# Public markets: concentrated earnings, leveraged ownership and the price of growth *Last updated: 2026-10-25 · Author: Codex / GPT-6 · Status: reviewed · Reviewers: Codex (gpt-6-astra) cross-check applied 2026-10-17; Claude (Opus 5.5) cross-check applied 2026-10-18 · Updated 2026-10-21 from signals/2026-10-21.md · Updated 2026-10-25 from signals/2026-10-25.md* ## Summary - On September 30, 2026, the top ten S&P 500 companies carried 40.6% of capitalization and 36.1% of trailing earnings. Their forward P/E was 20.9x against their own 20.8x chart average; the other 490 traded at 17.9x against 16.0x (112%). Those multiples imply a 36.9% forward-earnings share. On October 15, SPY's ten largest security lines were 38.63% of fund assets, including two Alphabet classes; Nvidia was 7.88%. These are different cohorts and dates (claim 1). - As of October 12, 2026, Micron and Nvidia supplied 36% of expected Q3 S&P 500 earnings growth, while 74% of constituents were expected to grow earnings. FactSet's October 16 blended Q3 growth was 29.0%; 9% had reported, of whom 84% beat EPS estimates. Broad growth coexists with concentrated incremental profits (claims 8–9). - The S&P 500 forward P/E was 19.0x on October 16, 2026, unchanged from September 30 (five-/ten-year averages 19.8x/19.1x; JPM's September 30 thirty-year average 17.2x). Nasdaq-100/SOX NTM P/Es were 21.36x/21.89x on October 13. Published September CAPE was 40.58 using estimated CPI; substituting released CPI-U of 336.940 for 335.511 gives approximately 40.41 (B arithmetic). Forecast earnings and historical real earnings remain different denominators (claims 3–6, 45–46). - Through October 16, 2026, Micron and AMD gained 207.92% and 146.89% YTD. Oracle/CoreWeave were down 60.94%/66.56% from their highest closes since January 2025 or listing, versus 48.33%/55.51% from their 2026 peaks; Vertiv/SK Hynix were down 45.13%/56.42% on either basis. Nasdaq-100 fell 5.49% in the week ended October 16 (A official Nasdaq levels; B arithmetic); the individual-share returns retain C. In a separate October 16–20 extension fetched October 21, NDX/SOX rebounded 1.04%/1.32% and M1/M1a/M1b gained 1.38%/1.79%/0.69%; October 21 intraday prices are excluded (claims 21, 57). M1/M1a/M1b were already 23.8%/14.5%/43.4% below their 2026 basket peaks (claims 11–21, 50). The rebound reversed after Meta's results: at the October 23 close M1/M1a/M1b stood at 97.02/99.09/93.63 (origin 100 on October 16; triggers at 60 need further falls of 38.2%/39.5%/35.9%), and official NDX was 7.26% below its October 6 peak. October 22 alone took M1 down 6.0%, with Lam −11.8%, CoreWeave −10.1%, GE Vernova −10.6% and Hut 8 −10.6%; Reuters reported SOX −6.7%, its worst day since June 23 (claims 60–63). - September 2026 margin debt was $1.480054tn, up 31.4% YoY; June's $1.502072tn was the record in FINRA's 1997–2026 series. TQQQ held $34.013bn on October 16 and targeted three times daily Nasdaq-100 returns. Margin calls and daily leverage resets can amplify losses; cap-weight drift requires no trading, and options turnover alone does not identify directional appetite (claims 22–26). - CoreWeave reported Q2 2026 revenue of $2.575bn and a $626m loss. On October 15 its five-year CDS reportedly hit 1,194.5bp intraday, its 2031 notes yielded 15.78%, and Meta-linked Hyperion/Beignet project bonds traded just under 92 cents. Later snapshots show CoreWeave CDS nearly 1,081bp and 2031 YTM 15.97% on October 20; Oracle CDS was 315/322bp on October 19. Broad HY/IG OAS was 323/84bp on October 15. These observations extend selective credit stress to traded project paper without measuring all funding channels (claims 37, 48–49). The stress reached an investment-grade hyperscaler after its results: Meta five-year CDS hit a record 114.79bp (Markit, +7.6bp) on October 22, with a separate ICE intraday high near 119.3bp; FactSet's dated October 21 snapshots put CoreWeave at 1,065.84bp and Oracle at 308.25bp. These are quotes, not credit events (claims 64–65). - Elice raised 201.1bn won ($151m), yet closed its October 20 debut at 61,500 won against a 90,500 won offer (−32%). The IPO observation extends evidence of weaker infrastructure risk appetite; it does not identify CoreWeave-like balance-sheet risk. Griffin, Fink and Sewing disagree about bubble timing, and Yardeni’s October 21 meter separates high conditions (83/100) from lower triggers (34/100), without calibrating this report’s forecasts (claims 52–54). - Listed Chinese labs show the equity-side repricing more sharply: Bloomberg reported on October 23 that MiniMax and Zhipu had lost about $150bn combined from their peaks, Zhipu 81% below its June high and MiniMax below its IPO price on October 22, before MiniMax rebounded 16.74% to HK$187.60 on October 23. CSI 300 closed −1.8% on October 22, 20% below its June high. Prediction venues moved little: Polymarket's bubble cards read 9% by end-2026 and 23% by June 30, 2027 on October 25 (claims 55, 66–67). ## Key claims | # | Claim | Value | As of | Source | Type | Grade | |---|-------|-------|-------|--------|------|-------| | 1 | Index concentration, with units distinguished | Top-ten companies: market cap 40.6%, trailing earnings 36.1%; later SPY fund holdings: ten security lines (nine companies) 38.63%, Mag Seven 34.60%, NVDA 7.88%; JPM top-ten forward P/E 20.9x versus own chart average 20.8x (100%); other 490 17.9x versus 16.0x (112%); implied forward-earnings share (0.406/20.9)/(0.406/20.9 + 0.594/17.9) = 36.92% | 2026-09-30; SPY fund weights 2026-10-15 | [JPM, pp. 8–9](https://cdn.jpmorganfunds.com/content/dam/jpm-am-aem/global/en/insights/market-insights/guide-to-the-markets/mi-guide-to-the-markets-us.pdf); [State Street holdings, sum](https://www.ssga.com/us/en/individual/etfs/state-street-spdr-sp-500-etf-trust-spy) | secondary / primary | B concentration compilation and sums / A published P/Es and NVDA weight / B implied forward share | | 2 | Mag Seven earnings-share benchmark; **stale, forward earnings** | 25.2% of forward earnings versus 34.8% market cap; current matched-period Mag Seven earnings share not verified | 2025-11-28 | [RBC GAM, exhibit 19](https://www.rbcgam.com/documents/en/articles/market-outlook-new-year-2026.pdf) | secondary | B | | 3 | S&P 500 valuation | Forward NTM 19.0x, unchanged from September 30; five-/ten-year averages 19.8x/19.1x; State Street trailing 24.90x, FY1 20.59x (FY1 differs from rolling NTM); JPM 30-year average 17.2x, +1 SD 20.5x at 2026-09-30; 19.0x is (19.0−17.2)/(20.5−17.2) ≈0.55 SD above that average | 2026-10-16; State Street 2026-10-15 | [FactSet](https://insight.factset.com/sp-500-earnings-season-update-october-16-2026); [State Street](https://www.ssga.com/us/en/individual/etfs/state-street-spdr-sp-500-etf-trust-spy); [JPM pp. 5, 9](https://cdn.jpmorganfunds.com/content/dam/jpm-am-aem/global/en/insights/market-insights/guide-to-the-markets/mi-guide-to-the-markets-us.pdf) | primary estimates / primary | A published estimates / B SD arithmetic | | 4 | Nasdaq-100 / SOX next-twelve-month P/E | 21.36x / 21.89x; respective ten-/five-year means 23.09x / 22.12x | 2026-10-13, published 2026-10-16 | [Nasdaq dashboard](https://www.nasdaq.com/docs/global-indexes/global-markets-dashboard-oct-16-2026) | primary | A | | 5 | Shiller CAPE, checked against project CSV | September published vintage 40.58; substituting released September CPI-U 336.940 for estimated 335.511 gives ≈40.41 (40.580874×335.511/336.940, holding other inputs fixed; approximate CPI adjustment, not a fully rebuilt workbook). October 40.71 uses **October 2 close**, not latest close; October CPI remains estimated. Historical benchmarks: 44.20 December 1999, 27.32 October 2007 | 2026-09 / 2026-10 | [Shiller workbook](https://shillerdata.com/); [local extract](/data/theory-shiller-cape.csv); [BLS September CPI](https://www.bls.gov/news.release/archives/cpi_10142026.htm) | primary | A published vintage / B approximate CPI-adjusted CAPE | | 6 | Earnings-yield spreads, **not a modelled expected ERP** | Treasury 5.10%, real TIPS 2.78%; forward earnings yield 5.26%, less Treasury ≈+0.16 percentage points; September CAPE yield 2.46%, less TIPS ≈−0.32 pp | yields 2026-10-15; mixed-date valuation inputs | [DGS10](https://fred.stlouisfed.org/series/DGS10), [DFII10](https://fred.stlouisfed.org/series/DFII10), claims 3, 5; arithmetic | primary | A yields / B calculations | | 7 | Nvidia sales valuation, approximate outstanding-share basis | $214.13 close × 24.1bn shares ≈$5.16tn; TTM revenue $302.970bn = $215.938bn + $177.837bn − $90.805bn; P/sales ≈17.03x | close 2026-10-16; shares 2026-08-21; TTM to 2026-07-26 | [10-K](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm), [10-Q](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm), [prices](https://query2.finance.yahoo.com/v8/finance/chart/NVDA?range=2y&interval=1d) | primary / secondary | A disclosed financial inputs / B derived valuation | | 8 | Calendar-year consensus earnings hurdle | 2026 growth 32.8%; earlier EPS estimates $362.65 for 2026, $419.32 for 2027 imply 15.63% subsequent growth; Q3 blended growth 29.0%, with 9% reported and 84% of reporters beating EPS estimates | growth 2026-10-16; EPS 2026-10-02 | [FactSet](https://insight.factset.com/sp-500-earnings-season-update-october-16-2026); [State Street citing FactSet](https://www.ssga.com/no/en_gb/institutional/insights/mind-on-the-market-12-october-2026) | primary estimates / secondary | A / B | | 9 | Earnings dependence and contrary breadth evidence | MU + NVDA: 36% of expected Q3 growth; ten largest growth contributors (not necessarily largest by capitalization) roughly two-thirds; 74% of constituents expected to grow earnings | 2026-10-12 | [Citadel Securities](https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence) | secondary | B | | 10 | Strongest fundamentals-based rebuttal | Nasdaq describes SOX prices and earnings both up about 60% YTD, with another 75% earnings growth projected by end-next-year | 2026-09-03 | [Nasdaq](https://www.nasdaq.com/economic-institute/the-print/second-quarter-earnings-show-breadth) | primary research | A for published estimates | | 11 | NVDA price returns: 2025 / 2026 YTD / latest week / current drawdown | +38.88% / +14.82% / −6.61% / −10.50% | through 2026-10-16 | [Yahoo daily series](https://query2.finance.yahoo.com/v8/finance/chart/NVDA?range=2y&interval=1d) | secondary | C | | 12 | AVGO, same definitions | +49.28% / −5.01% / −9.07% / −31.73% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/AVGO?range=2y&interval=1d) | secondary | C | | 13 | TSM US ADR, same definitions | +53.88% / +35.78% / −8.98% / −15.07% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/TSM?range=2y&interval=1d) | secondary | C | | 14 | ORCL, same definitions | +16.96% / −34.21% / −9.31% / −60.94%; 2026-peak drawdown −48.33% ($248.15, 2026-06-01 → $128.23) | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/ORCL?range=2y&interval=1d) | secondary | C | | 15 | CRWV, listing-period / YTD / week / drawdown | +79.03% from first listing close to end-2025 / −14.27% / −25.21% / −66.56%; 2026-peak drawdown −55.51% ($137.98, 2026-05-06 → $61.39) | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/CRWV?range=2y&interval=1d) | secondary | C | | 16 | AMD, same definitions as NVDA | +77.30% / +146.89% / −13.05% / −18.58% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/AMD?range=2y&interval=1d) | secondary | C | | 17 | MU, same