1HeadlinesWhat does this mean?
Stage 3 (forced deleveraging), unchanged from the last report; earliest downgrade 11 October. AI economics: Watch, unchanged.
The bad
- Large technology companies were reported to support up to roughly $300bn of AI infrastructure exposure through residual-value guarantees and special-purpose financing structures.1
- Examples included large commitments associated with Meta, Broadcom and Nvidia-linked projects.
- This came immediately after the $18bn Project Jupiter financing problem, where debt was trading materially below par and banks were unable to distribute as much as intended.2
The good
- The important firewall was still holding: no broad corporate-credit dislocation, widespread margin calls or hyperscaler capex cancellation had yet appeared.3
Not what it looks like
- The $300bn is contingent exposure, not losses. If accelerator and data-centre residual values remain high, the structures work as intended; the guarantees only bite if AI demand disappoints and the equipment has to be sold into a distressed market.
2ScorecardWhat does this mean?
| Pillar | Status | Trend | Why |
|---|---|---|---|
| AI equities | Watch | Unchanged | No new equity evidence in this report, and the same close as the 19 September report: the semiconductor index was 18.5% below its 52-week high.4 The report notes the guarantees would bite when tech earnings and valuations are already under pressure. |
| Rates and macro | Elevated | Unchanged | Treasury yields around 5% are one of the three parts of the report’s fragility case; the 10-year closed at 5.01% on 18 September, inside the Elevated band.7 |
| Broad credit (the firewall) | Calm | Unchanged | No broad corporate-credit dislocation.3 Investment-grade spread 77bp and high-yield 268bp, both well inside Calm.312 |
| AI-specific credit | Elevated | Unchanged | No new AI credit pricing in this report. Project Jupiter’s debt was still trading materially below par.2 Status held from the 19 September report. The guarantees add a route for losses to travel back to the technology companies themselves. |
| AI financing flow | Elevated | Unchanged | Big Tech was reported to support up to roughly $300bn of AI infrastructure exposure through residual-value guarantees and special-purpose vehicles, with large commitments associated with Meta, Broadcom and Nvidia-linked projects.1 More of the build-out is being funded by SPV and debt investors on the strength of guarantees, rather than straight from Big Tech cash, and this follows banks failing to distribute the Jupiter debt.2 |
| Leverage and liquidity | Elevated | Unchanged | No widespread margin calls had appeared.3 No new forced-selling evidence in this report; status held from the 19 September report. |
| Private credit | Elevated | Unchanged | No new private-credit evidence in this report; status held from the 19 September report. |
| AI economics | Watch | