To The Moon

Executive Summary

Observation. US stocks fell sharply on 5 November, led by chipmakers. The S&P 500 lost 4.3%, the Nasdaq 5.5% and the semiconductor index 12% [1] (high).

Finding. Forced selling and profit-taking explain much of the drop, but the cause is contested (medium). Credit stress is visible in a few names and is not yet systemic. Data center delivery is a second constraint. AI revenue is still growing.

Call to Action. Watch three things: the October FINRA margin print (due in the third week of November) [6], AI credit spreads, and the 19 November data center loan auction [19]. Review each holding against its written thesis before acting on price.

What happened

The Nasdaq-100 lost $2.3 trillion and the VIX closed near 28, up about 70% [1] (high). Retail investors sold a net $358 million of individual stocks, the biggest day since March 2020 (Vanda Research, via [1]) (medium-high). A widely shared chart put the day’s US loss at $3.8 trillion, with chipmakers $1.15 trillion of it [2]. The direction matches our check [1], but we could not find the total independently (medium). A separate viral claim of $13.2 trillion lost since 2026 highs [3] is unverified (low). It mixes each stock’s own peak date.

Two explanations

Goldman’s trading desk pointed to forced selling from crowded, leveraged positions, with hedge fund gross exposure in the 97th percentile over five years [4]. Nomura said funds were locking in 2026 profits rather than reacting to weaker fundamentals [1]. The other view is political. After the Democratic midterm sweep, UBS estimated that about 30% of expected 2028 US data center capacity additions could be at risk from construction pauses [5]. We cannot separate the two yet (medium).

Margin calls

No consolidated US margin-call figure exists yet. FINRA is the source of record for margin debt [6]. It fell to $1.417 trillion in July from a record $1.502 trillion in June, the largest monthly drop on record [9] (high). It then rebounded to $1.454 trillion in August [8] and $1.480 trillion in September [7], about 1.5% below the June record (high). The September print predates the sell-off, so little leverage had come out beforehand. In an earlier episode, Goldman reported 1.2 million leveraged retail accounts in South Korea margin-called, about 3.4% of adults [10] (medium). On 6 November, Korean and Taiwanese futures were down at least 7% [4].

Market value lost by US stocks on 5 Nov. Source: Social Capital post, TradingView data. Chipmakers in dark red.

Off-balance-sheet obligations of five US tech firms, trillions of dollars. Source: Nikkei research, as shown in a Global Markets Investor post. Some figures are estimates.

Debt and credit

Five-year default swaps on CoreWeave hit a record 1,194.5 basis points on 15 October (ICE Data Services, via [11]) (high). Oracle’s hit a record 288 [12] (high). A record does not mean equal distress. Off-balance-sheet commitments reached $3.2 trillion for five firms at end-September, up 90% from March [13]. WSJ counted about $3 trillion across nine firms against roughly $600 billion of trailing capex [14]. The scopes differ (medium). These are commitments, not debt issued. Perplexity cites an Octus tracker putting investment-grade AI infrastructure debt issued above $700 billion [20] (low-medium, not checked).

Banks and Blackstone will provide $63 billion for an SPV that buys Broadcom chips and leases them to Anthropic [15, 16] (high). Anthropic makes Claude, one of our cross-validation models, so read that item with this in mind. Amazon’s 10-Q discloses a secured facility to borrow at its discretion from Q1 2027 through Q3 2028 [17]. Reporting puts the ceiling at about $80 billion [18]. It sits on the balance sheet and is undrawn (high). About $500 million of data center loans are up for auction, with bids due 19 November [19] (high). That is small, so we read it as a price signal, not a stress event.

Perplexity’s research also cites about $300 billion of chip-backed SPVs from a single secondary source (low-medium), and adds Asia and Europe. Taiwan bank loans of about $126 billion for January to July, China AI capex of about $139 billion, and European tech bonds of $39.3 billion year to date [20]. We have not checked these (low-medium). The pattern is bank-led in Asia and bond-led in the US.

Bloomberg headline and AI summary on the loan auction, as shown in a user-supplied screenshot. Bids are due 19 November.

Reading the auction. The loans include debt tied to operators such as Vantage and CyrusOne. The seller is anonymous and no prices are published. The result on or after 19 November will show whether buyers will hold AI infrastructure paper at par, at a discount, or not at all.

Delivery risk

At least 47 large US data center projects were delayed, withdrawn or blocked in 2026, according to RegLayer, which calls its count conservative (reported by the FT) [21] (high). Data Center Watch counts more, at least 75 projects in Q1 [22] and 45 in Q2 [23], under a different definition (medium). Heatmap Pro put about $130 billion of cancellations in Q3, which a secondary summary says exceeded all of 2025. That is planned dollars, not capacity [24] (low-medium). Bernstein judges about 35% of the planned pipeline credible [21, 25] (medium-high). Dan Ives estimates 10% of the 800 to 1,000 planned sites will not happen. This is a sell-side view, and his range has moved between 5% and 15% [26] (low as a forecast).