definitions | +239.13% / +207.92% / −14.59% / −27.58% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/MU?range=2y&interval=1d) | secondary | C | | 18 | VRT, same definitions | +42.60% / +27.41% / −14.98% / −45.13% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/VRT?range=2y&interval=1d) | secondary | C | | 19 | ASML US line, same definitions | +54.36% / +53.43% / −7.80% / −17.49% | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/ASML?range=2y&interval=1d) | secondary | C | | 20 | SK Hynix Korean shares, KRW | +274.35% / +95.39% / −24.33% / −56.42%; year-end bases December 30 in both years, weekly base October 8 | through 2026-10-16 | [Daily series](https://query2.finance.yahoo.com/v8/finance/chart/000660.KS?range=2y&interval=1d) | secondary | C | | 21 | Week ending October 16 and chip losing streak | NDX −5.49% week (official EOD 29,187.5487 / 30,883.1478 − 1); 2026 peak 31,224.6854 on October 6 → −6.52%. SOX −11.41% week; eight consecutive SOX declines, October 7–16. Independently re-fetched October 16→20 week-to-date: official NDX 29,187.5487→29,490.2834 (+1.04%), SOX 11,137.2867→11,283.8120 (+1.32%); these are two-session moves, not a completed week | 2026-10-16; WTD through 2026-10-20, fetched 10-21 | [Nasdaq official NDX history](https://indexes.nasdaq.com/Index/History/NDX), POST HistoryData id=NDX, startDate=2026-01-01, endDate=2026-10-16, timeOfDay=EOD (199 observations); [Nasdaq official SOX history](https://indexes.nasdaq.com/Index/History/SOX), same POST with id=SOX and endDate=2026-10-20; [SOX daily series](https://query2.finance.yahoo.com/v8/finance/chart/%5ESOX?range=2y&interval=1d); [Bloomberg/SWI](https://www.swissinfo.ch/eng/stocks-fall-as-coreweave-selloff-fuels-ai-jitters%3A-markets-wrap/92225060) | primary NDX / secondary SOX | A official NDX and new SOX EOD / B official-index arithmetic / C SOX weekly return / B independently corroborated streak | | 22 | Margin borrowing | September $1.480054tn; June $1.502072tn, the highest in FINRA's January 1997–September 2026 series; September 2025 $1.126494tn → +31.39% YoY. October 2025 $1.183654tn → +25.04% over eleven months; September versus June −1.47% | 2026-09 | [FINRA](https://www.finra.org/rules-guidance/key-topics/margin-accounts/margin-statistics); [local CSV](/data/theory-finra-margin-debt.csv); [FINRA history workbook](https://www.finra.org/sites/default/files/2021-03/margin-statistics.xlsx) | primary | A / B arithmetic | | 23 | Retail participation, **one market maker's platform** | Weekly net purchases $9.1bn; semis $2.6bn, AI infrastructure $2.4bn, neoclouds $0.7bn; themes may overlap | week reported through 2026-10-11 | [Citadel Securities](https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence) | primary | A | | 24 | Leveraged ETF observable, not industry total | TQQQ net assets $34.013bn; targets 3× daily Nasdaq-100 performance | 2026-10-16 | [ProShares](https://www.proshares.com/our-etfs/leveraged-and-inverse/tqqq) | primary | A | | 25 | Options activity | OCC YTD options ADV 70.565m contracts, +20.6%; September SPX zero-day ADV 3.4m, record | 2026-09 | [OCC](https://www.theocc.com/newsroom/views/2026/10-02-september-2026-monthly-volume-report); [Cboe](https://www.cboe.com/insights/posts/market-metrics-that-matter-derivatives-september-2026-volume-highlights) | primary | A published statistics / B independently computed growth | | 26 | Passive share: denominator matters | Index funds 52% of long-term fund assets; domestic index funds own 19% of US equities; domestic equity ETF primary activity 5.6% of company-stock trading | 2025 year-end / full year | [ICI Fact Book, figures 2.5, 2.6, 4.3](https://www.ici.org/system/files/2026-04/2026-factbook.pdf) | primary | A | | 27 | Strategist positioning, dated rather than current consensus | Morgan Stanley mid-2027 S&P target 8,300, earnings growth assumptions 23% in 2026 / 12% in 2027; overweight equities | transcript 2026-05-15 | [Morgan Stanley](https://www.morganstanley.com/Themes/outlooks) | primary | A for stated forecast | | 28 | Hedge-fund long-share changes, **not net positioning** | Bridgewater NVDA 4,693,003→3,866,195 shares (−17.62%); CRWV 364,025→not listed. Citadel NVDA 18,664,861→11,815,467; ORCL 2,302,838→4,756,029 | 2026-03-31→2026-06-30 | [Bridgewater Q1](https://www.sec.gov/Archives/edgar/data/1350694/000135069426000002/0001350694-26-000002.txt), [Q2](https://www.sec.gov/Archives/edgar/data/1350694/000135069426000003/0001350694-26-000003.txt); [Citadel Q1](https://www.sec.gov/Archives/edgar/data/1423053/000110465926062477/0001104659-26-062477.txt), [Q2](https://www.sec.gov/Archives/edgar/data/1423053/000110465926097200/0001104659-26-097200.txt) | primary | B arithmetic | | 29 | Nvidia completed sales versus other disposals | Stevens sold 1,366,000 shares ≈$300.15m; Kress 34,918 ≈$7.65m under a trading plan. Huang's September filing reports tax withholding/gifts, not open-market sales | filings 2026-09-18/22 | [Stevens Form 4](https://www.sec.gov/Archives/edgar/data/1045810/000119903926000016/wk-form4_1790110579.xml), [Kress](https://www.sec.gov/Archives/edgar/data/1045810/000158867026000014/wk-form4_1789765959.xml), [Huang](https://www.sec.gov/Archives/edgar/data/1045810/000119764926000012/wk-form4_1789765856.xml) | primary | B aggregated sales | | 30 | Oracle completed sale versus cancelled plan | Sicilia sold 22,562 shares ≈$3.42m under a plan; Ellison cancelled his sale plan; none sold under it | 2026-09 | [Sicilia Form 4](https://www.sec.gov/Archives/edgar/data/1341439/000134143926000094/form4.xml); [cancellation 8-K](https://www.sec.gov/Archives/edgar/data/1341439/000119312526389753/d20034d8k.htm) | primary | A / B arithmetic | | 31 | CoreWeave insider sales: CFO execution and broader filing-based reports | Nitin Agrawal 22,579 shares ≈$1.81m under pre-existing plan. Reuters: CEO Intrator/affiliates sold nearly $1bn (>10m shares) since March 2025 IPO, mostly preset plans; about $40m sell-to-cover. Bloomberg via MT Newswires: co-founders sold $3.3bn since IPO. Broader reports not Form-4 reconstructed here; overlapping scopes must not be added | Agrawal 2026-10-12; reports 2026-10-16 | [Form 4](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000449/form4.xml); [Reuters via AOL](https://www.aol.com/articles/coreweaves-ceo-sold-nearly-1-164715000.html); [MT Newswires citing Bloomberg](https://vanguard.factsetdigitalsolutions.com/markets/newsStory?newsId=450e90f4d667b0a0¬ation=8941848&tabName=indexes_newsStory) | primary / secondary | B arithmetic / B filing-based reports, not reconstructed | | 32 | Broadcom completed chairman sales | Henry Samueli 702,190 shares ≈$250.01m, aggregated sale-coded lines; gifts excluded; pre-existing plan | 2026-09-23 | [Form 4](https://www.sec.gov/Archives/edgar/data/1730168/000110465926110983/tm2626134-1_4seq1.xml) | primary | B arithmetic | | 33 | Cerebras IPO: allocation versus aftermarket | Offer $185; first close $311.07 (+68.15%); October close $127.21 (−31.24% from offer, −59.11% from first close) | IPO 2026-05-14; current 2026-10-16 | [Issuer pricing](https://www.cerebras.ai/press-release/cerebras-systems-announces-pricing-of-initial-public-offering); [daily prices](https://query2.finance.yahoo.com/v8/finance/chart/CBRS?range=2y&interval=1d) | primary / secondary | A offer / C returns | | 34 | CoreWeave IPO performance | Offer and first close $40; first-day return 0%; latest $61.39, +53.48% versus offer despite drawdown in claim 15 | IPO 2025-03-28; current 2026-10-16 | [Issuer](https://www.coreweave.com/news/coreweave-announces-pricing-of-initial-public-offering); claim 15 prices | primary / secondary | A offer / C returns | | 35 | Another infrastructure IPO: Forgent Power Solutions | Offer $27; first close $29 (+7.41%); latest $31.68 (+17.33% versus offer) | IPO 2026-02-05; current 2026-10-16 | [Issuer](https://ir.forgentpower.com/news-events/press-releases/detail/89/forgent-power-solutions-announces-pricing-of-initial-public-offering); [daily prices](https://query2.finance.yahoo.com/v8/finance/chart/FPS?range=2y&interval=1d) | primary / secondary | A offer / C returns | | 36 | IPO window and lab expectations | Q2: 48 listings/$104.8bn; Q3: 30/$32.8bn, including $26.5bn SK Hynix US offering; IPO Index −11% in Q3. Earlier Anthropic Q4-2026 / OpenAI 2027 expectations remain **reported, unverified**. October 19 Reuters expects Anthropic mid-November and reports accelerated OpenAI ~$30bn talks at >$1.4tn with SoftBank/MGX; no closed round, fixed official listing date or priced IPO. Acceleration/counterparties are new; the fundraising concept and November expectation predate this window | review 2026-10-01; Reuters update 2026-10-19 | [Reuters October 19](https://www.marketscreener.com/news/openai-seeks-to-accelerate-fundraising-ahead-of-anthropic-ipo-sources-say-ce785cdbdf8cf321); [Renaissance](https://www.renaissancecapital.com/review/3Q26USReview_Public.pdf); [Anthropic confirms confidential filing, no fixed date](https://www.anthropic.com/news/confidential-draft-s1-sec) | secondary / primary filing announcement | B market data / C schedules | | 37 | Credit dispersion and CoreWeave financing | HY OAS 3.23%, IG 0.84%; CRWV five-year CDS reported intraday high 1,194.5bp on October 15, up from 452bp in June; subsequent CNBC snapshots ≈1,082bp October 19 / nearly 1,081bp October 20. Oracle five-year CDS 315bp (CNBC) / 322bp (FT) October 19, distinct unsynchronized observations, not official closes. **Added 2026-10-25:** FactSet via CNBC dated October 21 snapshots CoreWeave 1,065.84bp and Oracle 308.25bp (141bp on January 2), fixing time unspecified; Meta 114.79bp Markit October 22 (claim 64). An undated CNBC 1,145bp CoreWeave mark is excluded; Q2 revenue $2.575bn, net loss $626m, net interest $640m, debt carrying amount $35.068bn = $31.405bn recourse + $3.663bn non-recourse, excluding separately presented leases | OAS 2026-10-15; CDS 2026-10-15/19/20/21; financials 2026-06-30 | [CNBC October 22, FactSet](https://www.cnbc.com/2026/10/22/ai-debt-and-french-bonds-stress-test.html); [FRED HY](https://fred.stlouisfed.org/series/BAMLH0A0HYM2), [IG](https://fred.stlouisfed.org/series/BAMLC0A0CM); [CNBC October 19](https://www.cnbc.com/amp/2026/10/19/coreweave-oracle-ai-credit-stress.html), [October 20](https://www.cnbc.com/amp/2026/10/20/coreweave-debt-janus-henderson-lloyd.html), [FT via Financial Post](https://financialpost.com/financial-times/ai-debt-worries-spark-surge-in-default-insurance-trading); [TNW citing ICE/Bloomberg](https://thenextweb.com/news/huang-coreweave-backstop-stockholm); [10-Q](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm) | primary / secondary | A directly reported financials and OAS / B summed debt carrying amount / C reported five-year intraday CDS | | 38 | Historical warning sequence and current comparable | Ritter IPO counts/mean first-day returns: 1999 476/71.2%, 2000 380/56.3%, 2001 80/14.0%. Baa–Aaa: March 2000 0.69pp, December 0.81pp, October 2007 0.82pp, September 2026 0.44pp. New Century failed April 2007 before October equity peak; Nasdaq Composite later lost 77.9% from March 2000 to October 2002 | historical / latest 2026-09 | [Ritter table 1](https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf), [AAA CSV](https://fred.stlouisfed.org/graph/fredgraph.csv?id=AAA), [BAA CSV](https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAA), [Fed history](https://www.federalreservehistory.org/essays/subprime-mortgage-crisis), [NASDAQCOM