1HeadlinesWhat does this mean?
Stage 2 (AI-specific credit stress), unchanged from the last report; earliest downgrade 8 November. AI economics: Watch, unchanged.
The bad
- South Korea’s KOSPI fell 7.4%, with Samsung Electronics down 7.7% and SK Hynix down 8.8%. The Korean market was approximately 25% below its July highs.4
- Broadcom CEO Hock Tan said the biggest prospective constraint on AI development was capital availability, rather than semiconductor or memory shortages.5
- His estimates suggested the industry was spending around $200bn a year generating AI inference tokens against $150bn in associated revenue. These are management estimates rather than independently audited industry accounts.6
- The IMF highlighted that hyperscalers now account for approximately 10% of global investment-grade bond issuance, compared with less than 1% in 2024.12
The good
- The systemic-credit threshold has still not been crossed. Broad spreads remained relatively contained, and there was no confirmed widespread capex cancellation or margin-call cascade.3
2ScorecardWhat does this mean?
| Pillar | Status | Trend | Why |
|---|---|---|---|
| AI equities | Watch | Unchanged | The semiconductor index closed 16.4% below its 52-week high on 12 October, inside the Watch band.7 The following day the KOSPI fell 7.4%, with Samsung Electronics down 7.7% and SK Hynix down 8.8%, leaving Korea about 25% below its July highs.4 |
| Rates and macro | Elevated | Unchanged | US 10-year at 5.24%, inside the Elevated band.10 No rates commentary in the report. |
| Broad credit (the firewall) | Calm | Unchanged | Investment-grade spread 83bp and high-yield 311bp, both inside Calm; the report finds broad spreads relatively contained.315 |
| AI-specific credit | Stressed | Unchanged | No new AI-specific credit pricing in this report; Stressed held from the 11 October report. The IMF’s finding that hyperscalers are now about 10% of global investment-grade issuance raises the stakes of any repricing.12 |
| AI financing flow | Elevated | Unchanged | Broadcom’s Hock Tan named capital availability, not chip or memory supply, as the biggest prospective constraint on AI development, new evidence that financing could limit expansion.5 Status held from the 11 October report. |
| Leverage and liquidity | Elevated | Unchanged | No margin-call cascade confirmed.3 No new leverage evidence in this report; status held from the 11 October report. |
| Private credit | Elevated | Unchanged | No private-credit evidence in this report; status held from the 11 October report. |
| AI economics | Watch | Unchanged | Hock Tan’s estimates put industry spending on generating AI inference tokens at around $200bn a year against $150bn of associated revenue. These are management estimates rather than audited accounts, but they reinforce concern that expansion depends on whether customer revenues justify the capital. No confirmed widespread capex cancellation.63 |
3Change since last reportWhat does this mean?
Compared against the 11 October 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 (11 Oct) | This report | Change | |
|---|---|---|---|
| Stage | 2 (evidence 8 Oct) | 2 (evidence 8 Oct) | Unchanged. Stage 2 holds, entry evidence last confirmed on 8 October (Firmus pulling its IPO). No new forced-selling evidence, and no margin-call cascade.3 |
| AI economics | Watch | Watch | Unchanged. Worse inside it: Hock Tan’s estimates of $200bn spending against $150bn revenue on AI inference, though these are unaudited management figures.6 |
| AI equities | Watch | Watch | Unchanged. Worse inside it: the semiconductor index is 16.4% below its high, against 14.1% at the last report’s close, and Korean chip stocks sold off heavily the following day.74 |
| Rates and macro | Elevated | Elevated | Unchanged. The 10-year was flat at 5.24%.10 |
| Broad credit | Calm | Calm | Unchanged. High-yield and investment-grade unchanged at 311bp and 83bp.153 |
| AI-specific credit | Stressed | Stressed | Unchanged. No new pricing evidence, but the IMF’s issuance figures widen the route by which AI credit stress could reach mainstream bond portfolios.12 |
| AI financing flow | Elevated | Elevated | Unchanged. Worse inside it: a leading chip CEO now names capital, not hardware, as the binding constraint.5 |
| Leverage and liquidity | Elevated | Elevated | Unchanged. No new evidence. |
| Private credit | Elevated | Elevated | Unchanged. No new evidence. |
Net: no status moved. Pressure built inside AI equities, financing flow and AI economics, and the IMF’s issuance figures widened the potential transmission route from an AI slowdown into mainstream bond portfolios. The risk of transition increased, but the transition itself was not demonstrated.
4Transmission chainWhat does this mean?
- 1 AI equity correction Partial evidence
- 2 AI-specific credit stress Confirmed 8 Oct Earliest downgrade 8 Nov
- 3 Forced deleveraging Partial evidence
- 4 Firewall breached Partial evidence
- 5 Systemic liquidity event Not present
| Stage | Status |
|---|---|
| 1 — AI equity correction | Partial. The semiconductor index is 16.4% below its 52-week high, still within the 20% tripwire, though the KOSPI’s 7.4% fall the next day shows the correction deepening.74 |
| 2 — AI-specific credit stress | Confirmed. Confirmed from 22 September (Project Jupiter failing to syndicate on 18–19 September and the AI investment-grade premium on 22 September). Last confirmed by Firmus pulling its IPO, reported 8 October; earliest downgrade 8 November. |
| 3 — Forced deleveraging | Partial. Stage 3 lapsed in the 11 October report, a month after the 11 September forced-selling evidence. No margin-call cascade is confirmed in this report.3 |
| 4 — Firewall breached | Partial. High-yield at 311bp and CCC at 1,253bp are wide but broad spreads remain relatively contained, well short of the 450bp and 130bp tripwires.15163 |
| 5 — Systemic liquidity event | Not present. No funding-market malfunction reported. Hyperscalers’ growing share of global investment-grade issuance is a route by which an AI slowdown could reach mainstream bond portfolios.12 |
5JudgementWhat does this mean?