At CoreWeave, contracted power was about 4.2 gigawatts as of 11 August [27] against 1.5 gigawatts online [28], so about 2.7 gigawatts is contracted but not yet built. The dates differ and “offline” overstates it (medium-high). Management says moratoriums change where capacity is built, not demand [29]. Delivery lag is the risk that connects the debt and the demand sides of this snapshot.

Returns and slop

Anthropic’s run-rate topped $65 billion at end-July [30]. OpenAI told investors about $50 billion at end-September, unaudited [31]. Run-rates are not audited revenue (medium). S&P 500 earnings are up nearly 40% year on year, though not all of that is AI (Apollo’s chief economist, via [1] and [32]) (medium). On slop, Pangram found one in four long-form social posts fully AI-generated [33, 34], and Graphite found most AI articles do not surface in search [35]. The evidence shows trust and attention erosion in consumer channels, not an enterprise revenue loss (medium-low). The debt is underwritten mainly on enterprise compute contracts, so the link between slop and the debt side is indirect.

Reviewing our holdings

A price drop is not a thesis change, and we separate the two before acting on either. For each holding, we write down the value proposition in one sentence, the long-term objective it serves, and the evidence that would show the thesis is broken. We then test this week’s news against that list. Falling prices, wider credit spreads and forced selling [1, 4, 10, 11] tell us about price and positioning. Evidence that a thesis may have changed looks different: delivery lag of the kind at CoreWeave [27, 28], a pipeline Bernstein judges only 35% credible [21], or commitments that outrun revenue [13, 14]. If none of our written break conditions has been met, holding is a decision we can defend. If one has, we say so and revisit the objective, the time horizon and the position size (judgment, not data).

Holding is not the same as ignoring. We ask three questions of each position. Would we buy it today at this price, knowing what we know now? Does it depend on debt, financing or delivery timing that this snapshot shows under strain? Is it sized so that a further fall like 5 November, when the semiconductor index dropped 12% in a day [1], would not force us to sell?

Hedging. A hedge reduces exposure at a cost, and the cost is higher now. The VIX closed near 28 after a jump of about 70% [1], and higher volatility generally makes options more expensive. We do not have current quotes. An index hedge does not protect a single stock, and a hedge loses money if the market recovers. Availability and cost differ by market and broker in Asia and the US, so we check them before relying on one (low confidence on specifics).

Leverage. We avoid it. This week shows why. Leveraged ETFs reset daily and can lose value in choppy markets [36], their rebalancing can add to selling when prices fall [37], and margin calls force sales at the worst time [4, 10]. A leveraged holder can be forced out of a sound thesis by price alone.

Affordability. We invest only what we can afford to lose, and we keep enough cash for the next few years of needs so that we are never forced to sell into a fall. This is a personal rule, not a forecast. On 6 November, Korean and Taiwanese futures were down at least 7% [4], and the memory and chip names most tied to the AI trade sit in those markets. For Asian holdings we also check currency exposure, since home-currency returns depend on both price and exchange rate (low confidence on direction).

This section describes our review process. It is not a recommendation to buy, sell or hedge any security.

What we do not know

Total margin-call volume, the October FINRA print, the main cause of the drop, and how much of the delayed pipeline is permanent.

References

Source tiers: P = primary or original source. R = reputable secondary reporting. A = aggregator, blog, social post or AI research file. Where a primary source is listed, we read the figure there or at the secondary source cited with it.

[1] (R) Yahoo Finance, Stock market today, 5 Nov 2026. finance.yahoo.com/markets/live/stock-market-today-thursday-november-5-dow-sp-500-nasdaq-plunge-chip-stocks-melt-down-193726491.html

[2] (A) Social Capital on X, 6 Nov 2026 (user-supplied screenshot; TradingView data).

[3] (A) Bull Theory on X (user-supplied screenshot).

[4] (R) MarketIndex, ASX 200 Live, 6 Nov 2026 (Goldman trading desk). marketindex.com.au/news/asx-200-live-today-friday-6th-november

[5] (R) The American Bazaar, 5 Nov 2026 (UBS estimate, citing Reuters). americanbazaaronline.com/2026/11/05/nasdaq-faces-1-8-trillion-selloff-risk-as-ai-stocks-grapple-with-regulatory-uncertainty-after-midterms-491636/

[6] (P) FINRA, Margin Statistics (source of record, Rule 4521(d)). finra.org/investors/margin-statistics

[7] (R) TokenPost, U.S. margin debt rises $26.2B in September (FINRA data). tokenpost.com/news/investing/35483

[8] (R) Advisor Perspectives, Margin Debt Increases 2.6% in August (FINRA data). advisorperspectives.com/dshort/updates/2026/09/17/margin-debt-finra-august-2026

[9] (A) thetrading.tools, Margin debt fell $84.8B in July (FINRA data). thetrading.tools/chart-of-the-day/2026-09-10-margin-debt-record-drop

[10] (R) Yahoo Finance, Margin & Leverage: Inside the Violent Reset of the AI Trade (Goldman). finance.yahoo.com/markets/stocks/articles/margin-leverage-inside-violent-reset-164500093.html

[11] (R) The Next Web, CoreWeave credit swaps at a record (ICE Data Services), 15 Oct 2026. thenextweb.com/news/huang-coreweave-backstop-stockholm