CSV](https://fred.stlouisfed.org/graph/fredgraph.csv?id=NASDAQCOM), [JPM historical peaks](https://cdn.jpmorganfunds.com/content/dam/jpm-am-aem/global/en/insights/market-insights/guide-to-the-markets/mi-guide-to-the-markets-us.pdf) | primary / secondary | A facts / B calculations | | 39 | Illustrative unwind arithmetic, **not forecast** | Fixed initial top-ten cohort prices −77.9%, others flat, no reconstitution → index −31.63%, group weight 13.12%. Assume this cohort also has 36.1% of forward earnings; halve its earnings, hold others fixed → index forward earnings −18.05%; assumed forward P/E 19.0→15.1 → price −34.87%. Initial $100 redemption represents $40.60 top-ten exposure, not a cash-stock sales estimate | static starting weights from claim 1 | claims 1, 3, 38; [JPM pp. 4, 9](https://cdn.jpmorganfunds.com/content/dam/jpm-am-aem/global/en/insights/market-insights/guide-to-the-markets/mi-guide-to-the-markets-us.pdf); arithmetic. The 36.1% forward share is close to 36.9% implied by JPM's 20.9x/17.9x forward P/Es; 15.1x is JPM's forward P/E at the 9 October 2007 market peak, not a current fair-value estimate. Halving earnings, a flat remainder and reuse of that historical multiple are scenario assumptions | primary / secondary inputs | B conditional arithmetic | | 40 | Published implied-growth comparison, **stale model** | RBC's discounted cumulative earnings comparison (not cash distributions or a full terminal-value DCF) required Mag Seven earnings CAGR 23.8% over five years or 15.4% over ten; discount rate 6.35%, ex-group trend growth 6.19% | 2025-11-28 assumptions | [RBC GAM exhibit 20](https://www.rbcgam.com/documents/en/articles/market-outlook-new-year-2026.pdf) | primary model | A for model output, not outcome | | 41 | CoreWeave convertible pricing: financing-access counterexample | Priced $3.7bn of 2.875% convertible notes due 2033, upsized from $3bn; expected net proceeds $3.6445bn before offering expenses, less $498.8m capped calls = approximately $3.1457bn remaining without option exercise | 2026-09-18 pricing announcement | [CoreWeave issuer announcement](https://investors.coreweave.com/news/news-details/2026/CoreWeave-Prices-Upsized-3-7-Billion-Convertible-Senior-Notes-Offering/default.aspx) | primary | A terms / B subtraction | | 42 | Micron supplier profits and cash generation | Fiscal-Q4 revenue $54.23bn; GAAP net income $37.70bn; adjusted free cash flow $33.20bn; GAAP gross margin 86.8%; FQ1 FY2027 guidance: revenue $61.5bn ± $1.5bn, approximately 85.95% GAAP gross margin | quarter ended 2026-09-03; released 2026-09-30 | [Micron results](https://investors.micron.com/news/press-release/2026/Micron-Technology-Inc--Reports-Record-Fiscal-Fourth-Quarter-and-Full-Year-2026-Results/default.aspx) | primary | A reported figures | | 43 | ASML supplier results | Q3 sales €11.662bn; net income €4.960bn | Q3 2026; released 2026-10-14 | [ASML results](https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026) | primary | A | | 44 | TSMC supplier results | Q3 revenue $46.72bn; net profit margin 51.1% | Q3 2026; released 2026-10-15 | [TSMC management report](https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-10/f83780000ddc0c3b48e57a87eab6814c4ab35058/3Q26%20ManagementReport.pdf) | primary | A | | 45 | Amazon earnings quality and cash after capex | Q2 operating profit $27.5bn; net income $62.6bn includes $53.4bn pre-tax non-operating other income, primarily Anthropic investments; trailing free cash flow −$7.6bn. Inclusion of these items in each index EPS series unverified | quarter / trailing year ended 2026-06-30; released 2026-07-30 | [Amazon results](https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-Second-Quarter-Results/default.aspx); [State Street EPS methodology](https://www.ssga.com/no/en_gb/institutional/insights/mind-on-the-market-12-october-2026) | primary company / secondary index methodology | A company figures; index-normalization gap unverified | | 46 | Released inflation versus Shiller CPI estimate | September CPI-U 336.940, +3.7% YoY, +0.6% seasonally adjusted MoM; Shiller workbook September estimate 335.511. Claim 5 retains CAPE 40.58 as the published vintage and adds ≈40.41 as an approximate CPI-adjusted sensitivity | September 2026; BLS release 2026-10-14; workbook fetched 2026-10-17 | [BLS dated release](https://www.bls.gov/news.release/archives/cpi_10142026.htm); [Shiller workbook](https://shillerdata.com/) | primary | A observations / B CPI-adjustment arithmetic | | 47 | Corporate bid and implied correlation: contrary positioning evidence | $1.3tn YTD repurchase authorizations (Russell 3000); estimated 24% of S&P 500 weight outside buyback blackout windows; one-month implied correlation 7.6%. Authorizations are not executions; blackout estimates are not binding marketwide rules | 2026-10-12; published 2026-10-14 | [Citadel Securities](https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence) | secondary / compiled estimates | B | | 48 | Observable credit perimeter and loss recognition | Broad HY/IG OAS, reported CDS and traded bond marks (claims 37, 49) document dispersion and selective stress, including project paper; they do not measure the extent or absence of contagion in private credit, construction finance, collateral haircuts or overseas funding. Equity is contractually junior to debt; equity and debt prices can recognize losses concurrently | 2026-10-18 assessment of 2026-10-15 observations | [FRED HY](https://fred.stlouisfed.org/series/BAMLH0A0HYM2), [IG](https://fred.stlouisfed.org/series/BAMLC0A0CM); [reported CDS](https://thenextweb.com/news/huang-coreweave-backstop-stockholm); [CoreWeave financing disclosures](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm); claim 49 bond reports | analysis based on primary / secondary observations | A OAS / C CDS and bond-print inputs; analytical limits, not an additional measured observation | | 49 | Traded CoreWeave and project-bond stress | CoreWeave 9.75% 2031 notes yielded 15.78% (CNBC); 9.625% 2032 notes recovered from 76 to almost 81 cents, yield 14.6% (WSJ). Subsequent CNBC observations: 2032 YTM 14.9% on October 19; 2031 YTM 15.97% on October 20. Meta-linked Hyperion/Beignet project bonds just under 92 cents; CoreWeave bonds below 80 (FT via Folha). Different securities and intraday snapshots, not a single simultaneous curve. **Added 2026-10-25:** Morningstar/PitchBook (October 22) retrospectively identifies October 20 lows of 79.25 / ~15.97% for the 2031 notes and 77.5 / 15.70% for the 2032 notes; older observations newly traced, not this week's closing tape. IFR's October 23 'below 80 / above 16% in recent days' has no security/date and is not a mark | 2026-10-15; new 2026-10-19/20; retrospective 2026-10-20 reported 10-22 | [Morningstar/PitchBook, direct fetch restricted](https://www.morningstar.com/markets/coreweave-debt-tumbles-anew-korean-neocloud-ipo-tanks); [CNBC October 19](https://www.cnbc.com/amp/2026/10/19/coreweave-oracle-ai-credit-stress.html), [October 20](https://www.cnbc.com/amp/2026/10/20/coreweave-debt-janus-henderson-lloyd.html); [CNBC](https://www.cnbc.com/2026/10/15/coreweave-credit-default-swaps-hit-distress-levels-as-investors-sour-on-ai-debt.html); [WSJ](https://www.wsj.com/livecoverage/stock-market-today-dow-sp-500-nasdaq-10-15-2026/card/coreweave-bonds-extend-slide-MLE9nNEYezUmliPXS9KA); [FT via Folha](https://c-level.folha.uol.com.br/financas/2026/10/investidores-se-desfazem-de-titulos-de-data-centers-enquanto-mercado-questiona-receitas-de-ia.shtml) | secondary | C individual prints | | 50 | Daily equal-weight basket paths and 2026-peak drawdowns | M1/M1a/M1b = 100.0000 at October 16 close; October 9 = 114.57/111.68/119.78 → week −12.7%/−10.5%/−16.5%. 2026 peaks 131.26 (June 2) / 116.99 (October 7) / 176.70 (June 1) → current drawdowns −23.8%/−14.5%/−43.4%; triggers at 60 correspond to −54.3%/−48.7%/−66.0% from those peaks. M1 was 70.85 on 2025-12-31. M1 = NVDA AVGO TSM ORCL CRWV AMD MU VRT; M1a = five chip names NVDA AVGO TSM AMD MU; M1b = ORCL CRWV VRT | through 2026-10-16; register vintage 0.8b | [Basket CSV](/data/ai-basket-performance.csv), [frozen baselines](/data/forecast-observations.csv), [register](/forecasts/register); I_t = I_previous × mean(P_i,t/P_i,previous), back-cast and scaled to 100 at origin; constituent histories re-fetched from Yahoo query2 on 2026-10-18 | secondary price inputs / derived construction | B reproducible basket arithmetic; C underlying vendor prices | | 51 | Rates, spreads and financing-sensitive equity divergence | October 9→15: ten-year Treasury 5.24%→5.10% (−14bp), HY OAS 3.11%→3.23% (+12bp). October 13→14 Treasury 5.21%→5.19%, despite CPI release. October 9→16: M1b −16.5% versus M1a −10.5%. Pattern fits a credit/financing channel better than a rates shock, without identifying causality. **Added 2026-10-25:** the same pattern recurred on October 22: Treasury ten-year par 5.13%→4.96% (−17bp) while M1b fell 6.88% and M1a 5.52% (claim 61); no new HY OAS observation is added here | rates/spreads through 2026-10-15; equities through 2026-10-16; repeat check 2026-10-21→22 | [FRED DGS10 CSV](https://fred.stlouisfed.org/graph/fredgraph.csv?id=DGS10), [HY CSV](https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAMLH0A0HYM2); [Treasury par table](https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?field_tdr_date_value=2026&type=daily_treasury_yield_curve); claims 46, 50, 61 | primary rates/spreads / secondary equity inputs; analysis | A rate/spread levels / B arithmetic; channel interpretation is inference | | 52 | Elice IPO: fresh risk-appetite observation | October 20 Seoul debut closed 61,500 won against 90,500 won offer (−32.04%, reported as −32%); 201.1bn won ($151m) raised. Secondary reporting cites Reuters/company filing and Bloomberg; financing structure is not established as equivalent to CoreWeave's | 2026-10-20 | [TNW, linked original reporting](https://thenextweb.com/news/elice-ipo-32-percent-debut-korea-ai-data-centre-neocloud) | secondary | B reported prices/fundraising; B return arithmetic | | 53 | Fresh manager views differ, without identifying positioning | Griffin's October 20 warning calls AI investment almost certainly a bubble. October 21 Sewing rejects an imminent burst while allowing company setbacks; Fink argues faster datacenter construction lowers compute-cost inflation. These are attributed views, not measured returns or net exposures | remarks 2026-10-20/21 | [Bloomberg via The Edge](https://theedgemalaysia.com/node/822715) | secondary attributed opinions | B attribution | | 54 | Yardeni AI Bubble Meter, analyst model | Conditions 83/100; Triggers 34/100. Its high-conditions/low-triggers backtest spans 54 quarters since 1965: 9% subsequent two-year bear incidence and 18.3% median real gain. Historical modeled incidence is not a calibrated probability for this report | 2026-10-21 | [Yardeni original briefing archive, October 21](https://archive.yardeni.com/morning-briefing-2026/) | primary analyst model/opinion | A publication; C modeled inference | | 55 | Polymarket bubble contract: dated displays | October 21 display: 10% by December 31, 2026 (Yes 10.1¢), 22% by June 30, 2027; year-end cumulative volume $2,488,198 / whole event $3,111,645. October 18 opened display was 7%/17% (API updated October 17: 6.9%/16.5% midpoints). No verified 7→8→11% daily path or causal attribution to Griffin; current