ($1,150.89) to close -6.59% $1,075.08, and AAPL guided December GM down to 45-46% with memory costs named as the main pressure — the 'demand destruction extends across endpoints' branch is live (Tuesday Asia: KOSPI -3.4%, SKH -5.2%), but AAPL's +154% inventory build adds the unresolved twist. |
| 2026-11-02 | Doha's 48-hour window x the 11/3 midterms is oil's optionality peak: if Brent's $110-118 platform breaks for real (consecutive settles below ~$108) on an accepted US-Iran reopening, half of AI duration's macro headwind unlocks via a December-hike repricing; if the window fails into the election with the post-election strike preparations (CNBC 10/28) still live, Brent's round trip from $110 will be faster than the fall from $117 — supply-disruption premia reprice non-linearly. | 2026-11-09 | PENDING | Pending — verdict by 2026-11-09. Interim (11/03): First leg done: a draft US-Iran deal circulated Monday and Brent settled ~$104-105 (-8%, biggest drop since July) — the first settle below the ~$108 threshold; Tuesday morning held flat (~$104.8) despite post-window US 'defensive' strikes on Qeshm Island (WSJ/CNN/Guardian) — futures are voting for the deal; Iran's two-day window expires on election night. |
| 2026-11-02 | November Korean exports + MU's December call adjudicate the volume/price fork: if semiconductor exports stay at record runs MoM while MU's LTA language extends the annuity frame (sold-out/2027 locks), 2027 memory-earnings consensus built on price extrapolation is systematically low and October's -25% was a multiple event; export momentum fading MoM or soft LTA wording hands the pricing-top narrative its volume leg too. | 2026-12-01 | PENDING | Pending. |
| 2026-11-01 | AAPL's 11/2 print is the memory shortage's downstream bill: if Apple guides gross margin down on memory BOM (or quantifies the squeeze), demand destruction officially extends from PCs (Intel/Dell/HPQ already paid) to all consumer devices and the 'shortage is good for memory, bad for everyone else' trade gets its receipt; if Apple absorbs the cost and holds margin guidance, the shortage's downstream bill has a payer and October's hardware selloff reads as overdiscounted. | 2026-11-16 | PENDING | Pending — verdict by 2026-11-16. Interim (11/03): Resolved into the receipt branch: Apple guided December GM to 45-46% (from 47.9%, prior 50.1%), CFO Parekh named memory costs as the main pressure, Ternus said costs keep rising 'in the December quarter and beyond,' inventory +154% to $14.5B — the bill has a payer, and the payer is pre-buying more. |
| 2026-11-01 | Kioxia's +0.8% QoQ Q3 guidance is sandbagging convention, not a demand reading: if Monday's PTS slide (~¥12,500) is absorbed and SNDK/WDC stabilize, the 'NAND pricing top' narrative was overpriced against an 80.4% margin, sold-out books and a first dividend — a failed-breakdown setup for the memory complex into MU's December call; if Kioxia keeps falling and SNDK follows, pricing deceleration enters the main narrative and DRAM sentiment catches it. | 2026-11-15 | PENDING | Pending — verdict by 2026-11-15. Interim (11/03): The negative branch extended: Kioxia closed Monday -15.78% at ¥11,345 (at session lows, worse than the intraday range), SNDK -6.59% to a new low, WDC -7.52%, MU -5.78% — pricing deceleration entered the main narrative and is spreading to DRAM sentiment (SKH Seoul -5.2% Tuesday); Nomura's 'fundamental mispricing' call and KB's 'W-bottom' mark the sell-side defense forming at the lows. |
| 2026-11-01 | Supplier-underwritten credit becomes AI's marginal funding system: if Anthropic prices mid-November into Blackstone's junk-grade reality (roadshow unlaunched, window 11/16-18) while AVGO's $63B 'Project Northstar' vehicle and NVDA's backstop talks proceed, guarantor concentration (one Hock Tan, one Jensen) becomes a priced systemic variable and neocloud spreads stay wide despite Bessent's reassurances; an IG-grade print for Anthropic post-IPO, or IG-caliber pricing on the $63B vehicle, falsifies the tiering. | 2026-11-30 | PENDING | Pending — verdict by 2026-11-30. Interim (11/03): Tiering evidence compounded Monday: ORCL filed a $20B stock-plus-bond raise for AI (first mega-cap hyperscaler common-stock sale of the cycle, stock -8.76%), CRWV printed an all-time low ($50.15, -44%/month) as CPP 'gets selective' on AI debt, and Anthropic's amended S-1 lifted commitments to $895.4B — the guarantor-concentration variable stays priced. |
| 2026-10-30 | Samsung's January capital-return plan resolves Korea's two-tier structure: if Samsung announces a multi-year return program at meaningful scale relative to SK hynix's ₩100T (e.g., a combined dividend+buyback-and-cancel framework above ₩50T/year or an explicit cancellation schedule for the ₩35T buyback), the Samsung-vs-SKH valuation discount converges and the 'record profits, zero returns' discount framework breaks; a token or deferred-again plan confirms the two-tier inside Korea, and Samsung keeps trading at a return-policy discount despite shortage-through-2028 fundamentals. | 2027-01-20 | PENDING | — |
| 2026-10-30 | Friday's US session is round two of 'tiering vs AI-demand-top': if Amazon's after-hours loss compresses at the open (down less than ~3%), GOOGL holds above $280 and SOX holds 11,400, the tiering rule survives the GOOGL shock and hyperscaler dispersion remains the dominant trade; if AMZN opens down >6%, GOOGL breaks $270 and SOX loses 11,200 with memory giving back Wednesday entirely, the market is pricing an AI-demand top and October's credit-shock low (SOX 10,732) re-enters the range. | 2026-11-06 | PENDING | Interim (10/31, resolution data in): SOX -7.20% to 10,688.8 — broke BOTH the 11,400 ruler and the October low of 10,732; GOOGL -1.40% $279.84 (held the ~$280 line); AMZN -1.41% close (the -6% AH compressed); INTC -28.58% (worst since 1974) was the trigger, not GOOGL. Reading: the 'tiering' verdict held at the platform level (GOOGL/AMZN ~-1.4%) and failed at the semiconductor level — the 'demand-top' narrative now owns the semi tape. November's first sessions decide whether 10,689 was a positioning-flush low or a trend start; formal close 11/6. Interim (11/2): Monday's first prints lean repair — US futures ES +0.4%/NQ +0.8%, KOSPI +1.63%, SKH +2.54% — but Kioxia's -13% extension keeps the demand-top branch alive inside NAND; final reading on the 11/6 verdict date. |
