1HeadlinesWhat does this mean?
Stage 3 (forced deleveraging), unchanged from the last report; earliest downgrade 11 October. AI economics: Watch, unchanged.
The bad
- The US 10-year Treasury reached 5.225%, a 19-year high, after rising roughly 20bp in two days. The 30-year reached 5.502%, its highest since 2004.1
- The rate shock was becoming global, with long-dated yields rising sharply across Japan, Australia and Europe while the dollar strengthened.
- Large bond investors were moving towards shorter-duration and higher-quality assets and becoming explicitly more selective about AI debt.
- Regulators were examining large trading firms’ exposures following July’s AI and semiconductor deleveraging.
The good
- Still no fresh evidence of widespread AI capex cancellation, another major data-centre financing failure, multiple new margin calls, or broad corporate-credit contagion.3
- The firewall held: the investment-grade spread closed at 79bp and high-yield at 280bp, both inside Calm.12
- The semiconductor index closed within 20% of its 52-week high for the fifth session running on 24 September, which clears the Stage 1 tripwire. It finished 14.6% below the high.4
2ScorecardWhat does this mean?
| Pillar | Status | Trend | Why |
|---|---|---|---|
| AI equities | Watch | Unchanged | The semiconductor index closed 14.6% below its 52-week high on 24 September, inside the Watch band and the fifth consecutive close within 20% of the high.4 The report itself has no new equity evidence. |
| Rates and macro | Elevated | Unchanged | The 10-year closed at 5.18% and the 30-year at 5.47%, inside the Elevated band but 7bp short of Stressed on the 10-year.78 The report records the 10-year touching 5.225%, a 19-year high, after a roughly 20bp rise in two days, and the 30-year at 5.502%, its highest since 2004.1 Long-dated yields were rising sharply across Japan, Australia and Europe and the dollar was strengthening. |
| Broad credit (the firewall) | Calm | Unchanged | Investment-grade spread 79bp and high-yield 280bp, both inside Calm, and no broad corporate-credit contagion.312 |
| AI-specific credit | Elevated | Unchanged | No new AI credit pricing in this report; status held from the 23 September report. The report counts a widening AI-specific credit premium as one input to sharply higher required returns on AI infrastructure, and bond investors were becoming explicitly more selective about AI debt. |
| AI financing flow | Elevated | Unchanged | No deals priced or pulled in this report; status held from the 23 September report. Enormous forthcoming AI debt supply and already-stressed individual data-centre financings meet bond investors moving to shorter-duration, higher-quality assets, but there is no fresh evidence of another major data-centre financing failure. |
| Leverage and liquidity | Elevated | Unchanged | Regulators were examining large trading firms’ exposures following July’s AI and semiconductor deleveraging, but there is no fresh evidence of multiple new margin calls. Status held from the 23 September report. |
| Private credit | Elevated | Unchanged | No new evidence in this report; status held from the 23 September report. |
| AI economics | Watch | Unchanged | No fresh evidence of widespread AI capex cancellation. The pressure in this report is on the cost side: a 5.2% 10-year, a widening AI premium and heavy debt supply together mean sharply higher required returns on AI infrastructure. |
3Change since last reportWhat does this mean?
Compared against the 23 September 2026 report, back-scored the same way. Scores for this report were back-scored on 4 November 2026 from its own text and from FRED and Yahoo Finance closing data, because it predates the scorecard format.
| Last report (23 Sep) | This report | Change | |
|---|---|---|---|
| Stage | 3 (evidence 11 Sep) | 3 (evidence 11 Sep) | Unchanged. Stage 3 still rests on JPMorgan curbing lending to Situational Awareness after its margin call and forced liquidation, reported 11 September. No new forced-selling evidence in this report, so the earliest downgrade is 11 October. |
| AI economics | Watch | Watch | Unchanged. No capex cancellation evidence in either report; this one adds pressure on required returns from the rates side. |
| AI equities | Watch | Watch | Unchanged. The 24 September close was the fifth in a row within 20% of the high, which clears the Stage 1 tripwire, although the drawdown widened slightly to 14.6%.4 |
| Rates and macro | Elevated | Elevated | Unchanged. Worse inside it: the 10-year rose roughly 20bp in two days to a 19-year high of 5.225% and the shock spread to Japan, Australia and Europe.1 It closed at 5.18%, 7bp short of Stressed.7 |
| Broad credit | Calm | Calm | Unchanged. High-yield 280bp and investment-grade 79bp, a little wider than the last report’s close but well inside Calm.12 |
| AI-specific credit | Elevated | Elevated | Unchanged. No new AI credit pricing. The last report’s SoftBank CDS reading and $11.1bn high-yield bond are not updated here. |
| AI financing flow | Elevated | Elevated | Unchanged. No new deals. The last report noted that SoftBank’s bond sale completed with good demand; this one notes bond investors growing more selective about AI debt. |
| Leverage and liquidity | Elevated | Elevated | Unchanged. Regulators were examining trading firms’ exposures after July’s deleveraging, but no new margin calls were reported. |
| Private credit | Elevated | Elevated | Unchanged. No new evidence in either report. |
Net: no pillar status moved and the stage held at 3. The movement was in rates, where the 10-year climbed towards the Stressed threshold and the shock went global, and in the chain, where Stage 1 fell back to partial as the semiconductor index met its five-close clearing condition. The last report’s AI credit pressure was carried rather than extended.