[12] (R) Reuters via KSL, AI credit default swap spreads widen sharply. ksl.com/article/51636602/ai-credit-default-swap-spreads-widen-sharply

[13] (R) TechNews, Big Tech AI off-balance-sheet commitments, 4 Nov 2026 (chart credited to Nikkei research). technews.tw/2026/11/04/tech-giants-off-balance-sheet-ai-commitments-surge/

[14] (A) AI Weekly summary of WSJ analysis. aiweekly.co/alerts/wsj-nine-big-tech-firms-carry-3t-in-ai-related-off-balance-sheet-commitments

[15] (P) Broadcom Investor Relations, $63 billion financing structure release, 26 Oct 2026. investors.broadcom.com/node/64786/pdf

[16] (R) Reuters via KSL, Wall St banks, Blackstone line up $63 billion. ksl.com/article/51641301/wall-st-banks-blackstone-line-up-63-billion-to-finance-broadcom-ai-chips-for-anthropic

[17] (P) Amazon Form 10-Q, quarter ended 30 Sep 2026 (Loan Facility note). sec.gov/Archives/edgar/data/0001018724/000101872426000040/amzn-20260930.htm

[18] (R) Bloomberg via The Edge Malaysia, 31 Oct 2026. theedgemalaysia.com/node/824042

[19] (A) AI Stock Wire, $500M data center loan auction (summarising Bloomberg). aistockwire.com/blog/data-center-loans-auction-500-million-spacex-spcx-ai-debt-november-2026

[20] (A) Perplexity research file, ‘AI Debt Leverage and SPVs: A Country-Level Assessment’, 6 Nov 2026. Not independently checked.

[21] (R) Financial Times via Techmeme, RegLayer 47+ projects and Bernstein 35%, 18 Oct 2026. techmeme.com/261018/p6#a261018p6

[22] (P) Data Center Watch, Q1 2026 report summary. datacenterwatch.org/q1-2026-1

[23] (R) Trending Topics, Data Center Watch Q2 2026. trendingtopics.eu/us-data-center-opposition-q2-2026/

[24] (R) Random Lengths, Heatmap Pro cancellation tally, 7 Oct 2026. randomlengthsnews.com/archives/2026/10/07/data-center-cancellations/77896

[25] (A) AI Tools Recap, Bernstein 35% credible pipeline. aitoolsrecap.com/Blog/bernstein-35-percent-data-center-pipeline-credible-2026

[26] (R) Asianet Newsable, Dan Ives on data center projects. newsable.asianetnews.com/markets/dan-ives-says-10-of-planned-u-s-data-centers-will-not-happen-calls-democratic-sweep-a-seismic-shift-for-the-ai-buildout-articleshow-syhxp0k

[27] (P) CoreWeave Form 424B5 (SEC), contracted power of about 4.2 GW. sec.gov/Archives/edgar/data/0001769628/000162828026062362/coreweave-424b5.htm

[28] (R) Tom’s Hardware, CoreWeave contracted versus online power. tomshardware.com/tech-industry/coreweave-ceo-mike-intrator-says-it-has-signed-an-a100-contract-running-into-2029

[29] (R) Blockspace, CoreWeave shifts data center sites amid moratoriums. blockspace.media/insight/coreweave-shifts-data-center-sites-amid-moratoriums/

[30] (R) Channel Insider, Anthropic run-rate, 18 Aug 2026. channelinsider.com/news-and-trends/news-anthropic-revenue-run-rate-outpaces-openai/

[31] (A) Implicator.ai, OpenAI run-rate, 9 Oct 2026. implicator.ai/openai-puts-revenue-run-rate-near-50-billion-sets-70-billion-year-end-target/

[32] (P) Apollo, Daily Spark, 3 Nov 2026 (earnings and nominal GDP; read via [1]). apollo.com/wealth/insights-news/insights/daily-spark/10-percent-nominal-gdp-growth-coming

[33] (R) AdNews, Pangram analysis of about 1M posts, 13 Jul 2026. adnews.com.au/news/linkedin-found-in-the-deep-end-of-the-ai-slop-pool

[34] (P) Pangram, AI in your feed (read via [33]). pangram.com/blog/ai-in-your-feed

[35] (R) TechRadar, Graphite study of AI-written articles. techradar.com/ai-platforms-assistants/the-internet-is-now-mostly-written-by-machines-study-finds

[36] (A) StockWireX, What $1.4 Trillion in Margin Debt Is Warning Investors (leveraged ETF daily reset). stockwirex.com/analysis/margin-debt-warning-signal/

[37] (A) Crypto Briefing, Leverage driving US stock rally sparks unease (leveraged ETF rebalancing). cryptobriefing.com/leverage-driving-us-stock-rally-unease/

AI cross-validation credits: research inputs from Perplexity, Grok and Kimi. Review by Grok and Kimi. Verification and drafting by Claude (built by Anthropic, a financing counterparty above).

@TigerStars

$Vanguard S&P 500 ETF(VOO)$

$Cboe Volatility Index(VIX)$