| 2026-10-30 | Kioxia's FQ2 print (~14:00 CST 10/30, post-window) rules on the intra-NAND bifurcation: if Kioxia beats on enterprise/data-center NAND with guidance for further enterprise price gains while TrendForce's consumer spot weakness (512Gb TLC -9.19%) persists, the 'enterprise strong / consumer weak' split is confirmed and NAND pricing power decouples from consumer spot; if Kioxia's guidance leans on consumer recovery or enterprise pricing disappoints, the DRAM/NAND divergence thesis weakens and the SKH-overhang discount on Kioxia deepens. | 2026-11-06 | PENDING | Interim (10/31, print landed in-window Friday 14:00 CST): FQ2 revenue ¥2.48T (+40.3% QoQ, +453.2% YoY), non-GAAP OP ¥1.9883T (80.2% margin; IFRS 80.4%) — single-quarter OP more than double the prior FULL fiscal year, beat guidance across the board; first-ever ¥90 interim dividend (Dec 8). But Q3 cumulative revenue guidance ¥6.7472T implies just +0.8% QoQ with OP below consensus — the first official 'price-hike deceleration' print from a major memory maker. Stock -7.14% ¥13,470, PTS overnight ~-7% to ~¥12,500; SNDK -13.47%, WDC -7.41% sympathy. The bifurcation tilts toward 'pricing-slope topping' — final reading Monday's PTS/SNDK reaction, formal close 11/6. Interim (11/2): the deceleration branch extended — Kioxia -11~-15% Monday (lowest since May) with NAND ASP +mid-20% QoQ vs +70% prior now the accepted reading; the enterprise/consumer split stays unresolved inside the JV itself (Kioxia +0.8% vs JV partner SNDK +12.4% guides). |
| 2026-10-29 | The capex-verdict rule written this week (reward capex discipline + cloud acceleration; punish capex escalation + revenue deceleration) gets its final test at Amazon tonight: if AWS prints at/above the ~41% street growth with capex language that is flat-to-down and the stock works while GOOGL's -10% open doesn't spread beyond its own tape, the tiering rule is confirmed and hyperscaler dispersion becomes the dominant trade into year-end; if Amazon beats on AWS but still sells off on capex, or GOOGL's decline drags SOX/memory down despite SKH's sold-out-through-2029 fundamentals, the market is pricing an AI-demand top, not tiering. | 2026-11-06 | INTERIM | Interim (10/30-31): Friday's session stress-tested the rule. Amazon's -6% after-hours print COMPRESSED to a -1.41% close ($240.35) — the market ultimately rewarded AWS's actual acceleration over the group's -$42.4B FCF (the least-punished of the week's three prints), while GOOGL stabilized at -1.40% ($279.84). But the complex-wide 'AI cash burn' selloff (Reuters) took everything down in between: SOX -7.2% broke its October low. The rule survived at the platform level and failed at the semiconductor level — final scoring at the 11/6 window with Monday's AAPL print as the capstone. Interim (10/30): the print is in — AWS $46.0B +39.3% (above the $45.7B dollar consensus though below the ~41% growth street; fastest in nearly five years per GeekWire), record 13.9% operating margin, but EPS missed, FCF -$42.4B, 2026 capex held at ~$220B, light Q4 guide: -6% after hours. So neither clean branch: AWS beat in dollars AND the stock sold off on capex/FCF. GOOGL's -12.64% close DID spread beyond its own tape — but asymmetrically: SOX held 11,518 (>11,400) while memory gave back ~half (MU -4.98%) and Friday Asia broke hard (KOSPI -5.13%, SKH -8.33%). Final verdict on Friday's US open: tiering survives if AMZN's loss compresses and SOX holds 11,400. |
| 2026-10-29 | SK hynix's 'sold out through 2029 / constrained through 2030' becomes the industry anchor only if Micron's December call corroborates it: if MU gives a matching multi-year LTA-coverage framing (and its 2027 sold-out ratio rises from ~75%), the memory sector re-rates on contract-annuity math and the YE27 P/E debate resolves higher; if MU guides more cautiously (capacity additions, China overhang), SKH's language gets read as single-company positioning and the re-rate stalls. | 2026-12-20 | PENDING | — |
| 2026-10-29 | Alphabet's -10% after-hours print is a positioning stress test for the AI complex: if the decline compresses at Thursday's open and SOX/memory hold their SKH-driven gains (SOX above ~11,400), the 'AI special-equity' rebuke remains name-specific and the two-tier framework holds; if GOOGL opens near -10% and takes SOX below ~11,200 with memory giving back its Wednesday gains, the rebuke generalizes and the October credit-shock low (SOX 10,732) comes back into play. | 2026-11-05 | INTERIM | Interim (10/31): Friday's close -1.40% at $279.84 — the shock did NOT compound at the platform level (no second-day capitulation; Reddit's -28% on Google-AI traffic interception even reads as the flip side of GOOGL's moat). Transmission instead surfaced in semis/hardware: SOX -7.20% to 10,688.8 (below 11,400 AND the 10,732 October low), AMD -10.9%, TSM -5.95%, MRVL -10%. The 'asymmetric transmission' branch (platform punished, semis hold) inverted into 'platform holds, semis break' — final verdict 11/5. Interim (10/30): GOOGL closed -12.64% at $283.82 (~$500B, biggest single-day value loss in its history) — the AH -10% deepened rather than compressed. SOX closed 11,518 (held above 11,400; the 'name-specific' leg technically survived in US semis), but memory gave back roughly half of Wednesday's gains (MU -4.98% to $934.27, still above Tuesday's close; SNDK -5.32%) and the transmission went full-force through Asia on Friday: KOSPI -5.13% (sell-side sidecar), SK hynix -8.33%, Samsung -5.82%, Kioxia -8.31%, Nikkei -2.54%. Reading: US tiered, Asia indiscriminate. Final verdict 11/5 on whether Friday's US session holds the line. |