fetch did not expose full numerical resolution rules. **October 25 display:** 9% end-2026 (Yes ask 9.5¢), 23% June 30, 2027; volumes $2,496,737 / $37,684, event $3,122,417; page FAQ still says 10% for December, so the named card is preferred and the inconsistency flagged | accessed 2026-10-21; earlier snapshot 2026-10-18; latest 2026-10-25 per [signals](/signals/2026-10-25) item 65, not re-fetched in this update | [Polymarket primary page](https://polymarket.com/event/ai-bubble-burst-by); [earlier verified API/display](/signals/2026-10-18-market-views) | primary venue display | A displayed quotes/volume; historical path unverified | | 56 | Kalshi Anthropic IPO-announcement timing snapshot | 73% before December 1, 2026; 76% before January 1, 2027; 79% before February 1, 2027. These concern an official announcement, not a completed listing; the earlier IPO-first market in the October 18 log is a different contract. **October 25 display:** 68% before December 1 / 79% before January 1 / 87% before February 1 (asks 68¢/79¢/89¢), volume $4.122m; the near bucket fell 5 points while the later buckets rose | accessed 2026-10-21; latest 2026-10-25 per [signals](/signals/2026-10-25) item 67, not re-fetched (Kalshi refused connection on 2026-10-25) | [Kalshi primary contract](https://kalshi.com/markets/kxipoanthropic/anthropic-ipo/kxipoanthropic-date) | primary venue display | A displayed snapshot | | 57 | Current week's basket moves, separate extension from signals | October 16→20: daily-rebalanced M1/M1a/M1b +1.38%/+1.79%/+0.69%, levels 101.3792/101.7852/100.6948 versus frozen October 16 origin 100. Both October 19 and 20 closes available for all eight names; October 21 intraday values excluded. Prior completed October 9→16 week remains −12.7%/−10.5%/−16.5% (claim 50). NDX/SOX WTD +1.04%/+1.32% (claim 21). No forecast outcome is certified by this extension | through 2026-10-20 close; independently fetched 2026-10-21, beyond signals' closing-price verification | [Constituent close CSV](/data/ai-basket-performance.csv); [Yahoo NVDA representative vendor endpoint](https://query2.finance.yahoo.com/v8/finance/chart/NVDA?range=1mo&interval=1d); same exact ticker endpoints in CSV; claim 50 construction | secondary equity inputs / derived baskets | C underlying equity prices; B basket arithmetic | | 58 | S&P opacity warning, without a rating action | New October 19 remarks say interconnected financing can obscure liabilities; no hidden-debt total quantified and no current rating change warranted. The six-firm >$7tn 2025–30 capex / negative combined FCF at least through 2027 thesis is existing context, not new realized spending | 2026-10-19 | [Reuters via Boursorama](https://www.boursorama.com/bourse/actualites/les-principaux-hyperscalers-pourraient-avoir-un-endettement-lie-a-l-ia-plus-important-qu-on-ne-le-pense-actuellement-met-en-garde-s-p-global-83db047e4cfe79ccfdc7bad05d42e61b) | secondary reporting of agency analysis | B attribution; C projections | | 59 | Elson's valuation-location argument | Original October 20 essay puts bubble risk in labs and compares sales multiples. Its 15% CoreWeave yield / 62% implied-default references use older inputs; no verified model/recovery assumptions establish a physical default probability. Opinion does not demonstrate impairment | 2026-10-20 | [Simply Put original essay](https://edwardelson.substack.com/p/a-tale-of-two-bubbles) | primary commentary / secondary numerical assertions | A view attribution; C valuation; D default-model rendering | | 60 | October 22 US close reaction (Reuters final wrap) | S&P 500 −1.21% to 7,663.43; Nasdaq Composite −2.60% to 26,008.86; Dow +0.02%; SOX −6.7%, its largest one-day fall since June 23, more than 26% below its June 22 record close and still +51.5% YTD. Meta −7.0%, Tesla −3.1%; Intel, AMD, Sandisk and Marvell each fell more than 8%. The value-loss figures are timed estimates and not certified closing losses: Reuters' separate CDS story says SOX was "set to erase" more than $688bn that day; the Reuters company report's ~$114bn Meta loss was a 15:06 EDT conditional figure; FT's $124bn was not established as a close | 2026-10-22 close; re-fetched 2026-10-25 | [Reuters via MarketScreener](https://www.marketscreener.com/news/s-p-500-nasdaq-end-sharply-lower-as-revived-ai-doubts-ignite-tech-selloff-ce785cdedc80ff25); [Reuters CDS story](https://www.marketscreener.com/news/meta-s-credit-risk-indicator-jumps-to-record-high-after-results-ce785cdedd89fe23); [Reuters company report](https://www.marketscreener.com/news/meta-falls-as-investors-balance-muse-growth-with-rising-ai-costs-ce785cdeda88f522) | secondary | B price reporting / C value estimates | | 61 | Daily basket and NDX path, October 16–23 | M1 100.0000 → 101.2729 / 101.3792 / 102.9865 / 96.7763 / 97.0201 (Oct 19–23); M1a → 101.9455 / 101.7852 / 103.9217 / 98.1866 / 99.0949; M1b → 100.1520 / 100.6948 / 101.4292 / 94.4484 / 93.6265. October 22 one-day M1/M1a/M1b −6.03%/−5.52%/−6.88%; week to October 23 −2.98%/−0.91%/−6.37%; further falls to the 60 trigger 38.2%/39.5%/35.9%. On the claim 50 peak basis, current drawdowns are about −26.1%/−15.3%/−47.0%. Official NDX 28,957.53 on October 23: −0.79% on the week, −2.90% on October 22, −7.26% from the October 6 peak of 31,224.69 (M3 trigger 21,857.28, a further −24.5%). Recomputed here from the constituent closes; no forecast resolved | closes 2026-10-21/22/23 | [Forecast observations](/data/forecast-observations.csv); [forecast check](../../reviews/2026-10-25-forecast-check.md); [basket CSV](/data/ai-basket-performance.csv); [Nasdaq official NDX](https://indexes.nasdaq.com/Index/History/NDX) | secondary equity inputs / primary NDX / derived | C vendor prices; B basket arithmetic; A NDX levels | | 62 | October 22 equipment, neocloud and power selloff | One-day closes October 21→22: LRCX −11.76%, KLAC −9.55%, AMAT −8.16%, ASML −6.93% (US lines); Tokyo Electron −6.99% (Tokyo session, which closes before the US open); CRWV −10.06%, HUT −10.63%, CIFR −8.87%, IREN −8.37%, NBIS −7.23%; GEV −10.64%, VRT −7.81%; META −6.98%; AMD −8.10%, MU −6.26%. Week October 16→23: LRCX −12.82%, KLAC −12.15%, 8035.T −11.03%, CRWV −11.65%, CIFR −11.58%, GEV −11.53%, HUT −9.62%, META −9.40%. Closest layer-row names still need further falls of 31.2% (LRCX) and 31.7% (KLAC) to reach 0.60× the October 16 close. Lam's September-quarter release (revenue $8.219bn, +54% y/y; China 35% of revenue) is dated October 21; the timing fits a reaction but does not identify the cause | closes 2026-10-21/22/23 | [Forecast observations](/data/forecast-observations.csv), per-ticker daily paths in note field; [forecast check](../../reviews/2026-10-25-forecast-check.md) | secondary | C vendor closes / B arithmetic | | 63 | Individual-name drawdowns, October 23 | 2026-peak basis: ORCL −49.16% ($248.15 → $126.16), CRWV −60.69% ($137.98 → $54.235), VRT −48.35% ($376.23 → $194.32). Since-2025/listing basis: ORCL −61.58%, CRWV −70.46%; NVDA −11.82% from its October 6 close high. CRWV's 54.235 is a sub-penny print that may be a last sale rather than a closing-cross price | 2026-10-23 close | [Basket CSV](/data/ai-basket-performance.csv); claims 14–15, 18 peaks | secondary | C prices / B arithmetic | | 64 | Meta credit protection after Q3 results | Five-year CDS 114.79bp, +7.6bp, a record and the largest one-day jump since at least 2025 (Markit via Reuters); Bloomberg/ICE intraday high ~119.3bp, +8.9bp, a separate vendor and time. Reuters, citing S&P Global Market Intelligence, says Oracle CDS was close to a record and Amazon's near its highest since March 2020, without numbers. Investment-grade issuer; quotes are not a rating action | 2026-10-22 | [Reuters via MarketScreener, re-fetched 2026-10-25](https://www.marketscreener.com/news/meta-s-credit-risk-indicator-jumps-to-record-high-after-results-ce785cdedd89fe23); [Bloomberg via Yahoo](https://finance.yahoo.com/technology/ai/articles/meta-credit-risk-gauge-hits-125603234.html), per [signals](/signals/2026-10-25) item 29 | secondary | B reporting / C individual marks | | 65 | Dated CoreWeave and Oracle CDS snapshots | FactSet via CNBC: CoreWeave 1,065.84bp and Oracle 308.25bp on October 21 (Oracle 141bp on January 2), fixing time unspecified. A separate CNBC article, updated October 23, gives CoreWeave "last" 1,145bp and an unsecured yield of ~15% with no fixing date or security; excluded from the CSV and from this row's values | 2026-10-21; reported 2026-10-22 | [CNBC October 22](https://www.cnbc.com/2026/10/22/ai-debt-and-french-bonds-stress-test.html); [CNBC updated article](https://www.cnbc.com/2026/10/21/coreweaves-market-confidence-erodes-as-bond-yields-spike-stock-sinks.html); per [signals](/signals/2026-10-25) items 30–31; re-fetch returned HTTP 403 on 2026-10-25; [credit CSV](/data/credit-spreads.csv) | secondary | B reporting / C marks / D close attribution of undated quotes | | 66 | China listed-model and index price falls | Bloomberg (October 23, 03:51 CEST): MiniMax and Zhipu together lost about $150bn from their peaks; Zhipu −81% from its June high; MiniMax fell below its IPO price on Thursday October 22, erasing a March peak gain of up to 650%. Bloomberg cites continued token-price cuts, the capability gap to US leaders and possible DeepSeek/Moonshot listings. MiniMax then closed at HK$187.60 on October 23, +HK$26.90 / +16.74% from HK$160.70, so the below-IPO observation belongs to October 22 only. CSI 300 closed −1.8% on October 22, 20% below its June high; Hang Seng Tech fell as much as 5.3% intraday | 2026-10-22/23 | [Bloomberg via Yahoo HK, re-fetched 2026-10-25](https://hk.finance.yahoo.com/news/%E4%B8%AD%E5%9C%8Bai%E6%A8%A1%E5%9E%8B%E9%96%8B%E7%99%BC%E5%95%86%E8%82%A1%E5%83%B9%E4%BC%BC%E6%9C%AA%E8%A6%8B%E5%BA%95-minimax%E5%92%8C%E6%99%BA%E8%AD%9C%E5%B8%82%E5%80%BC%E5%85%B1%E8%B7%8C1500%E5%84%84%E7%BE%8E%E5%85%83-015155881.html); [FT/LSEG MiniMax history](https://markets.ft.markitdigital.com/data/equities/tearsheet/historical?s=100%3AHKG), per [signals](/signals/2026-10-25) item 46; [Bloomberg via The Edge, re-fetched 2026-10-25](https://theedgemalaysia.com/node/822807) | secondary | B reported closes / C approximate peak-to-date measures | | 67 | Other venue cards, October 25 | Polymarket October 2026 Fed: 96% no change / 5% +25bp (asks 96¢/5¢); Kalshi: 93% hold / 5% hike (asks 94¢/6¢). Polymarket best model end-2026: Anthropic 50% / Google 39% / OpenAI 6%, event volume $2.975m. Polymarket's Anthropic-first IPO card (95%) came from a search snapshot only and is not certified; its rules resolve 50–50 if neither lists by December 31, 2027. Continuously changing displays, not closing probabilities, and not matched to register questions | accessed 2026-10-25 per [signals](/signals/2026-10-25) items 66–68; not re-fetched in this update | [Polymarket Fed](https://polymarket.com/event/fed-decision-in-october-20260617190323537); [Kalshi Fed](https://kalshi.com/markets/kxfeddecision/fed-meeting/kxfeddecision-26oct); [Polymarket model](https://polymarket.com/event/which-company-has-best-ai-model-end-of-2026); [Polymarket IPO](https://polymarket.com/event/will-anthropic-or-openai-ipo-first) | primary venue displays | A displayed quotes / C Polymarket IPO search snapshot | | 68 | Meta Class C distribution: handled as a split in the register | Meta's board approved a one-for-one nonvoting Class C stock dividend on October 19; controlling-shareholder consent was given October 20; the charter amendment becomes effective January 19, 2027, and record/payment dates need further board action. The register treats it as a two-for-one split: the October 16 META baseline (709.33) is halved from that date, so the hyperscaler-row trigger of 425.598 becomes 212.799. Price histories here are not adjusted before the ex-date. The dividend raises no cash | 8-K filed 2026-10-21; register rule 2026-10-25 | [Meta 8-K](https://www.sec.gov/Archives/edgar/data/1326801/000162828026066700/meta-20261019.htm); [register](/forecasts/register); [forecast check](../../reviews/2026-10-25-forecast-check.md) caveat 4 | primary / register methodology | A filing / B adjustment arithmetic | Grades concern evidence, not forecast reliability. Individual-share returns and the historical Yahoo SOX return retain C pending full exchange reconciliation; claim 21 uses A-grade official Nasdaq NDX and new October 16–20 SOX EOD levels (B return arithmetic). The Codex cross-check was applied on 2026-10-17 and the independent Claude cross-check on 2026-10-18. The prior review’s figures were web-fetched and basket arithmetic recomputed on October 18; claims 21 and 57 extend closes through October 20 in the October 21 update; claims 60–68 add the October 21–23 closes and October 22–25 market, credit, China and venue observations in the October 25 update. The existing return and stress arithmetic reproduces; the review's approximately 0.9-SD valuation characterization is corrected to 0.55 SD (claim 3). ## Analysis **What ownership concentrates.** Apollo's [January deck](https://www.apolloacademy.com/wp-content/uploads/2026/01/011326-MarketConcentration.pdf) establishes the concern; the updated measurements extend it. Company aggregation, observation dates, SPY fund versus index weights and Alphabet's share classes explain differences among concentration series; the two observations are not a like-for-like concentration change. JPM's September 30 forward-P/E split, 20.9x for the top ten against their own 20.8x average and 17.9x for the rest against 16.0x, implies a 36.9% forward-earnings share (claim 1). This places the leaders' earnings and margins at the center of concentration risk; their multiples alone are near their own chart average. The Mag Seven is also an imperfect AI proxy: memory, networking, power equipment and foreign suppliers matter, while mega-caps earn substantial profits outside AI. Profitable businesses underpin the prices. Nvidia's consolidated statements support its sales valuation, and Nasdaq's research supplies a serious fundamentals-based rebuttal. Supplier profits, however, do not establish adequate returns for equipment purchasers. Demand funded by renewed financing can support suppliers until customers cut orders. CoreWeave's interest expense and losses expose that distinction. [Nvidia](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm), [Nasdaq](https://www.nasdaq.com/economic-institute/the-print/second-quarter-earnings-show-breadth), [CoreWeave](https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm). Amazon's July 30, 2026 release makes earnings quality observable: Q2 operating profit was $27.5bn, net income $62.6bn, and pre-tax non-operating other income $53.4bn, primarily from Anthropic investments; trailing free cash flow was −$7.6bn (claim 45). Operating profits, investment marks and cash after capex must be separated; subtracting a pre-tax gain from after-tax net income would not yield normalized profit. State Street says its Market EPS can use GAAP or non-GAAP figures; inclusion of these gains in each index series remains unverified. [Amazon](https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-Second-Quarter-Results/default.aspx), [State Street methodology](https://www.ssga.com/no/en_gb/institutional/insights/mind-on-the-market-12-october-2026). Fresher supplier disclosures strengthen the fundamentals rebuttal: Micron reported $33.20bn adjusted free cash flow and an 86.8% GAAP gross margin for fiscal Q4 on September 30; ASML reported €11.662bn Q3 sales and €4.960bn net income on October 14; TSMC reported $46.72bn Q3 revenue and a 51.1% net margin on October 15 (claims 42–44). Micron also guided FQ1 FY2027 revenue to $61.5bn ± $1.5bn at approximately 85.95% GAAP gross margin. These establish supplier profitability and, for Micron, cash generation; they leave purchasers' returns unresolved. The inference is that margin normalization must be a valuation sensitivity rather than treating extraordinary margins as perpetual. [Micron](https://investors.micron.com/news/press-release/2026/Micron-Technology-Inc--Reports-Record-Fiscal-Fourth-Quarter-and-Full-Year-2026-Results/default.aspx), [ASML](https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026), [TSMC](https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-10/f83780000ddc0c3b48e57a87eab6814c4ab35058/3Q26%20ManagementReport.pdf). **What prices discount.** Forward multiples depend on expected earnings; consensus requires another substantial increase after an exceptional year. CAPE smooths past real earnings and may understate structural improvements. Their disagreement alone proves neither a bubble nor safety. FactSet's 19.0x forward multiple on October 16 was unchanged from September 30; NTM and fiscal-year-one estimates also differ. The shorter FactSet averages (19.8x/19.1x over five/ten years) give a different baseline from JPM's thirty-year 17.2x and +1-SD 20.5x: 19.0x is approximately 0.55 SD above the latter mean (claim 3). FactSet's October 16 blended Q3 growth was 29.0%, with 9% reported and 84% of those beating EPS estimates (claim 8). [FactSet](https://insight.factset.com/sp-500-earnings-season-update-october-16-2026), [State Street](https://www.ssga.com/no/en_gb/institutional/insights/mind-on-the-market-12-october-2026). Neither earnings-yield spread is an expected equity risk premium: equities grow, bond principal is contractual, and the CAPE/TIPS comparison mixes dates. Small spreads nevertheless offer limited cushioning against forecast cuts. A unique implied growth rate requires margins, reinvestment, discount rates and terminal valuation. RBC's explicit exercise illustrates persistence requirements using older prices; Nvidia's sales multiple alone cannot identify growth. [RBC model](https://www.rbcgam.com/documents/en/articles/market-outlook-new-year-2026.pdf). **Performance and financing sensitivity.** Claims 11–21 summarize [the daily dataset](/data/ai-basket-performance.csv). Returns exclude dividends and currency conversion. TSM and ASML use US lines, SK Hynix domestic shares. Drawdowns compare the latest close with the highest close since the requested period began, or listing, rather than an intraday all-time high. Weekly bases use the prior Friday or preceding Korean session. The register's layer table uses the 2026-peak basis; this note's drawdown columns use the since-2025-or-listing basis. At October 16, Oracle/CoreWeave were −48.33%/−55.51% from 2026 peaks (June 1/May 6), versus −60.94%/−66.56% on the longer basis. Vertiv/SK Hynix were −45.13%/−56.42% on either basis (claims 14–15, 18, 20). The cross-section shows larger drawdowns in some financing-sensitive clouds and equipment suppliers alongside large memory-stock YTD gains. This is descriptive, not identification of what caused the moves. Index performance conceals those differences. Daily equal-weight back-casts put M1/M1a/M1b at −23.8%/−14.5%/−43.4% from their 2026 peaks, despite each being defined as 100 at the October 16 origin. Their common trigger of 60 therefore represents approximately −54.3%/−48.7%/−66.0% from those peaks (claim 50). M1b is financing-sensitive: Vertiv supplies equipment and is not itself shown here to be a distressed borrower. The chip losing streak independently matches [Bloomberg's wrap](https://www.swissinfo.ch/eng/stocks-fall-as-coreweave-selloff-fuels-ai-jitters%3A-markets-wrap/92225060). The October 14 BLS release put September CPI-U at 336.940, against 335.511 estimated in the downloaded Shiller workbook (claim 46). CAPE 40.58 remains the published vintage; substituting the released CPI gives approximately 40.41, holding other inputs fixed (claim 5). Inflation was 3.7% YoY, yet the ten-year fell from 5.24% on October 9 to 5.10% on October 15 while HY OAS widened from 3.11% to 3.23%. Over October 9–16, ORCL/CRWV/VRT's daily equal-weight basket fell 16.5% against 10.5% for the five chip names. This pattern fits a credit/financing channel better than a rates shock; the inference is descriptive, not causal, and the observation windows differ by one session (claim 51). [BLS](https://www.bls.gov/news.release/archives/cpi_10142026.htm), [Shiller](https://shillerdata.com/). **Marginal demand and positioning.** FINRA captures customer margin, not all market leverage. June 2026's $1.502072tn was the record in its January 1997–September 2026 workbook; September's $1.480054tn was 1.47% below it but 31.39% above September 2025 (claim 22). Citadel reports its own retail platform; overlapping themes cannot be added into national flows. OCC counts contracts rather than directional exposure; zero-day trading includes hedging. TQQQ establishes sizeable leveraged exposure, not industry-wide assets. [FINRA](https://www.finra.org/rules-guidance/key-topics/margin-accounts/margin-statistics), [Citadel](https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence), [OCC](https://www.theocc.com/newsroom/views/2026/10-02-september-2026-monthly-volume-report), [ProShares](https://www.proshares.com/our-etfs/leveraged-and-inverse/tqqq). Citadel's October 12 estimates also show $1.3tn YTD repurchase authorizations, only 24% of S&P 500 weight outside estimated blackout windows, and 7.6% one-month implied correlation (claim 47). Authorizations are not executed purchases and blackout estimates are not binding marketwide rules. The analytical countercase is that a returning corporate bid may cushion selling, while low priced correlation creates sensitivity to a correlation jump. [Citadel](https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence). Morgan Stanley's dated outlook and Citadel's newer commentary remain constructive, but do not establish strategist consensus or actual portfolio positioning. Sampled managers reduced Nvidia longs while Citadel increased Oracle. Holdings omit shorts, do not delta-adjust options, and precede the latest selloff. [Morgan Stanley](https://www.morganstanley.com/Themes/outlooks), [Citadel filing](https://www.sec.gov/Archives/edgar/data/1423053/000110465926097200/0001104659-26-097200.txt). Insider sales deserve monitoring when expansion depends on investor confidence. CoreWeave's sample now includes Reuters' October 16 filing-based report of nearly $1bn sold by Intrator and affiliates since IPO, mostly preset plans, alongside Bloomberg's broader $3.3bn co-founder tally relayed by MT Newswires (claim 31). These scopes overlap and cannot be added; neither total was reconstructed from Form 4s here. The samples do not establish aggregate flight: trading plans, withholding and gifts have different interpretations. Oracle's cancellation directly contradicts an indiscriminate exit narrative. [Cancellation](https://www.sec.gov/Archives/edgar/data/1341439/000119312526389753/d20034d8k.htm). **Issuance and returns.