Week ending 30 October 2026

Correction (4 November 2026): The key-indicators table now uses FRED and Yahoo Finance closes for 30 October, in the fixed indicator set every report carries. The original table used a newsletter’s reading of the week to 23 October. On 30 October the ICE BofA high-yield spread was 344bp, so the headline “112bp short of the Stage 4 tripwire” should read 106bp; the investment-grade spread was 94bp. Neither changes a status. The trend column now shows the change in status against the 30 October report, which was unchanged for every pillar; the pressure that built inside them is in the change section.

1Headlines?What does this mean?

Stage 3 (forced deleveraging), unchanged from the last report; earliest downgrade 23 November. AI economics: Stressed, unchanged.

2Scorecard?What does this mean?

3Change since last report?What does this mean?

Compared against the archive report of 30 October 2026, back-scored onto the eight pillars (the archive predates the scorecard format, so the baseline statuses are a reading of that report’s text rather than scores it published).

Net: no status moved. The week added pressure inside four pillars without crossing a threshold in any of them. That is what “worse inside it” on an unchanged scorecard means, and it is the pattern the series should expect most weeks: thresholds are crossed rarely, pressure builds often.

4Transmission chain?What does this mean?

1. 1 AI equity correction Confirmed
2. 2 AI-specific credit stress Confirmed
3. 3 Forced deleveraging Confirmed 23 Oct Earliest downgrade 23 Nov
4. 4 Firewall breached Not present
5. 5 Systemic liquidity event Not present

5Judgement?What does this mean?

The week’s most important fact for the AI economics axis is the Microsoft capex cut. Every prior report in this series recorded hyperscaler capex guidance being held or raised; this is the first instance of a hyperscaler trimming its own number. One data point is not a trend, and Microsoft’s explanation — third-party GPU delivery delays pushing some capacity into the first half of 2027, not a demand pullback — argues against reading it as the start of a broad retrenchment.² But it changes what a reader should watch for next: whether Microsoft’s FY2027 commentary (due at its next earnings call) shows capex resuming its prior trajectory or settling onto a lower path.

On the firewall, this week is a good illustration of why the change section carries a reason column and not just a status. The high-yield spread, at roughly 338bp in the latest full weekly reading, is inside the Calm range on the published thresholds, and the daily readings since then narrowed slightly, so against the last report the status is unchanged. But the spread has widened 72bp in about five weeks (266bp on 21 September), the fastest sustained move of the cycle, and an “unchanged” against the last report would hide that if the diff recorded status alone. The threshold hasn’t been crossed; the distance to it has roughly halved since September.

AI-specific credit remains the most unambiguous pillar: CoreWeave’s CDS near 1,195bp, Meta’s at a record 124bp, and a persistent several-hundred-basis-point premium for a basket of AI borrowers over the broad high-yield index are all evidence that lenders are pricing a specific, differentiated risk. That is discrimination, in the sense the methodology uses the word: it is not yet contagion, because the broad indices have not followed. Whether it stays that way is largely a function of whether the credit desks’ “wait for the midterms” posture — effectively a pause in high-yield AI issuance until after 3 November — resolves into renewed supply at similar prices, or into buyers staying away even after the political uncertainty clears.

6Key indicators?What does this mean?

Market data as at the close on 30 October 2026, compared with the close used by the 30 October 2026 report (29 October 2026). Reported figures are carried forward from the report that last cited them until a newer one appears.

What would change our view

Would make it worse (up toward Stage 4): US high-yield spread closing above 450bp, or investment-grade spread above 130bp, for five consecutive trading days; a hyperscaler or large-cap AI name failing to roll over maturing debt at any price; a second documented wave of hedge-fund deleveraging independent of the one already confirmed on 22–23 October.

Would make it better (down to Stage 2): A month with no new forced-deleveraging evidence (no reported margin calls, fund liquidations or prime-broker withdrawals) takes the stage from 3 to 2; the clock started on 23 October, so the earliest that can happen is 23 November. Beyond that: the already-pulled high-yield data-centre deals returning to market at similar or tighter spreads once the midterm-election uncertainty clears; CoreWeave’s CDS and bond yields retracing meaningfully from current levels without a change in its disclosed debt obligations.

Would move the AI-economics axis: Confirmation in Microsoft’s next quarterly commentary of whether the capex cut was a one-quarter supply-chain delay (as management says) or the start of a lower spending path; Alphabet’s Q4 free cash flow either recovering toward positive or deteriorating further; any hyperscaler explicitly cancelling, rather than delaying, committed AI infrastructure.

7Sources?What does this mean?