| 2026-10-26 | SpaceX's $40B terms as the credit-tiering ruler: if the deal prints at investment-grade-caliber spreads while CoreWeave CDS holds above 10%, the two-tier AI credit market is structural and 2027 GPU capacity consolidates around IG-backed balance sheets (SpaceX/MSFT-tier), with pure-plays financing at 15%+ or clearing out; if SpaceX's terms reveal high-yield economics, or neocloud spreads collapse to single digits after the 10/28 FOMC, October's tiering was a positioning snapshot, not a regime. | 2026-11-20 | PENDING | Interim (10/28): post-FOMC risk-on — NBIS +11.2%/+7.0% and CRWV +7.9%/+7.3% over two sessions (TradingView: 'Anthropic boost + SKH stunner'; Cramer urges caution). CDS prints not yet taken; the ruler verdict stays open to 11/20. |
| 2026-10-26 | SK hynix's 10/28 call resolves the Kioxia question: if the stake sale and ~₩16T special dividend are confirmed, Korea's capital-return template extends beyond buybacks, the 'return vacuum' narrative dies, and Kioxia's overhang becomes a priced variable into its 10/30 print; a denial or deferral keeps the buyback-vacuum framing intact into Samsung's 10/29 results and hands the cycle-top reading (selling NAND exposure at the peak) more weight. | 2026-10-30 | PARTIAL | PARTIAL. The branch resolved as 'not confirmed': SK hynix's 10/28 call did NOT announce the Kioxia stake sale or the ₩16T special dividend (the rumor is back to UNCONFIRMED; BigGo relayed a denial). But the second half of that branch didn't play out either — instead of the 'buyer vacuum' persisting, SKH resolved capital allocation bigger and more conventionally with the ₩100T ($75B) plan, and KOSPI rose 6.63% on the print. The Kioxia overhang itself is NOT resolved: Kioxia -8.31% intraday Friday 10/30 ahead of its FQ2 print (~14:00 CST), and the governance variable (SKH holds the stake; a sale remains an option) carries forward as a standing overhang into 2026. |
| 2026-10-26 | Seagate's 10/27 print closes the HDD profit-quality interrogation: volume-and-price strength with WDC-grade guidance (EPS consensus $7.42) proves the post-WDC selloff was sympathy positioning and the HDD pair re-rates; flat pricing or conservative guidance confirms the 'supply response' leg and the storage/HDD multiple center steps down again — with read-throughs to MU/SNDK/SKHY into the Korean prints. | 2026-11-02 | PARTIAL | Verdict (11/2): PARTIAL — the fundamentals leg delivered (revenue +61% YoY, record $1.5B FCF, guidance above estimates; WDC-grade volume-and-price strength) but the re-rating leg was crushed by sector beta: STX finished Friday $644.58 (-7.98%, below its pre-print level) and WDC $366.11 as Kioxia's NAND deceleration + Intel's BOM warning dragged the whole storage complex down. The HDD multiple center stepped down WITH the complex rather than re-rating — the interrogation was answered 'yes on earnings, no on multiple', and the read-through into MU/SNDK/SKHY stayed negative through month-end. |
| 2026-10-25 | Seagate 10/27 print: if STX shows volume-AND-price strength with WDC-grade guidance (+70% growth ballpark) and the stock holds, WDC's Thursday-Friday drop was sympathy-positioning and the storage/HDD repair resumes; volume-up-price-flat, soft guidance or a supply-response framing confirms the 'profit quality' leg and the sector's valuation midpoint steps down another notch. | 2026-11-01 | PARTIAL | PARTIAL (closed 11/1). The fundamentals delivered: revenue +61% YoY, record $1.5B FCF, guidance above estimates. But the re-rate branch failed — not on profit quality, on sector beta: after a brief recovery (+4.51% Wednesday to ~$700), Friday's global chip rout took STX -7.98% to $644.58, ~7.7% BELOW its pre-print level, WDC -7.41% alongside. The interrogation stays unresolved because the market refused to pay for HDD earnings strength in an AI-cash-burn tape; the 'supply response' branch never got a clean fundamental test (pricing stayed strong). Carry-forward: HDD multiple vs memory-complex beta into WDC/STX December guidance. |
| 2026-10-25 | Pentagon x Anthropic into the IPO: if the public S-1 discloses the procurement fight prominently AND the designation stays enjoined through pricing, political risk gets priced as a disclosed-but-contained overhang; if the designation is reinstated (en banc/SCOTUS) or extends to other labs before mid-November pricing, a federal-revenue and political-risk discount enters lab-equity pricing for the first time — ANTP's $1.5T-vs-$2T debate gets a new, negative term. | 2026-11-20 | PENDING | Interim (10/26): the public S-1 is FILED (~400 pages, 90+ pages of risk factors). Data-center construction moratoriums in US jurisdictions are named as a specific growth risk (CryptoBriefing dissection of the filing); the Pentagon procurement fight's disclosure detail not yet verified — read on roadshow documents. Watch the pricing. |
| 2026-10-25 | GPU-collateral depreciation test: if new chip-collateralized loans signed over the next two months print with longer amortization (6-8yr) or tighter spreads, or CRWV yields return to single digits, Huang's 8-year thesis wins and the $500B financing pool reflates neocloud capex; if lenders hold 4-year marks and spreads stay wide, the financing flywheel tightens and second-tier neoclouds clear — watch actual deal terms, not rhetoric. | 2026-12-15 | PENDING | — |
| 2026-10-24 | SK hynix 10/28 capital-allocation framework: if a new dividend/buyback scale is announced AND Solidigm's US expansion is funded on the parent balance sheet, 'returns + sovereignty' becomes the Korean duo's template and the memory valuation floor gets set by capital-allocation policy before spot prices; a spin-off relaunch or a slash to expansion/return guidance falsifies the sovereignty read. | 2026-10-29 | HIT | HIT (resolved 10/28): SK hynix announced a ₩100T ($75B) shareholder-return plan for this year — ≥₩30T cash dividends + >₩60T buybacks-and-cancellations (including the completed ₩40T buyback pending cancellation); Q3 board approved ₩14T dividends plus a fresh ₩20T buyback. Caliber notes: the rumored Kioxia stake sale/₩16T special dividend was NOT confirmed; ₩10T of the new buyback is employee comp excluded from the tally (effective ~₩90T). Market: Seoul +10.89%, ADR +13.30%, KOSPI +6.63%. |