** Cerebras separates successful fundraising from aftermarket losses. CoreWeave remains above its offer despite its peak drawdown; Forgent offers a contrasting successful infrastructure listing, although electrical equipment is not exclusively AI. SK Hynix's US offering is an additional listing of an existing public issuer. Quarterly totals therefore do not count newly created AI companies. Anthropic confirms confidential submission, not a fixed launch; lab listing dates remain expectations. [Anthropic](https://www.anthropic.com/news/confidential-draft-s1-sec), [Renaissance](https://www.renaissancecapital.com/review/3Q26USReview_Public.pdf). **Credit and unwind mechanics.** The CoreWeave equity/CDS divergence discussed in [the company note](/research/stack/coreweave) is possible without either market being irrational. Equity retains residual upside; creditors need timely repayment. Dilution, refinancing and collateral priority affect them differently. CDS includes liquidity and risk premia, rather than directly measuring actuarial default frequency. Equity is contractually junior to debt, but equity and debt prices can recognize losses concurrently; margin calls and collateral terms determine when leveraged holders must reduce exposure. On September 18, 2026, CoreWeave priced $3.7bn of 2.875% convertible notes due 2033, upsized from $3bn (claim 41). This documents financing access at pricing, not confirmed settlement or continuing October access; the conversion option prevents a direct coupon comparison with straight-debt borrowing rates. The financing counterexample qualifies the distress interpretation of the reported October 15 CDS high. Subsequent traded prices sharpen the October evidence: CoreWeave's 2031 notes yielded 15.78%, its 2032 notes rebounded from 76 to almost 81 cents at 14.6%, and Meta-linked Hyperion/Beignet bonds traded just under 92 cents (claim 49). Selective distress thus appears in traded bonds and project paper as well as the reported CDS quote. [Issuer announcement](https://investors.coreweave.com/news/news-details/2026/CoreWeave-Prices-Upsized-3-7-Billion-Convertible-Senior-Notes-Offering/default.aspx). Broad US corporate OAS, reported CDS and bond marks establish dispersion and selective stress; they cannot establish the extent or absence of contagion. They do not measure private-credit marks, construction finance, collateral haircuts or all overseas funding. This is a limitation of the evidence in claims 37 and 49, recorded in claim 48, not evidence that those unobserved channels are either healthy or impaired. [FRED HY](https://fred.stlouisfed.org/series/BAMLH0A0HYM2), [IG](https://fred.stlouisfed.org/series/BAMLC0A0CM), [reported CDS](https://thenextweb.com/news/huang-coreweave-backstop-stockholm). Claim 39 separates price and earnings stress. A 77.9% price decline in a fixed initial top-ten cohort, with other prices flat and no reconstitution, produces an index decline of 31.63% and a remaining cohort weight of 13.12%. Separately, assume this cohort also represents 36.1% of forward earnings, halve those earnings, and hold other forward earnings fixed. Index forward earnings then fall 18.05%; an assumed P/E change from 19.0x to 15.1x produces a 34.87% price decline. The 36.1% forward share borrowed from trailing data is close to the 36.9% implied by JPM's 20.9x/17.9x forward P/Es. JPM records 15.1x at the October 9, 2007 market peak; reusing it as a terminal multiple, halving earnings and holding the remainder flat are scenario assumptions, not a verified current fair-value estimate (claim 39). These conditional calculations are not forecasts. Passive ownership does not mechanically accelerate this: capitalization-weighted holdings drift with prices without trading. Redemptions reduce fund exposure, but an ETF can transfer a basket in kind; neither secondary-market ETF turnover nor a redemption fixes the amount of underlying cash-market selling. Daily leveraged funds may rebalance exposure after losses. The $40.60 in claim 39 measures initial top-ten exposure in a $100 redemption, not cash-stock sales. ICI's 19% domestic index-fund ownership statistic is not a census of all passive ownership and does not support claims that passive funds own most US stocks. [ICI](https://www.ici.org/system/files/2026-04/2026-factbook.pdf). **Historical ordering.** Dot-com issuance and first-day returns warned about risk appetite; their collapse largely confirmed the turn. Broad quality spreads supplied weak early warning. In housing, deterioration in collateral and mortgage funding, then New Century's failure, preceded the equity peak. Today's comparable signals are cooling IPO aftermarkets, selective financing distress and narrow earnings dependence. Broad OAS levels alone cannot settle contagion; selective distress does not establish a systemic freeze. [Precedents](/research/theory/historical-precedents), [Ritter](https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf), [Fed history](https://www.federalreservehistory.org/essays/subprime-mortgage-crisis). ## Conflicts and open questions - **October 25 update versus the October 21 rebound:** the two-session rebound recorded on October 21 (claim 57) reversed. M1 lost 6.0% on October 22 and closed the week to October 23 at 97.02; NDX fell 2.90% that day (claims 61–62). The ten-year Treasury fell 17bp on the same day, which repeats the October 9–16 pattern of falling rates with financing-sensitive equity underperforming (claim 51). That reading stays descriptive. Strong supplier evidence coexists with the equipment selloff: Lam reported $8.219bn quarterly revenue (+54% y/y) and a preliminary 2027 WFE outlook above $200bn on October 21, the day before its shares fell 11.8%, so the price move cannot be read as weaker reported supplier results. - **Value-loss estimates versus closes:** Reuters' ~$114bn Meta loss was a 15:06 EDT conditional figure, FT's $124bn was not established as a close, and Reuters' >$688bn SOX figure was a "set to erase" estimate. None is a certified closing market-cap change; the note uses closing price returns only (claim 60). - **SOX versus NDX drawdown bases:** Reuters puts SOX more than 26% below its June 22 record close; NDX is 7.26% below its October 6 peak. The peaks differ, so the two drawdowns are not comparable stress measures, and neither is the M3 test (claims 60–61). - **MiniMax below-IPO observation versus the October 23 rebound:** Bloomberg's below-IPO statement is dated October 22; MiniMax's +16.74% close on October 23 must not be overwritten by it, and the roughly 87% peak drawdown circulating in scans is not a synchronized closing calculation (claim 66). - **Polymarket card versus FAQ:** the October 25 bubble card shows 9% by end-2026 while the same page's FAQ says 10%; the named card is used (claim 55). - **Meta Class C — open caveat closed for scoring:** the forecast check warned that the January 19, 2027 distribution could trigger the hyperscaler row mechanically. The register now treats it as a two-for-one split from that date (claim 68). Histories in this note's CSVs are unadjusted until then; a combined A+C tracking method was not adopted. - **Credit quotes versus credit events:** Meta's record CDS, CoreWeave's 1,065.84bp and Oracle's 308.25bp are pricing. No missed payment, distressed exchange, rating cut to junk or bankruptcy is recorded, and selective bank access continued (Nebius $3.5bn, TeraWulf ~$150m added committed capacity; [signals](/signals/2026-10-25) items 25–26) (claims 64–65). - **October 21 update versus a uniformly worsening market narrative:** the new two-session price extension records a modest basket/index rebound after the old week's selloff, while Elice's new debut and continuing CoreWeave/Oracle credit stress show persistent financing sensitivity (claims 21, 37, 49, 52, 57). Sewing/Fink and Yardeni offer counterviews to Griffin's warning; those opinions neither establish healthy purchaser returns nor date a crash (claims 53–54). The earlier rates-versus-credit inference in claim 51 remains tied to October 9–16, not demonstrated anew for October 19–20. - **Prediction-market comparison:** Polymarket's changed 10%/22% display replaces the 7%/17% opened display for current monitoring, with both dates retained. The October 21 fetch did not re-establish full rules, historical daily paths or a Griffin cause. Kalshi's 73%/76%/79% announcement buckets are distinct from the earlier IPO-first contract and from a completed listing. Neither venue supplies a matched probability or resolution for M1/M1a/M1b/M2/M3 or the credit forecasts (claims 55–56). - **Fundamentals and earnings quality — differing emphasis:** the original note treated profitable suppliers and compressed multiples as substantial counterevidence, with recurring cash after customers' financing and replacement costs an unresolved future test. The cross-check gives greater weight to already observed investment-related profits and negative cash after capex at Amazon (claim 45), while fresh supplier results and CoreWeave's convertible pricing strengthen the countercase (claims 41–44). Both views remain relevant: supplier cash generation is real, purchaser returns and index earnings normalization remain unresolved. - **Measurement:** current matched-period Mag Seven earnings share was not verified. Forward/trailing shares and growth contributions differ. The original note flagged an Nvidia segment/consolidated discrepancy; the cross-check did not reproduce it. In the fetched July 26, 2026 10-Q, six-month Compute & Networking revenue of $162.850bn plus Graphics revenue of $14.987bn reconciles to consolidated revenue of $177.837bn; the earlier warning is corrected. [Nvidia 10-Q](https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm). - **Stress inputs — Codex versus Claude:** Codex (2026-10-17) found the 36.1% forward share unsupported as an observation because it came from trailing data, and treated terminal 15.1x as an assumption. Claude (2026-10-18) derives a nearby 36.9% forward share from JPM's 20.9x/17.9x P/Es and identifies 15.1x as the October 9, 2007 peak multiple. Both findings are retained: the 36.1% transfer is an explicit approximation, and reusing a historical 15.1x is still a scenario choice. Claude gives the earnings/margin leg greater weight; Codex emphasized consistent denominators. Neither stress calculation is a forecast. - **NDX provenance — changed grade:** Codex retained C for exact NDX weekly returns because it reproduced Yahoo rather than exchange history. Claude fetched Nasdaq's 199 official 2026 EOD observations and recommends A. The October 18 re-fetch reproduces those levels: 29,187.5487 at October 16 and 31,224.6854 at the October 6 peak, versus Yahoo's 29,187.550781 and 31,224.470703. Both yield −5.49% for the week and −6.52% from peak at two decimals. The CSV now uses Nasdaq for NDX levels, with B arithmetic; individual equities and SOX remain C. - **Rates versus credit — Codex versus Claude:** Codex interpreted September CPI of +3.7% YoY as a contemporaneous discount-rate/cost risk. Claude observes that Treasury yields fell 14bp while HY OAS widened 12bp from October 9 to 15, with M1b underperforming M1a through October 16, favoring a financing-channel reading. The note adopts that description while retaining inflation as a macro risk; these co-movements do not identify causality (claims 46, 51). - **Long-history valuation — review arithmetic corrected:** Claude adds JPM's thirty-year context and describes 19.0x as about +0.9 SD; the published 17.2x mean and 