4Transmission chainWhat does this mean?
- 1 AI equity correction Partial evidence
- 2 AI-specific credit stress Confirmed
- 3 Forced deleveraging Confirmed 11 Sept Earliest downgrade 11 Oct
- 4 Firewall breached Not present
- 5 Systemic liquidity event Not present
| Stage | Status |
|---|---|
| 1 — AI equity correction | Partial. The semiconductor index closed within 20% of its 52-week high for the fifth consecutive session on 24 September, clearing the tripwire first met on 3 September. It is still 14.6% below the high, a Watch-level drawdown, so the stage is partial rather than absent.4 |
| 2 — AI-specific credit stress | Confirmed. Confirmed from 22 September: Project Jupiter failing to syndicate on 18–19 September plus the AI investment-grade premium recorded on 22 September. This report adds a widening AI-specific credit premium and bond investors growing explicitly more selective about AI debt. |
| 3 — Forced deleveraging | Confirmed. Entry evidence 11 September: JPMorgan curbing lending to Situational Awareness after its margin call and forced liquidation. No multiple new margin calls in this report; regulators were examining trading firms’ exposures after July’s deleveraging. Earliest downgrade 11 October. |
| 4 — Firewall breached | Not present. Investment-grade 79bp and high-yield 280bp, far below the 130bp and 450bp tripwires, and no broad corporate-credit contagion.312 |
| 5 — Systemic liquidity event | Not present. The global duration shock is a repricing of government bonds, not a malfunction in funding markets; no repo stress, redemption gates or official intervention were reported. |
5JudgementWhat does this mean?
This report’s meaningful deterioration is in the rates and liquidity backdrop.2 The US 10-year reached 5.225%, a 19-year high, after rising roughly 20bp in two days, and the 30-year reached 5.502%, its highest since 2004.1 The rate shock was also becoming global, with long-dated yields rising sharply across Japan, Australia and Europe while the dollar strengthened. That crossed an important threshold, because the AI financing equation now looked like this:
10-year Treasury 5.2% + widening AI-specific credit premium + enormous forthcoming AI debt supply + already-stressed individual data-centre financings = sharply higher required returns on AI infrastructure.
Large bond investors were moving towards shorter-duration and higher-quality assets and becoming explicitly more selective about AI debt, and regulators were examining large trading firms’ exposures following July’s AI and semiconductor deleveraging. But there was still no fresh evidence of widespread AI capex cancellation, another major data-centre financing failure, multiple new margin calls, or broad corporate-credit contagion.3
So the assessment moves from “AI-specific credit deterioration with a hostile rates backdrop” to “AI-specific credit deterioration plus a genuine global duration shock.” The headline stays at 3 on the September forced-selling evidence. The one improvement is in equities: the semiconductor index has cleared the Stage 1 tripwire, so the price leg of the chain is easing even as the cost of money rises.4
6Key indicatorsWhat does this mean?
Market data as at the close on 24 September 2026, compared with the close used by the 23 September 2026 report (22 September 2026). Reported figures are carried forward from the report that last cited them until a newer one appears.