| 2026-10-24 | The 10/28 same-day double event: if MSFT raises FY27 capex language AND the stock closes up while the Fed holds with December-hike language hardened by oil, the 'AI equity discount' was Meta-specific and October's second leg files as positioning; 'prudent/pacing' language — or MSFT falling on a raise — confirms duration compression leg two. | 2026-10-29 | HIT | HIT (resolved 10/28): the day landed as scripted — SK hynix's call (record quarter, sold out through 2029, ₩100T return) + FOMC hold (split vote, meetings cut 8→6) + Microsoft's print (Azure +48%, capex cut to ~$175B). KOSPI +6.63% to 6,012.44 on the SKH print; SOX +1.75%; Alphabet's -10% after-hours print arrived the same night as the capex-escalation counterexample. |
| 2026-10-24 | China dual-front DRAM: if CXMT delivers its 4F2 server RDIMM by 12/31 AND YMTC's Fab 3 ramp shows customer evidence by mid-Q1 2027 (Counterpoint says 5k wpm of 7,500 MT/s LPDDR5X has run since Oct 2025; 9,600 MT/s next), 2028 conventional-DRAM oversupply odds reprice materially (worse for MU than SKH given HBM/LTA mix); if both slip, the second-front risk defuses. | 2027-01-15 | PENDING | — |
| 2026-10-22 | MSFT's FY27 capex language (10/28): if Microsoft raises capex guidance or FY27 language alongside Meta's raised floor and equity pivot, 'demand falsifies the debt narrative' holds and the October selloff files as positioning clear-out; if 'prudent/pacing' language appears, the top signal upgrades and duration compression gets a second leg. | 2026-11-04 | PENDING | — |
| 2026-10-22 | Meta's financing sequence: if Q4 capex lands within the implied $49-59B and the gap is funded via the first Class C issuance (or stacked debt) at reasonable discount, 'AI-dedicated equity' becomes a hyperscaler standard and the AI credit ceiling moves up; if Q4 capex falls below $49B or the Class C plan stalls under litigation, the financing-innovation read flips to a top signal. | 2027-02-05 | PENDING | — |
| 2026-10-22 | US-Iran tentative deal, two-week verification: if Brent follows WTI below $100, half of AI duration's macro headwind unlocks (December-hike consensus still priced); if talks collapse and Brent reclaims $110+, oil's second-round transmission stacks onto the December hike — the $12 Brent-WTI spread says the market believes half. | 2026-11-05 | PENDING | Interim (10/31): no deal inside the window — the opposite leg extended. Brent settled $117.64 Friday (+2.4%, six-month-high settle; ~+14% for October) as Iran said four tankers struck mines in Hormuz and one caught fire; the US is framed as weighing fresh strikes AFTER the 11/3 midterms (CNBC 10/28). The 'Brent follows WTI below $100' branch is effectively dead; final close 11/5 on the settlement prints. Pending — verdict by 2026-11-05. Interim (10/24): deal collapsed in 48 hours — Houthi attacks reached Saudi oil (Aramco HQ cleared per Energy News Beat), 7 tankers attacked in Hormuz (AP 10/22), UAE oil spill confirmed; WTI's highest settlement since April ($103.44), Brent >$114 intraday. The 'second-round transmission' branch is live; watch whether the US re-strikes and whether Brent holds $110+ into the FOMC. Interim (11/2): the biggest reversal of the window — Brent -6% Monday morning to ~$110.6 from Friday's six-month-high settle ($117.64) on Iran's official 'reopen Hormuz within two days if US accepts key demands' statement (third day of Doha talks; Qatar mediating a $6bn asset release per the Guardian; Araghchi: US response decides). NOT yet the sub-$100 branch: still above $110 and Washington hasn't accepted; verdict 11/5 as scheduled. Interim (11/03): The unlock branch arrived violently: Brent settled ~$104-105 Monday (-8%, first settle under $108; WTI ~$96.5 back under $100) on a circulating draft deal — half of AI duration's macro headwind is in the process of unlocking pending a second settle and signature; post-window US 'defensive' strikes on Qeshm left oil flat (the market treats them as tactical). |
| 2026-10-21 | Anthropic pricing tiers: final IPO pricing (mid-November target, draft S-1 filed 10/20) >= $2T closes the private-repricing chain; ~$1.5T = midpoint down, channel open; <$1T or delay = duration compression leg two. Front-run read: if OpenAI's reported $30B raise at $1.4T lands first at that level or better, the first public mark moved BEFORE the IPO. | 2026-11-30 | PENDING | Pending — verdict by 2026-11-30. Interim (11/03): Amended S-1 (Monday, in-window): Q3 revenue ~$17B (+45-47% QoQ, slowing from +145%), back to an operating loss ($567M), net loss $23.3B (mostly non-cash; 9M $104B), purchase commitments $895.4B (+$87B in a week); post-window TokenPost: valuation expectations now $1.5-1.6T — the midpoint-down branch is the base case unless the roadshow reprices upward. |
| 2026-10-21 | Tonight's trio (TSLA/META/NFLX, 10/21 after close): if FY27 CapEx language holds or rises, October's selloff resolves as positioning clearance falsified-by-demand; if any of the three introduces 'prudent'/'pace' language or trims CapEx windows, the debt narrative gets a demand-side confirmation and the top-signal upgrades. | 2026-11-07 | PENDING | Interim (10/22): META delivered the demand-side evidence — FY26 capex narrowed UP to $135-145B (Q3 capex $34.91B, first quarter above operating cash flow; FCF -$2.12B, first negative in nine; buybacks $0 YTD), but paired it with a NEW non-voting Class C share (effective 1/19/27) specifically to fund AI spending. TSLA beat with positive FCF yet cut spending 'ahead of a potential SpaceX deal' (Reuters). NFLX missed Q4 ad/revenue guidance and fell to a 26-month low with street-wide target cuts. Score so far: one raise (META), one cut (TSLA, merger-driven), one miss (NFLX, non-capex). MSFT 10/28 decides the verdict. |