Alert: meaningful deterioration in semiconductor markets, accompanied by new evidence that financing, not hardware availability, could constrain AI expansion.
The selloff moved to Korea: the KOSPI fell 7.4%, Samsung Electronics 7.7% and SK Hynix 8.8%, leaving the market about 25% below its July highs.4 At the same time Broadcom’s Hock Tan said the biggest prospective constraint on AI development was capital availability rather than semiconductor or memory shortages.5 His estimates of roughly $200bn a year spent generating AI inference tokens against $150bn in associated revenue are management estimates rather than independently audited industry accounts, but they reinforce concern that continued AI expansion increasingly depends on whether customer revenues justify the capital required.6
The IMF’s finding that hyperscalers now account for approximately 10% of global investment-grade bond issuance, against less than 1% in 2024, materially increases the potential transmission route from an AI investment slowdown into mainstream bond portfolios.12
The systemic-credit threshold has still not been crossed. Broad spreads remained relatively contained and there was no confirmed widespread capex cancellation or margin-call cascade.3 The risk of transition had increased, but the transition itself was not yet demonstrated.
6Key indicatorsWhat does this mean?
Market data as at the close on 12 October 2026, compared with the close used by the 11 October 2026 report (9 October 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) | 12,237; 16.4% below its 52-week high (22 Jun) | 12,572 | −2.7% | Stage 1 when more than 20% below its 52-week high for five closes | Confirmed, Yahoo Finance ^SOX7 |
| Nasdaq 100 | 30,665 | 30,883 | −0.7% | Context | Confirmed, Yahoo Finance ^NDX8 |
| KOSPI | 6,384 | 6,626 | −3.6% | Context | Confirmed, Yahoo Finance ^KS119 |
| Rates and macro | |||||
| US 10-year Treasury | 5.24% (9 Oct) | 5.24% | ±0.00pp | Elevated 4.75–5.25%; Stressed above 5.25%, each for five closes | Confirmed, FRED DGS1010 |
| US 30-year Treasury | 5.60% (9 Oct) | 5.60% | ±0.00pp | Context | Confirmed, FRED DGS3011 |
| Brent crude | $129.95 | $130.03 | −0.08 | Context | Confirmed, FRED DCOILBRENTEU12 |
| US dollar index (broad) | 121.60 (9 Oct) | 121.60 | ±0.0% | Context | Confirmed, FRED DTWEXBGS13 |
| Yen per US dollar | ¥158.34 (9 Oct) | ¥158.34 | ±0.00 | Context | Confirmed, FRED DEXJPUS14 |
| Broad credit | |||||
| ICE BofA US investment-grade spread | 83bp | 83bp | ±0bp | Watch at 100bp; Stage 4 above 130bp for five days | Confirmed, FRED BAMLC0A0CM3 |
| ICE BofA US high-yield spread | 311bp | 311bp | ±0bp | Watch at 350bp; Stage 4 above 450bp for five days | Confirmed, FRED BAMLH0A0HYM215 |
| ICE BofA CCC & lower spread | 1,253bp | 1,253bp | ±0bp | Early warning only | Confirmed, FRED BAMLH0A3HYC16 |
| ICE BofA CCC & lower effective yield | 17.36% | 17.35% | +0.01pp | Early warning only | Confirmed, FRED BAMLH0A3HYCEY17 |
| 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 | almost 100bp (1 Oct report) | almost 100bp | no new reading | n/a | Carried from the 1 October report |
| 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 | Beignet bonds (Meta Hyperion, $27bn) ~91c, ~7.55% yield (1 Oct report) | Beignet bonds (Meta Hyperion, $27bn) ~91c, ~7.55% yield | no new reading | n/a | Carried from the 1 October report |
Yahoo Finance ^SOX, daily closes to 12 Oct.
FRED DGS10, daily closes to 9 Oct.
FRED BAMLC0A0CM, daily closes to 12 Oct.
FRED BAMLH0A0HYM2, daily closes to 12 Oct.
Faint vertical lines mark each of the last 19 reports. Dashed red lines are the published thresholds.
What would change our view
Would make it worse (up toward Stage 3): Broad high-yield above 450bp or investment-grade above 130bp for five consecutive trading days, particularly as hyperscaler issuance now makes up a much larger share of the investment-grade market; a confirmed margin-call cascade or other forced selling, which would put Stage 3 back.
Would make it better (down to Stage 1): A month with no new sign of AI-specific credit stress and named AI debt back above 90 cents; the earliest that can happen is 8 November.
Would move the AI-economics axis: Confirmed widespread capex cancellation; independent evidence on whether AI inference revenue is catching up with the spending behind it.
7SourcesWhat does this mean?
- International Monetary Fund — Global Financial Stability Report, October 2026 — Resilience, Risk, and Transformation (2026-10-13) —
official-source - Reuters — IMF cites AI slowdown, spiking yields as looming global market risks (2026-10-13) —
reconstructed-match - Federal Reserve Bank of St. Louis (FRED) — ICE BofA US Corporate Index Option-Adjusted Spread (continuous) —
official-data-series - NBC News / Associated Press — South Korea’s Kospi index sinks more than 7% on semiconductor share sell-off over AI boom worries (2026-10-13) —
reconstructed-match - CNBC — Intel, Broadcom CEOs warn capital crunch could become AI bottleneck (2026-10-12) —
reconstructed-match - Computer Weekly — Capital, not chips, may be the real brake on the AI build-out (2026-10-13) —
reconstructed-match - 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.