1. The sourcing process generalises cleanly. Every figure above was found via live search with its original URL in hand, which is actually easier than the archive’s reconstruction work: there was no “match this claim to a citation after the fact” step, so every source here is preserved-original-url rather than reconstructed-match.
2. The template holds up against real, messy data. The broad-credit spreads this week are a genuine edge case — moving fast but still inside the Calm band — and the scorecard’s status-plus-trend design handled it without needing a special case.
3. Carrying state forward changed the verdict. A first draft of this report, written without a baseline, called the headline Stage 2 because that was the highest stage this week’s evidence confirmed on its own. Once the 30 October archive report was back-scored as the previous state, the persistence rule did its job: Stage 3 was confirmed on 23 October and nothing has yet cleared it, so the headline is Stage 3, unchanged, with a dated earliest downgrade. That is the right answer, and it only came out of the comparison. The baseline here is back-scored from prose rather than read from a published block, and several “Previous” indicator values are approximate for the same reason; from the next report on, both come straight from the frontmatter of this one.
4. CNBC — AI turmoil: neoclouds struggle, credit risks jump, Solidigm pulls IPO (2026-10-23) — preserved-original-url
5. CNBC — Nvidia’s role as AI financier faces first big test as neoclouds take market hit (2026-10-24) — preserved-original-url
6. DCM Insider — DCM Insider Weekly, 26th October 2026 (2026-10-26) — preserved-original-url
7. LPL Research — Rate and Credit View: Financing the Future, AI’s Impact on the Bond Market (2026-10-28) — preserved-original-url
8. AIStockWire — AI borrowers’ bond spreads widened this week, our data shows (2026-10-19, underlying data as of 2026-10-16) — preserved-original-url
9. CNBC — CoreWeave’s market confidence erodes as bond yields spike, stock sinks (2026-10-21) — preserved-original-url
10. Alphabet Inc. — Q3 2026 Earnings Release (SEC Exhibit 99.1) (2026-10-28) — official-source
11. Federal Reserve Bank of St. Louis (FRED) — 10-Year Treasury Constant Maturity Rate (continuous) — official-data-series
12. Federal Reserve Bank of St. Louis (FRED) — ICE BofA US Corporate Index Option-Adjusted Spread (continuous) — official-data-series
13. Federal Reserve Bank of St. Louis (FRED) — ICE BofA US High Yield Index Option-Adjusted Spread (continuous) — official-data-series
14. Yahoo Finance — PHLX Semiconductor Index (^SOX) price history (continuous) — market-data-series
15. Yahoo Finance — Nasdaq 100 (^NDX) price history (continuous) — market-data-series
16. Yahoo Finance — KOSPI (^KS11) price history (continuous) — market-data-series
17. Federal Reserve Bank of St. Louis (FRED) — 30-Year Treasury Constant Maturity Rate (continuous) — official-data-series
18. Federal Reserve Bank of St. Louis (FRED) — Crude Oil Prices: Brent - Europe (continuous) — official-data-series
19. Federal Reserve Bank of St. Louis (FRED) — Nominal Broad U.S. Dollar Index (continuous) — official-data-series
20. Federal Reserve Bank of St. Louis (FRED) — Japanese Yen to U.S. Dollar Spot Exchange Rate (continuous) — official-data-series
21. Federal Reserve Bank of St. Louis (FRED) — ICE BofA CCC & Lower US High Yield Index Option-Adjusted Spread (continuous) — official-data-series
22. Federal Reserve Bank of St. Louis (FRED) — ICE BofA CCC & Lower US High Yield Index Effective Yield (continuous) — official-data-series

Note on this report: what the repeatability test did and didn’t prove

This report was built from scratch — live web search against real published sources, dated 23–30 October 2026 — without referring back to the earlier 31-report archive, as a test of whether the format and sourcing process described in part 0 could be executed independently rather than inherited from the original ChatGPT-produced material. Three honest findings from doing that:

1. The sourcing process generalises cleanly. Every figure above was found via live search with its original URL in hand, which is actually easier than the archive’s reconstruction work: there was no “match this claim to a citation after the fact” step, so every source here is preserved-original-url rather than reconstructed-match.
2. The template holds up against real, messy data. The broad-credit spreads this week are a genuine edge case — moving fast but still inside the Calm band — and the scorecard’s status-plus-trend design handled it without needing a special case.
3. Carrying state forward changed the verdict. A first draft of this report, written without a baseline, called the headline Stage 2 because that was the highest stage this week’s evidence confirmed on its own. Once the 30 October archive report was back-scored as the previous state, the persistence rule did its job: Stage 3 was confirmed on 23 October and nothing has yet cleared it, so the headline is Stage 3, unchanged, with a dated earliest downgrade. That is the right answer, and it only came out of the comparison. The baseline here is back-scored from prose rather than read from a published block, and several “Previous” indicator values are approximate for the same reason; from the next report on, both come straight from the frontmatter of this one.

Not investment advice. Scoring, tripwires and the persistence rule are set out in How to read the AI Market Watch; terms are explained in the explainers.