| 2026-10-21 | SK hynix 10/28 call: if official HBM4 capacity/stacking numbers and LTA wording confirm 2027 lock-ins at rising prices, the 'wrongly sold vs supply break' fork resolves toward wrongly sold (KOSPI at a six-month low, pair at 2-3x P/E = positioning read); vague or hedged LTA language keeps the credit-side reading alive. | 2026-10-29 | HIT | HIT (closed 10/30, one day late — the 10-29 edition reported the outcome but the ledger wasn't updated). SK hynix's 10/28 call delivered the LTA anchor in full: DRAM+HBM 'effectively sold out through 2029', constraints projected through 2030, LTAs + contracts under discussion covering three years of output, plus a ₩100T ($75B) return plan — the 'wrongly-sold vs supply-cut' fork resolved toward wrongly-sold: SKH Seoul +10.89%, SKHY ADR +13.30%, KOSPI +6.63% (10/28), with sell-side re-rating starting ('YE27 5x P/E makes no sense'). Caveat noted in the ledger: no wpm-level HBM4 capacity number was disclosed (the 50k wpm caliber stays unverified). Follow-on: MU's December call is the industry-level final test (item dated 10-29). |
| 2026-10-18 | Nscale pricing action: if Nscale lists within two weeks at or near the ~$35B frame despite Glomfjord's six-month delay, the primary market is 'expensive but open' and the junk-shutdown narrative downgrades to spread-widening; a pulled, discounted or further-delayed deal confirms funding availability (not price) is the bottleneck and duration compression enters leg two. | 2026-11-06 | PENDING | Interim (10/21): two-sided evidence landed inside the window — Microsoft pulled out of Nscale's UK Essex project over power ('in the 2030s', Telegraph/SiliconANGLE) and investors pushed back on the ~$30-35B valuation (Bloomberg/TNW), while Bloomberg/CNBC report Anthropic in talks for a SECOND $45B deal on top of the $44.6B booked. Watch upgraded: pricing action vs counterparty re-negotiation is now the live fork. Interim (10/26): Microsoft is reported to have CANCELED the UK AI data-center project over power delays (TradingView Sunday relay — consistent with the 10/21 Essex exit; if the same project, this is the formal cancellation). SpaceX's closed $40B GPU debt (Reuters/Bloomberg, Sunday) tightens the same vise from the financing side; Nscale's pricing action itself still pending into the 11/6 verdict. Interim (10/26): Microsoft is reported to have CANCELED the UK AI data-center project over power delays (TradingView Sunday relay — consistent with the 10/21 Essex exit; if the same project, this is the formal cancellation). SpaceX's closed $40B GPU debt (Reuters/Bloomberg, Sunday) tightens the same vise from the financing side; Nscale's pricing action itself still pending into the 11/6 verdict. |
| 2026-10-18 | Earnings-week CapEx language: if TSLA (10/21), META and NFLX — then MSFT (~10/28) — hold or raise FY27 AI CapEx language through the credit scare, the demand side falsifies the debt narrative and October marks a positioning flush; any 'pace/cautious' hedging converts October's credit noise into an earnings-cycle top signal. | 2026-11-06 | PENDING | Pending — verdict by 2026-11-06. Interim (10/22): META raised the FY26 floor to $135-145B (implied Q4 $49-59B) and invented a funding vehicle — equity, not 'prudence'; TSLA cut spending but attributed to the SpaceX merger; NFLX weakness is ad-side. The 'capitulation in capex language' branch has NOT triggered; the 'demand falsifies debt narrative' branch needs MSFT 10/28. |
| 2026-10-18 | Brent vs $110: if Brent holds above $110 into next weekend as Hormuz attacks continue (two tankers mined Saturday), December-hike pricing hardens and the AI duration shock gets a second macro leg; a fall back below $100 removes the largest macro headwind for AI equities. | 2026-10-30 | HIT | HIT. Brent held above $110 through the verdict window: settlements ran $115.81 (10/26), $113.88 (10/27), $115.32 (10/28), $114.88 (10/29) — Friday 10/30 intraday ~$112.3 but still above the line. The macro leg played out as specified: no Reuters/AP-grade ceasefire confirmation materialized (weekend 'framework' reports stayed single-source), PCE printed 3.6% with the December-hike consensus intact, and AI duration assets traded under an open rate channel all week (10Y 4.95→5.05→4.99). Follow-on watch: whether the post-midterm window (CNBC: US reportedly told Gulf states to prepare for strikes after 11/3) breaks the $110-115 platform either way. |
| 2026-10-17 | Fed + credit fork: if the Fed holds on 10/27-28 while CoreWeave yields normalize to single digits (and big tech raises CapEx at earnings), October marks an Aug-2024-style financing-panic V-bottom and multiple compression ends; a hike with yields sustained above 12% (or new funding failures) opens duration compression, leg two. | 2026-11-13 | PENDING | Pending — verdict by 2026-11-13. Interim (11/03): Spreads have NOT normalized — the opposite: ORCL's $20B raise (with an equity leg) was punished -8.76%, CRWV hit an all-time low, CPP turned selective — leg-two duration compression is running through equity dilution rather than bond yields (10Y 4.89); the V-bottom branch needs the December-hike repricing (now live via the oil break) to offset. |
| 2026-10-17 | Korea full-Q3 verdict: if Samsung/SKH confirm rising LTA prices and an official HBM4 capacity number (the 50k wpm frame), memory is a contract annuity and this -25% was a multiple-event mispricing; soft LTA language or a capacity miss hands the CXMT fear fundamental backing and the memory valuation midpoint steps down. | 2026-11-13 | PENDING | Pending — verdict by 2026-11-13. Interim (10/24): SKH voluntarily scrapped Solidigm's US IPO (10/23) and put capital allocation on the official 10/28 agenda; BofA sees bigger returns for the duo; KOSPI -3.52% Thursday (buyback vacuum + US contagion) then +0.30% Friday with SKH +4.42% — the first differentiated positive read of October. The 10/28 call now resolves three things at once: HBM4 capacity, capital-allocation template, and the LTA wording. |