20.5x +1-SD level imply +0.55 SD. Codex's five-/ten-year comparisons remain valid but incomplete. The top-ten average of 20.8x is a separate chart baseline and must not be confused with the index's thirty-year average (claims 1, 3). - **Coverage gaps:** complete exchange reconciliation, industry leveraged-ETF assets, current net hedge-fund exposure and aggregate insider sales remain unavailable. Form 144 notices are not executions; even Form 4 sale lines can cover taxes. Options turnover alone does not establish bullish positioning, qualifying the original Summary's risk-appetite interpretation. - **Credit/issuance — differing scope:** the original Summary judged broad credit contagion limited; the cross-check judges two broad OAS levels and one CDS quote insufficient to establish its extent or absence, especially in private credit, construction finance, collateral terms and overseas funding. Codex (2026-10-17) limited the observed conclusion to dispersion from broad OAS and CDS. Claude (2026-10-18) adds traded bond yields and project-bond prices, supporting observed selective distress beyond that earlier perimeter (claim 49); the extent of contagion remains unmeasured. Its earlier equity-first wording suggested a loss sequence; the cross-check distinguishes contractual seniority from concurrent equity/debt repricing. September convertible pricing supports financing access on that date, not continuing October access (claim 41). - **Credit-source limits:** no direct ICE CDS history was fetched. TNW's financial context differs from SEC financials; only its attributed five-year intraday quote is used. Summed carrying debt excludes separately presented leases and is distinct from principal. Neither lab's launch date nor a current national IPO first-day mean was verified. - **Valuation and timing:** the original Summary generalized forward-multiple compression; the cross-check distinguishes past compression evidence from the unchanged September 30–October 16 S&P snapshot. Shiller uses an earlier October close and estimated CPI despite the newer BLS release; Claude adds an approximate 40.41 CPI-adjusted September CAPE alongside the 40.58 published vintage retained by Codex (claims 5, 46). Earnings, holdings and margin data have distinct lags; the weekly selloff does not by itself identify AI-demand deterioration. ## Signals to watch | Signal | Current value | As of | Source | Why it matters | |--------|---------------|-------|--------|----------------| | Top-ten concentration | Companies 40.6% cap / 36.1% earnings; later SPY fund's ten largest security lines 38.63% | 2026-09-30 / 10-15 | claim 1 | Compare consistent definitions and earnings support. | | CAPE | September 40.58 published / ≈40.41 CPI-adjusted; October 40.71 uses October 2 close | workbook re-fetched 2026-10-18; September CPI released 10-14 | claims 5, 46 | Separate the published vintage from approximate inflation adjustment. | | Margin debt | $1.480054tn, +31.39% YoY, −1.47% from June's series-record $1.502072tn | 2026-09 | claim 22; FINRA workbook | Watch deleveraging from a record borrowing base. | | IPO first-day and aftermarket returns | Elice −32% debut, 61,500/90,500 won; 201.1bn won raised. Cerebras −31.24% from offer at October 16; Forgent +7.41% debut | Elice 2026-10-20; earlier comparators 10-16 | claims 33, 35, 52 | Next: subsequent closing sessions; Anthropic mid-November remains reported, unverified (claim 36); no official listing date. Separate funding raised from aftermarket returns. | | Insider execution | Sample: Stevens $300.15m, Samueli $250.01m, Sicilia $3.42m, Agrawal $1.81m; Intrator/affiliates nearly $1bn since IPO; broader co-founders $3.3bn | September–October filings; broader reports 2026-10-16 | claims 29–32 | Do not add overlapping scopes; distinguish plans, taxes and gifts. | | Individual-name drawdown | October 23: since 2025/listing NVDA −11.82%, ORCL −61.58%, CRWV −70.46%; 2026 peaks ORCL −49.16%, CRWV −60.69%, VRT −48.35%. October 16: NVDA −10.50%, ORCL −48.33%/−60.94%, CRWV −55.51%/−66.56%, VRT −45.13%, SK Hynix −56.42% | 2026-10-23; earlier 2026-10-16 | claims 11–20, 63; CSV | Keep the register's 2026-peak basis distinct from the longer basis. | | Earnings dependence and breadth | MU/NVDA 36% of expected Q3 growth; blended index growth 29.0%, 9% reported, 84% beating | 2026-10-12 / 10-16 | claims 8–9 | Track revisions alongside reporting breadth. | | Credit dispersion | Meta CDS record 114.79bp Markit / ~119.3bp ICE intraday October 22; FactSet October 21 CRWV 1,065.84bp, ORCL 308.25bp. Earlier: CRWV CDS ≈1,082bp October 19 / nearly 1,081bp October 20; 2032 YTM 14.9% October 19 / 2031 15.97% October 20; ORCL CDS 315/322bp October 19. Prior October 15 HY/IG 3.23%/0.84%, CRWV CDS high 1,194.5bp, project bonds just under 92¢ | 2026-10-15/19/20/21/22 | claims 37, 49, 58, 64–65 | Next: dated financing and rating disclosures; quotes are not defaults or synchronized closes. | | M1 baskets: current week and earlier peaks | October 23 levels 97.0201/99.0949/93.6265 (week −2.98%/−0.91%/−6.37%; October 22 −6.03%/−5.52%/−6.88%); NDX −7.26% from October 6 peak. Earlier: October 20 levels 101.3792/101.7852/100.6948 (WTD +1.38%/+1.79%/+0.69%). October 16: M1 −23.8% (peak 131.26, June 2); M1a −14.5% (116.99, October 7); M1b −43.4% (176.70, June 1); each was 100 on October 16 | current closes 2026-10-23; peak comparisons 10-16 | claims 50, 57, 61; CSV and register construction | Next: closes after the October 28–29 hyperscaler results; further falls of 38.2%/39.5%/35.9% reach the trigger of 60. | | Rates versus financing, historical comparison | Ten-year −14bp to 5.10%; HY OAS +12bp to 3.23%; M1b/M1a week −16.5%/−10.5% | October 9–15 rates/spreads; October 9–16 equities | claim 51 | Financing-channel pattern is descriptive, not causal. | | Leader valuation and margins | Top-ten/rest forward P/E 20.9x/17.9x; implied forward-earnings share 36.9%; index 19.0x versus thirty-year 17.2x | 2026-09-30 | claims 1, 3 | Test earnings/margin normalization alongside multiple sensitivity. | | Supplier guidance | Micron FQ1 FY2027 revenue $61.5bn ± $1.5bn; approximately 85.95% GAAP gross margin | released 2026-09-30 | claim 42 | Follow realized cash and margins against guidance. | | Prediction-market comparison | Polymarket bubble 9% end-2026 / 23% June-2027 (10%/22% October 21; 7%/17% October 18); Kalshi Anthropic announcement 68%/79%/87% before Dec 1/Jan 1/Feb 1 (73%/76%/79% October 21); October Fed hold 96% Polymarket / 93% Kalshi | accessed 2026-10-25 | claims 55–56, 67 | Next: dated venue snapshots; first current deadlines Dec 1 (Kalshi) / Dec 31 (Polymarket). Definitions/horizons differ from the register. | | Bubble views and model triggers | Griffin warning versus Sewing/Fink counterviews; Yardeni Conditions 83/100, Triggers 34/100 | 2026-10-20/21 | claims 53–54 | Next: subsequent primary commentary and realized financing/earnings; model scores are not this report's probabilities. | | Near-term earnings and macro calendar | FOMC October 27–28 (decision October 28); Microsoft and Alphabet October 28 (Alphabet call 4:30pm ET); Amazon October 29 after the close (call 5pm ET); Q3 advance GDP and September PCE October 29, 8:30am ET. Schedules, not actuals | checked 2026-10-25 | [Microsoft official](https://www.microsoft.com/en-us/investor/events/fy-2027/earnings-fy-2027-q1); [Alphabet announcement](https://www.marketscreener.com/news/alphabet-announces-date-of-third-quarter-2026-financial-results-conference-call-ce785dd9dc89f62d); [Amazon official](https://www.aboutamazon.com/news/company-news/amazon-earnings-q3-2026-report); [Fed calendar](https://www.federalreserve.gov/monetarypolicy/fomccalendars.htm); [BEA schedule](https://www.bea.gov/news/schedule) | Next: released earnings, capex and cash-flow guidance; the first test of C2/D3 inputs after Meta's capex raise. | | China listed-model stress | MiniMax HK$187.60 (+16.74%) October 23 after a below-IPO day on October 22; Zhipu −81% from June high; combined ~$150bn lost; CSI 300 −20% from June high | 2026-10-22/23 | claim 66 | Next: dated closes and any DeepSeek/Moonshot listing filings; keep peak-to-date estimates separate from closes. | | Meta corporate action | One-for-one Class C distribution, charter effective January 19, 2027; register halves the 709.33 baseline (trigger 425.598 → 212.799) from that date | 8-K 2026-10-21; register 2026-10-25 | claim 68 | Next: board-set record and payment dates; apply the adjustment only from the ex-date. | ## Sources Sources below were accessed for the original note and Codex cross-check on **2026-10-17**. The Claude cross-check was applied on **2026-10-18**, with fresh fetches for JPM, FactSet, Micron, BLS, Shiller, FINRA history, FRED rates/HY, Nasdaq NDX, Yahoo constituent histories and the bond/insider reports added here. CNBC was reached by direct HTTP fetch after the web tool failed; WSJ and FT/Folha supplied the quoted bond snapshots. The 2026-10-21 update re-fetched the linked prediction venues, TNW Elice reporting, Bloomberg/The Edge, Yardeni archive, Elson, Reuters/Boursorama, CNBC October 19/20, Reuters October 19 and FT/Financial Post. A separate complete eight-name close extension for October 19–20 and official Nasdaq NDX/SOX EOD were fetched on October 21; this extends the note beyond the signals entry’s price-ledger verification. The 2026-10-25 update re-fetched the Reuters October 22 close wrap and Meta CDS story, Bloomberg's China-model and CSI 300 reports, and the Treasury par-yield CSV; the CNBC FactSet article returned HTTP 403 and Kalshi refused the connection, so those values rest on the verified [October 25 signals log](/signals/2026-10-25). Equity closes for October 21–23 come from the forecast check's Yahoo spark pull (grade C) and Nasdaq's official NDX EOD. Live pages can change. SEC XML transaction sums exclude gifts and withholding unless expressly described; the broader insider reports were not reconstructed from filings in this pass. 1. J.P. Morgan Asset Management, *Guide to the Markets, U.S., Fourth Quarter 2026*, data 2026-09-30, pp. 4–5, 8–10 (re-fetched 2026-10-18): https://cdn.jpmorganfunds.com/content/dam/jpm-am-aem/global/en/insights/market-insights/guide-to-the-markets/mi-guide-to-the-markets-us.pdf 2. State Street, *SPY holdings and characteristics*, data 2026-10-15: https://www.ssga.com/us/en/individual/etfs/state-street-spdr-sp-500-etf-trust-spy 3. Apollo, Slok, Shah and Galwankar, *The extreme weight of AI in the S&P 500*, January 2026: https://www.apolloacademy.com/wp-content/uploads/2026/01/011326-MarketConcentration.pdf 4. RBC GAM, Savoie and Chornous, *High hopes for 2026*, New Year 2026, exhibits dated 2025-11-28: https://www.rbcgam.com/documents/en/articles/market-outlook-new-year-2026.pdf 5. FactSet, Butters, *S&P 500 Earnings Season Update*, 2026-10-16: https://insight.factset.com/sp-500-earnings-season-update-october-16-2026 6. Nasdaq, *Global Index Snapshot*, published 2026-10-16, data 2026-10-13: https://www.nasdaq.com/docs/global-indexes/global-markets-dashboard-oct-16-2026 7. Robert Shiller, *US Stock Price, Earnings and Dividends*, workbook downloaded from https://shillerdata.com/ ; direct file: https://img1.wsimg.com/blobby/go/e5e77e0b-59d1-44d9-ab25-4763ac982e53/downloads/1449043e-a1ed-4e2e-8e9c-7b9a2a064493/ie_data.xls?ver=1791297298731 8. Federal Reserve/FRED, Treasury and TIPS daily yields, latest observation 2026-10-15: https://fred.stlouisfed.org/series/DGS10 and https://fred.stlouisfed.org/series/DFII10 9. NVIDIA, fiscal annual/half-year consolidated statements, 2026-02-25 / 2026-08-26: https://www.sec.gov/Archives/edgar/data/1045810/000104581026000021/nvda-20260125.htm and https://www.sec.gov/Archives/edgar/data/1045810/000104581026000075/nvda-20260726.htm 10. State Street, *Strong earnings raise the bar*, 2026-10-12: https://www.ssga.com/no/en_gb/institutional/insights/mind-on-the-market-12-october-2026 11. Citadel Securities, Rubner, *October Update: Fundamentals in Focus*, 2026-10-14: https://www.citadelsecurities.com/news-and-insights/global-market-intelligence/october-update/?series=global-market-intelligence 12. Nasdaq, Mackintosh and Normyle, *Second-Quarter Earnings Show Breadth*, 2026-09-03: https://www.nasdaq.com/economic-institute/the-print/second-quarter-earnings-show-breadth 13. Yahoo Finance chart API, daily histories fetched through 2026-10-16. The cross-check used the same vendor's query2 endpoint after query1 returned HTTP 429; claims 11–21, 33 and 35 now link that fallback. Individual-share CSV rows retain their original query1 source URLs; NDX rows now use Nasdaq official EOD (source 41); representative fallback URL: https://query2.finance.yahoo.com/v8/finance/chart/NVDA?range=2y&interval=1d 14. Bloomberg via SWI, *Chip Stocks Post Longest Losing Streak Since 2008: Markets Wrap*, 2026-10-16: https://www.swissinfo.ch/eng/stocks-fall-as-coreweave-selloff-fuels-ai-jitters%3A-markets-wrap/92225060 15. FINRA, *Margin Statistics*, monthly table through September 2026: https://www.finra.org/rules-guidance/key-topics/margin-accounts/margin-statistics ; history workbook January 1997–September 2026, re-fetched 2026-10-18: https://www.finra.org/sites/default/files/2021-03/margin-statistics.xlsx 16. ProShares, *UltraPro QQQ*, daily snapshot 2026-10-16: https://www.proshares.com/our-etfs/leveraged-and-inverse/tqqq 17. OCC, *September 2026 Monthly Volume Data*, 2026-10-02: https://www.theocc.com/newsroom/views/2026/10-02-september-2026-monthly-volume-report ; Cboe, *September Volume Highlights*, 2026-10-12: https://www.cboe.com/insights/posts/market-metrics-that-matter-derivatives-september-2026-volume-highlights 18. ICI, *2026 Investment Company Fact Book*, 2026 edition: https://www.ici.org/system/files/2026-04/2026-factbook.pdf 19. Morgan Stanley, *Investing Through an Uneasy Boom*, transcript 2026-05-15 embedded in outlook hub: https://www.morganstanley.com/Themes/outlooks 20. Bridgewater and Citadel, SEC Forms 13F-HR, filed 2026-05-15 / 2026-08-14. Full submission URLs for both periods and managers are supplied in claim 28; they include the information tables. 21. NVIDIA insiders, SEC Forms 4, September 2026: https://www.sec.gov/Archives/edgar/data/1045810/000119903926000016/wk-form4_1790110579.xml ; https://www.sec.gov/Archives/edgar/data/1045810/000158867026000014/wk-form4_1789765959.xml ; https://www.sec.gov/Archives/edgar/data/1045810/000119764926000012/wk-form4_1789765856.xml 22. Oracle, Sicilia Form 4 filed 2026-09-23: https://www.sec.gov/Archives/edgar/data/1341439/000134143926000094/form4.xml ; Ellison plan cancellation, filed 2026-09-14: https://www.sec.gov/Archives/edgar/data/1341439/000119312526389753/d20034d8k.htm 23. CoreWeave, Agrawal Form 4 filed 2026-10-14: https://www.sec.gov/Archives/edgar/data/1769628/000176962826000449/form4.xml 24. Broadcom, Samueli Form 4 filed 2026-09-25: https://www.sec.gov/Archives/edgar/data/1730168/000110465926110983/tm2626134-1_4seq1.xml 25. Cerebras, *Pricing of Initial Public Offering*, 2026-05-13: https://www.cerebras.ai/press-release/cerebras-systems-announces-pricing-of-initial-public-offering 26. CoreWeave, *Pricing of Initial Public Offering*, 2025-03-27: https://www.coreweave.com/news/coreweave-announces-pricing-of-initial-public-offering 27. Forgent, *Pricing of Initial Public Offering*, February 2026: https://ir.forgentpower.com/news-events/press-releases/detail/89/forgent-power-solutions-announces-pricing-of-initial-public-offering 28. Renaissance Capital, *Third Quarter US IPO Review*, 2026-10-01: https://www.renaissancecapital.com/review/3Q26USReview_Public.pdf ; Anthropic, *Confidentially submits draft S-1*, 2026-06-01: https://www.anthropic.com/news/confidential-draft-s1-sec 29. ICE BofA/FRED, option-adjusted spreads, latest observation 2026-10-15: https://fred.stlouisfed.org/series/BAMLH0A0HYM2 and https://fred.stlouisfed.org/series/BAMLC0A0CM 30. The Next Web, Stanciuc, *Huang says CoreWeave needs no backstop*, 2026-10-15, secondary report citing Bloomberg/ICE: https://thenextweb.com/news/huang-coreweave-backstop-stockholm 31. CoreWeave, quarterly SEC financial statements, filed 2026-08-11: https://www.sec.gov/Archives/edgar/data/1769628/000176962826000366/crwv-20260630.htm 32. Jay Ritter, *Initial Public Offerings: Updated Statistics*, 2026-10-05, table 1: https://site.warrington.ufl.edu/ritter/files/IPO-Statistics.pdf 33. Moody's/Nasdaq via FRED, historical CSV downloads: https://fred.stlouisfed.org/graph/fredgraph.csv?id=AAA ; https://fred.stlouisfed.org/graph/fredgraph.csv?id=BAA ; https://fred.stlouisfed.org/graph/fredgraph.csv?id=NASDAQCOM 34. Federal Reserve History, *Subprime Mortgage Crisis*, historical essay: https://www.federalreservehistory.org/essays/subprime-mortgage-crisis 35. CoreWeave, *Prices Upsized $3.7 Billion Convertible Senior Notes Offering*, 2026-09-18: https://investors.coreweave.com/news/news-details/2026/CoreWeave-Prices-Upsized-3-7-Billion-Convertible-Senior-Notes-Offering/default.aspx 36. Micron, *Record Fiscal Fourth-Quarter and Full-Year 2026 Results*, 2026-09-30: https://investors.micron.com/news/press-release/2026/Micron-Technology-Inc--Reports-Record-Fiscal-Fourth-Quarter-and-Full-Year-2026-Results/default.aspx 37. ASML, *Q3 2026 results*, 2026-10-14: https://investor.asml.com/news-releases/news-release-details/asml-reports-eu11.7-billion-total-net-sales-and-eu5.0-billion-net-income-q3-2026 38. TSMC, *3Q26 Management Report*, 2026-10-15: https://investor.tsmc.com/english/encrypt/files/encrypt_file/reports/2026-10/f83780000ddc0c3b48e57a87eab6814c4ab35058/3Q26%20ManagementReport.pdf 39. Amazon, *Second Quarter Results*, 2026-07-30: https://ir.aboutamazon.com/news-release/news-release-details/2026/Amazon-com-Announces-Second-Quarter-Results/default.aspx 40. BLS, *Consumer Price Index — September 2026*, released 2026-10-14: https://www.bls.gov/news.release/archives/cpi_10142026.htm 41. Nasdaq, *NDX History / HistoryData*, official EOD; fetched 2026-10-18, 199 observations in 2026 (451 including 2024-12-30 through 2026-10-16 for consistent CSV bases): https://indexes.nasdaq.com/Index/History/NDX ; POST https://indexes.nasdaq.com/Index/HistoryData with id=NDX, startDate=2024-12-30, endDate=2026-10-16, timeOfDay=EOD. 42. CNBC, *CoreWeave credit-default swaps hit distress levels as investors sour on AI debt*, 2026-10-15, fetched 2026-10-18: https://www.cnbc.com/2026/10/15/coreweave-credit-default-swaps-hit-distress-levels-as-investors-sour-on-ai-debt.html 43. Wall Street Journal, Goldfarb, *CoreWeave Bonds Bounce Back After Steep Slide*, updated 2026-10-15 12:28pm ET, fetched 2026-10-18: https://www.wsj.com/livecoverage/stock-market-today-dow-sp-500-nasdaq-10-15-2026/card/coreweave-bonds-extend-slide-MLE9nNEYezUmliPXS9KA 44. Financial Times via Folha, Hodgson, Chan, Healy and Smith, *Investidores se desfazem de títulos de data centers enquanto mercado questiona receitas de IA*, 2026-10-15, fetched 2026-10-18: https://c-level.folha.uol.com.br/financas/2026/10/investidores-se-desfazem-de-titulos-de-data-centers-enquanto-mercado-questiona-receitas-de-ia.shtml 45. Reuters via AOL, Rooprai and Soni, *CoreWeave's CEO sold nearly $1 billion of stock since IPO, filings show*, 2026-10-16, fetched 2026-10-18: https://www.aol.com/articles/coreweaves-ceo-sold-nearly-1-164715000.html 46. MT Newswires, *US Equity Indexes Mixed This Week as CoreWeave Rattles AI Trade, Sparks Rotation Out of Technology*, 2026-10-16, relaying Bloomberg co-founder sales, fetched 2026-10-18: https://vanguard.factsetdigitalsolutions.com/markets/newsStory?newsId=450e90f4d667b0a0¬ation=8941848&tabName=indexes_newsStory 47. Reuters via MarketScreener, *S&P 500, Nasdaq end sharply lower as revived AI doubts ignite tech selloff*, 2026-10-22, re-fetched 2026-10-25: https://www.marketscreener.com/news/s-p-500-nasdaq-end-sharply-lower-as-revived-ai-doubts-ignite-tech-selloff-ce785cdedc80ff25 48. Reuters via MarketScreener, *Meta's credit risk indicator jumps to record high after results*, 2026-10-22 11:42 EDT, modified 14:47 EDT, re-fetched 2026-10-25: https://www.marketscreener.com/news/meta-s-credit-risk-indicator-jumps-to-record-high-after-results-ce785cdedd89fe23 49. Bloomberg via Yahoo, *Meta credit risk gauge hits record*, 2026-10-22, per signals 2026-10-25 item 29: https://finance.yahoo.com/technology/ai/articles/meta-credit-risk-gauge-hits-125603234.html 50. CNBC, *AI debt and French bonds stress test*, 2026-10-22, FactSet-dated October 21 CDS; per signals 2026-10-25 item 30 (HTTP 403 on 2026-10-25 re-fetch): https://www.cnbc.com/2026/10/22/ai-debt-and-french-bonds-stress-test.html ; CNBC updated article, 2026-10-21/23: https://www.cnbc.com/2026/10/21/coreweaves-market-confidence-erodes-as-bond-yields-spike-stock-sinks.html 51. Morningstar/PitchBook, *CoreWeave debt tumbles anew*, 2026-10-22, retrospective October 20 bond lows; direct fetch restricted: https://www.morningstar.com/markets/coreweave-debt-tumbles-anew-korean-neocloud-ipo-tanks 52. Reuters via MarketScreener, Meta company report, 2026-10-22: https://www.marketscreener.com/news/meta-falls-as-investors-balance-muse-growth-with-rising-ai-costs-ce785cdeda88f522 53. Bloomberg (Jeanny Yu) via Yahoo HK, *China AI Model Stocks' $150 Billion Drop Shows No Sign of Easing*, 2026-10-23, re-fetched 2026-10-25: https://hk.finance.yahoo.com/news/%E4%B8%AD%E5%9C%8Bai%E6%A8%A1%E5%9E%8B%E9%96%8B%E7%99%BC%E5%95%86%E8%82%A1%E5%83%B9%E4%BC%BC%E6%9C%AA%E8%A6%8B%E5%BA%95-minimax%E5%92%8C%E6%99%BA%E8%AD%9C%E5%B8%82%E5%80%BC%E5%85%B1%E8%B7%8C1500%E5%84%84%E7%BE%8E%E5%85%83-015155881.html ; FT/LSEG MiniMax price history: https://markets.ft.markitdigital.com/data/equities/tearsheet/historical?s=100%3AHKG 54. Bloomberg via The Edge, Chinese equities report, 2026-10-22, re-fetched 2026-10-25: https://theedgemalaysia.com/node/822807 55. Forecast observations and first scoring check, 2026-10-25: /data/forecast-observations.csv ; reviews/2026-10-25-forecast-check.md 56. Meta, Form 8-K on the Class C stock dividend, filed 2026-10-21: https://www.sec.gov/Archives/edgar/data/1326801/000162828026066700/meta-20261019.htm ; register corporate-action rule: /forecasts/register 57. Prediction venues read 2026-10-25 per signals 2026-10-25 items 65–68: https://polymarket.com/event/ai-bubble-burst-by ; https://polymarket.com/event/fed-decision-in-october-20260617190323537 ; https://kalshi.com/markets/kxfeddecision/fed-meeting/kxfeddecision-26oct ; https://kalshi.com/markets/kxipoanthropic/anthropic-ipo/kxipoanthropic-date ; https://polymarket.com/event/will-anthropic-or-openai-ipo-first ; https://polymarket.com/event/which-company-has-best-ai-model-end-of-2026 58. US Treasury, daily par yield curve, 2026 (CSV download fetched 2026-10-25): https://home.treasury.gov/resource-center/data-chart-center/interest-rates/TextView?field_tdr_date_value=2026&type=daily_treasury_yield_curve