| Indicator | Latest | Previous | Change | Tripwire | Source |
|---|---|---|---|---|---|
| AI equities | |||||
| PHLX Semiconductor Index (SOX) | 12,493; 14.6% below its 52-week high (22 Jun) | 12,690 | −1.6% | Stage 1 when more than 20% below its 52-week high for five closes | Confirmed, Yahoo Finance ^SOX4 |
| Nasdaq 100 | 30,479 | 30,732 | −0.8% | Context | Confirmed, Yahoo Finance ^NDX5 |
| KOSPI | 7,081 (23 Sep) | 7,018 | +0.9% | Context | Confirmed, Yahoo Finance ^KS116 |
| Rates and macro | |||||
| US 10-year Treasury | 5.18% | 4.96% | +0.22pp | Elevated 4.75–5.25%; Stressed above 5.25%, each for five closes | Confirmed, FRED DGS107 |
| US 30-year Treasury | 5.47% | 5.29% | +0.18pp | Context | Confirmed, FRED DGS308 |
| Brent crude | $120.92 | $114.89 | +6.03 | Context | Confirmed, FRED DCOILBRENTEU9 |
| US dollar index (broad) | 120.55 | 119.59 | +0.8% | Context | Confirmed, FRED DTWEXBGS10 |
| Yen per US dollar | ¥158.92 | ¥157.55 | +1.37 | Context | Confirmed, FRED DEXJPUS11 |
| Broad credit | |||||
| ICE BofA US investment-grade spread | 79bp | 77bp | +2bp | Watch at 100bp; Stage 4 above 130bp for five days | Confirmed, FRED BAMLC0A0CM3 |
| ICE BofA US high-yield spread | 280bp | 268bp | +12bp | Watch at 350bp; Stage 4 above 450bp for five days | Confirmed, FRED BAMLH0A0HYM212 |
| ICE BofA CCC & lower spread | 1,112bp | 1,075bp | +37bp | Early warning only | Confirmed, FRED BAMLH0A3HYC13 |
| ICE BofA CCC & lower effective yield | 16.04% | 15.48% | +0.56pp | Early warning only | Confirmed, FRED BAMLH0A3HYCEY14 |
| AI-specific credit | |||||
| AI investment-grade premium over the broad index | ~115bp vs 78bp broad IG (22 Sep report) | ~115bp vs 78bp broad IG | no new reading | n/a | Carried from the 22 September report |
| Oracle 5-year CDS | Not yet reported in the series | — | — | n/a | — |
| Meta 5-year CDS | Not yet reported in the series | — | — | n/a | — |
| CoreWeave 5-year CDS | Not yet reported in the series | — | — | n/a | — |
| SoftBank 5-year CDS | ~385bp (from ~280bp in June) (23 Sep report) | ~385bp (from ~280bp in June) | no new reading | n/a | Carried from the 23 September report |
| Named data-centre project debt | Project Jupiter (~$18bn) quoted 89–91c (19 Sep report) | Project Jupiter (~$18bn) quoted 89–91c | no new reading | n/a | Carried from the 19 September report |
Yahoo Finance ^SOX, daily closes to 24 Sep.
FRED DGS10, daily closes to 24 Sep.
FRED BAMLC0A0CM, daily closes to 24 Sep.
FRED BAMLH0A0HYM2, daily closes to 24 Sep.
Faint vertical lines mark each of the last 12 reports. Dashed red lines are the published thresholds.
What would change our view
Would make it worse (up toward Stage 4): A 10-year close above 5.25%, which would make rates Stressed; another major data-centre financing failure or multiple new margin calls; broad corporate-credit contagion, with high-yield above 450bp or investment-grade above 130bp for five consecutive trading days.
Would make it better (down to Stage 2): A month with no new forced-selling evidence after the Situational Awareness margin call and liquidation reported on 11 September, so not before 11 October, provided Stage 2’s own evidence is still confirmed.
Would move the AI-economics axis: Evidence of widespread AI capex cancellation, or of AI infrastructure projects being delayed because the higher required returns no longer clear.
7SourcesWhat does this mean?
- Reuters — The 5% threshold (2026-09-24) —
reconstructed-match - Reuters — SoftBank raises $11.1 billion in world’s biggest high-yield corporate bond sale (2026-09-24) —
reconstructed-match - Federal Reserve Bank of St. Louis (FRED) — ICE BofA US Corporate Index Option-Adjusted Spread (continuous) —
official-data-series - Yahoo Finance — PHLX Semiconductor Index (^SOX) price history (continuous) —
market-data-series - Yahoo Finance — Nasdaq 100 (^NDX) price history (continuous) —
market-data-series - Yahoo Finance — KOSPI (^KS11) price history (continuous) —
market-data-series - Federal Reserve Bank of St. Louis (FRED) — 10-Year Treasury Constant Maturity Rate (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — 30-Year Treasury Constant Maturity Rate (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — Crude Oil Prices: Brent - Europe (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — Nominal Broad U.S. Dollar Index (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — Japanese Yen to U.S. Dollar Spot Exchange Rate (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — ICE BofA US High Yield Index Option-Adjusted Spread (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — ICE BofA CCC & Lower US High Yield Index Option-Adjusted Spread (continuous) —
official-data-series - Federal Reserve Bank of St. Louis (FRED) — ICE BofA CCC & Lower US High Yield Index Effective Yield (continuous) —
official-data-series
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.