| 2026-10-17 | Big-tech CapEx defense: if META/TSLA/NFLX then MSFT/GOOGL raise or hold FY27 CapEx language through the credit scare, the demand side falsifies the debt narrative and October was a positioning flush; any 'pace/cautious' hedging converts October's credit noise into an earnings-cycle top signal. | 2026-11-13 | PENDING | Pending — verdict by 2026-11-13. Interim (10/22): META raised the FY26 capex floor to $135-145B and paired it with the Class C funding vehicle — the defense is being funded by equity innovation rather than debt. Awaiting MSFT/GOOGL. |
| 2026-09-26 | MU 9/30 five-in-one print: if management proactively addresses 'power delay → demand timing' (and/or discloses lab direct-buying), interconnection schedules have entered the demand model and the memory re-rate extends with structure; explicit denial of timing impact falsifies the FM-transmission chain into 2027 demand. | — | PARTIAL | 9/30 print: $54.2B record quarter ($40B DRAM), GAAP NI $37.7B, Q1FY27 guided $61.5B; management disclosed ~75% of 2027 supply sold out, 'tight through 2028' — the structure disclosures landed (hit on that leg) and MU printed its all-time high 10/1 ($1,097.40). But the re-rate did not extend: an October credit shock (CXMT 4F² + CoreWeave 15% + hot CPI) took MU -20% from peak by 10/16. Power-delay/lab-direct language couldn't be verified from CI transcripts. Partial: disclosure hit, extension falsified by exogenous credit shock. |
| 2026-09-26 | Force-majeure diffusion (two-week window): if a second FM notice appears, Blue Owl-type project debt stays below 90c, or SB Energy's IPO stays shelved — the 'power bottleneck becoming contractual' is a trend, not an event; debt back to par with no second FM downgrades it to a one-off. SoftBank's post-9/29-settlement secondary is the companion read. | — | PARTIAL | No second FM notice found through 10/16; Blue Owl project-debt marks not retrievable in CI; SB Energy's IPO stayed shelved. But the thesis validated through a different instrument: CoreWeave straight-bond yields ~15% (10/14), neocloud equities -38~-42% from peaks, 'AI debt fears' going mainstream, 9fin reporting DC lenders debating construction-risk materiality. The financing layer repriced delay/duration risk — via balance sheets, not project bonds. Partial: trend confirmed, channel different. |
| 2026-09-26 | Application-layer bargaining: if Nebius's Oct 1 hikes (17-21%) hold volumes for 1-2 weeks AND Microsoft's late-Oct FY27Q1 gives a Copilot adoption/ARPU disclosure, the token-deflation narrative yields to 'total-bill expansion' (Jevons structure); volume attrition or no Copilot data keeps the deflation frame alive. | — | PENDING | 10/17 update: no public volume-retention data since the 10/1 hikes; NBIS -38.5% from the August peak moved on sector credit, not churn evidence. MSFT FY27Q1 (~10/28) is the deciding read — watch extended. |
| 2026-09-25 | MU 9/30 three-theme print: if HBM4 share/2027 contract-price language lands strong AND management addresses power-delay timing (or discloses lab direct-buying), the memory re-rate extends into October with structure, not just slope; if guidance leans only on price slope with no contract-structure disclosure, the 'high-level convergence' camp regains the narrative. | 2026-10-09 | PARTIAL | Structure delivered (75% of 2027 sold out, sold-out HBM, multi-year language), price slope strong ($61.5B guide) — but the October re-rate ran the other way on credit/China shocks, so 'extends with structure' failed on price even though disclosure passed. Same basis as the 9/26 five-in-one score. |
| 2026-09-25 | Force-majeure diffusion: if a second FM notice appears, Blue Owl-type project debt stays below 90 cents, or SB Energy's IPO stays shelved past mid-October, 'power bottlenecks going contractual' becomes a trend and AI DC financing spreads widen structurally; if debt recovers toward par with no new notices, score it a contained one-off. | 2026-10-22 | HIT | Markers: no second FM (absent), Blue Owl marks unverifiable in CI, SB Energy still shelved past mid-October (hit) — and AI DC financing spreads widened structurally through the neocloud channel: CRWV ~15% yield, SoftBank equity -20%, Anthropic roadshow talk $2T->$1.5T, Goldman/Gartner consolidation framing. 'Power bottlenecks going contractual' upgraded to 'capital going structural.' Scored hit on the spread-widening arm. |
| 2026-09-25 | NBIS acceptance test: if utilization holds through the Oct 1 hikes (17-21%) and CRWV's next contracts confirm premium short-dated pricing, compute pricing power re-rates CRWV/NBIS and extends GPU/memory procurement visibility; if churn appears or CRWV new contracts price flat, 'pricing power' downgrades to temporary scarcity premium. | 2026-10-16 | PARTIAL | Rate hikes took effect 10/1 as notified; no public utilization/churn data since; CRWV's premium short-dated pricing not re-confirmed in new contracts within the window. Sector repricing was driven by credit, not demand elasticity. Indeterminate on retention: partial. |
| 2026-09-24 | Lab dis-intermediation: if MU's 9/30 call and Samsung's October results disclose the first lab-direct memory contracts (or the 1GW direct lease signs), the memory pricing benchmark starts migrating from channel to terminal LTAs and the intermediary pricing power of NVDA/cloud margins erodes; if 2027 direct buying fails to materialize at reported scale or the 1GW talks collapse, 'dis-intermediation' downgrades to a bargaining chip. | 2026-10-31 | PARTIAL | PARTIAL (closed 10/31 as planned). Direction confirmed at document level: Anthropic's public S-1 (10/26) discloses direct leases + a $567B compute plan with self-built data centers (Fortune read-through) — 'lab dis-intermediation' graduated from DIGITIMES rumor to SEC filing. But the contract legs never landed inside the window: MU's 9/30 call and Samsung's 10/29 call disclosed NO first lab direct-procurement contracts, and Anthropic's reported ~1GW direct lease (Apollo/Stream talks) remained unsigned. File-level evidence yes, contracts no — hence partial. |
| 2026-09-24 | Two-track AI financing: if SoftBank's bonds hold near par after the 9/29 settlement (9-10% prints standing), the marginal funding cost for credits without convert access caps neocloud supply growth and forces CRWV/NBIS-class players toward convert/equity structures; if spreads compress rapidly toward convert levels, score it 'expensive but smooth' — sector-positive. | 2026-10-22 | HIT | SoftBank bond secondary not directly observable in CI, but the mechanism landed emphatically: CoreWeave straight-bond yields ~15% vs CRWV converts at 2.875% — the marginal funding cost for credits without convert access now caps neocloud supply growth (Gartner: most won't survive independently). 'Expensive but smooth' did not materialize: SoftBank equity -20% since 9/26 (9984.T ¥4,944). Hit on the wide-spread arm. |
| 2026-09-24 | Muse take-rate: if the 9/29 Sensor Tower weekly holds #1 while no additional major retailer joins Amazon's blockade, the agent take-rate story converts into GMV expectations (META/SHOP/payments chain); if rankings decay or the blockade spreads, the download-chart rally marks down as narrative positioning. | 2026-10-08 | PARTIAL | 10/17 note: no Sensor Tower weekly or blockade-breadth data visible through the gap; unresolved — kept pending, flagged for manual follow-up. 10/25 close-out (partial): no public Sensor Tower weekly data ever resumed through the verdict window; the Amazon blockade escalated into a 'standoff' (CNBC 9/23) which is the weaker half of the falsification branch, but rankings data stayed unavailable. Scored partial on data availability, not on direction. |
| 2026-09-23 | Token price war: if OpenAI/Anthropic API volume (tokens consumed) inflects up over the next two quarters while agent workloads scale (Muse retention, per-user costs at Connect), the -20/-50% cuts read as 'deflation for penetration' and the compute/memory second curve steepens; if volume fails to respond, the cuts print as ARPU suicide and model-layer revenue estimates come down. | 2026-10-21 | PARTIAL | 28-day interim (scored partial, volume leg unresolved): price deflation continued into late October — DeepSeek V4.1-Flash (10/17, faster and cheaper) and MiniMax M3.1 Flash undercutting DeepSeek/Opus 5 on business benchmarks; no lab discloses token consumption, so the volume response stays publicly unobservable. Indirect reads point the 'deflation for penetration' way (SMCI $4B above guidance, Reuters' $35B+ China order book, Anthropic in talks for a second $45B compute deal) — consistent with, but not proof of, a steepening compute/memory second curve. |
| 2026-09-23 | Alibaba compute stack: if Zhenwu V900 hits 1Q27 mass production with cluster deployments at scale (and 20GW/2032 stays in official plans), China compute self-sufficiency gets a hard date and NVDA/AVGO China TAM takes a structural cap; a slipped timetable or a big benchmark-versus-claims gap kills the signal. | 2027-04-30 | PENDING | — |
| 2026-09-23 | Agent victim list: if Meta Connect (9/23) discloses Muse monetization and the 'consumer inertia' discount (Intuit, retail banks) deepens while agent retention holds, the inertia-premium deflation becomes a durable cross-sector trade; if Connect is devices-only and rankings decay, both the victim list and the CPU rally mark down as narrative. | 2026-10-07 | PARTIAL | 10/17 note: verdict date (10/7) passed with no CI-visible weekly-chart or inertia-discount data through the gap; unresolved — kept pending pending data, flagged for manual follow-up. 10/25 close-out (partial): Connect (9/23) was not devices-only — agent-commerce take-rate mechanics were disclosed (Reuters) and the 'inertia discount' hit bank/Intuit names on 9/23 — but no durable cross-sector data emerged through October and the Muse narrative faded from Tier-1 coverage during the credit scare; the victim-list trade was overwhelmed by macro. Scored partial: mechanism confirmed, durability unproven. |
| 2026-09-22 | Meta Connect (9/23): if Muse monetization or per-user infrastructure economics are disclosed and weekly app-chart retention holds, the CPU rally gets a fundamental leg and the memory second leg (server DRAM/enterprise SSD) starts; if Connect is devices-only and retention decays, Monday's Arm +17/INTC +12 prints as narrative positioning. | 2026-09-29 | PARTIAL | Connect did disclose the agent take-rate (9/23); memory's second leg did start (MU peaked 10/1, DRAM sold out) — but the CPU-rally leg (ARM -31.6% from the 9/22 peak, INTC faded) unwound with October's credit shock before the two-quarter test completed. Partial: one of two legs confirmed, then both repriced by an exogenous shock. |
| 2026-09-22 | SoftBank $11B junk bond (~9/24 pricing): the spread is the market's grade of OpenAI-adjacent credit — a wide print (or pulled deal) re-rates ORCL/CRWV/NBIS backlogs at the customer-credit level; a clean print at moderate spread confirms the market will keep paying for OpenAI. | 2026-10-06 | PARTIAL | Priced 9/23 at $11.1B multi-tranche: USD notes ~9-10% (3.5/5.5/7.5yr), EUR mid-8%; book >2x ($20B+); settles 9/29, funds the third $10B OpenAI tranche 10/1. 9984.T opened +7% on Tokyo's reopen. Verdict: the market keeps paying for OpenAI — but at a near-10% explicit premium. 'Willing, not cheap.' 10/17 addendum: 'willing, not cheap' degraded further toward 'tolerance strained' — 9984.T ¥4,944 on 10/16 vs ¥6,150 on 9/25 (-20%), alongside the neocloud credit shock; bond secondary itself not observable in CI. |
| 2026-09-22 | Texas permit freeze: if the audit/permit halt lasts >30 days or other states copy (post-California), 'permit wait time' becomes a priced scarcity in DC supply and already-permitted sites/assets (EQIX/DLR, site-locked neoclouds) re-rate; if permits resume within 30 days, it was noise. | 2026-10-22 | HIT | Closed early (10/21): Inside Climate News (10/19, 404 Media corroborating) documented that Texas kept permitting data centers after Abbott's pause — the halt was never a hard 30-day freeze, and no other state copied with one. Per the claim's own mapping, that resolves to the 'noise' branch: political-risk premium deflated, and already-permitted assets (EQIX/DLR) showed no scarcity re-rate — both fell with the sector. The priced-scarcity thesis stays shelved; the real Texas constraint remains water/power enforcement (Paxton's 156-site letters), not permit waits. |
| 2026-09-21 | Summit 9/24: if the AI dialogue/notification mechanism makes it into a joint readout with no new export-control tightening, the policy-acceleration vector is confirmed and the forced-slowdown channel stays closed. | 2026-09-28 | HIT | 9/24: US proposed an AI notification mechanism (Bessent: mutual notification of major AI developments/threats); leaders' readout includes AI-cooperation language (Xi: keep AI under human control); trade truce extended 2 months; no new export-control tightening. Mechanism implementation unresolved (Yahoo: 'questions abound'). SSE -1.22% = the market's 'underwhelm' verdict. Policy-acceleration vector confirmed; forced-slowdown channel stays closed. |
| 2026-09-21 | CXMT G5: if LPDDR5X from G5 shows up at scale in Chinese flagship phones by 1Q27 while mobile DRAM prices flatten relative to server, the shortage-through-2030 consensus takes a supply-side haircut. | 2026-10-21 | PARTIAL | Final 30-day read (10/21): partial. The supply-side haircut happened — but through the 4F² server-RDIMM vector and CXMT's blockbuster post-IPO print (record Q3 revenue, up to $19.8B nine-month net income guided, buyback pledged; shares -7.8% Monday on slowing growth), not through verified G5 LPDDR5X at scale in Chinese flagships. Mobile-flagship node remains unverified; that leg rolls into the 12/31 4F² deadline and the 1Q27 flagship cycle. |
| 2026-09-21 | Ceasefire trade: if the US-Iran framework fails within two weeks (WTI back above $100), Monday's futures-led rally gets given back; if it holds, ~$94 marks the war premium de-rating. | 2026-10-05 | HIT | Framework broke at week three (window was two): Hormuz attacks escalated sharply from 10/14 (NYT), Brent touched $110 on 10/16, WTI closed $101.41 (+10.5% wk) — far above the $100 break line; S&P gave back the futures-led rally (-1.9% on the week). Outcome as predicted, timing three days late. Hit with a timing caveat. |
| 2026-09-20 | Trump-Xi summit week: 'quiet = success' is the priced-in outcome; any concrete deal on AI diffusion, procurement or oil rewrites Monday's sector relative performance. | 2026-09-27 | HIT | Quiet outcome delivered: truce extension + hotline proposal, no concrete AI-diffusion/procurement/oil deal; Monday's sector relative performance held. SSE -1.22% on the day confirms 'quiet = success' was the correct base case. |
| 2026-09-20 | Pace-the-frontier transmission test: if no lab publicly defers/cuts frontier compute orders through October (Nscale IPO pricing + Rubin rack orders as revealed preference), the slowdown stays a narrative event (Form 1). | 2026-10-20 | HIT | Scored 10/17 (3 days early; evidence complete): no lab publicly deferred/cut frontier compute orders through the window — Anthropic's roadshow is live, Rubin entered production with CoreWeave shipping NVL72 at scale (9/30). The slowdown stayed a narrative event (Form 1); the October selloff came from financing, not frontier caution. |
| 2026-09-20 | Index-flow check: if SNDK gives back more than half of Friday's +10.99% within 5 sessions after the 9/21 S&P 100 effective date, positioning dominated the memory melt-up going into MU's 9/30 print. | 2026-09-28 | HIT | Final 5/5 (9/25 close $1,777.80): -0.8% vs the 9/18 close ($1,791.82), never approached the half-giveback line (~$1,703). Two -3% days inside the window but no breakdown — index-flow momentum quality passed. |
| 2026-09-18 | Micron FY26Q4 (9/30): HBM4 share and 2027 contract-price guidance decide whether the memory cycle is 'accelerating' or 'converging at a high level'. | 2026-10-05 | PARTIAL | Neither pure acceleration nor convergence: TrendForce 4Q26 kept price increases positive but converging (AI-server led, consumer weak); MU's structure disclosures (75% sold out) argue 'high-level sustain with contracts.' The market resolved the fork violently in October — via credit/China shocks, not via the price data. Partial. |
| 2026-09-18 | TrendForce 4Q26 forecast (~9/24–25): if DRAM gains keep converging while NAND flattens, the 'supercycle' acceleration premise needs re-rating. | 2026-09-28 | PARTIAL | Published 9/30: 'AI server demand sustains memory contract price increases in 4Q26, consumer-side pressure persists' — DRAM gains kept converging (the claim's leg confirmed), NAND/eSSD stayed firm. But October's supercycle re-rating came from credit and China-supply shocks, not from this convergence. Partial: direction confirmed, driver different. |
| 2026-09-18 | CoreWeave convertible pricing coupon is the direct read on neocloud cost of capital — a rich coupon re-rates the whole sector. | 2026-09-25 | HIT | Priced 9/19 at 2.875% (upsized to $3.7B) — cheaper than feared. Settled 9/22 at $4.2B with the $500M option exercised in full; worker recorded the settlement same day. |
| 2026-09-18 | Nebius raising on-demand GPU rates on 10/1 tests whether neocloud pricing power is industry-wide. | 2026-10-08 | PARTIAL | Hikes took effect 10/1 as notified (second in three months) — industry-wide pricing power confirmed on the rate card. Acceptance (volume retention) remained unobservable publicly through 10/16; NBIS's -38.5% drawdown from peak tracked sector credit, not demand. Partial. |
Updated 2026-11-03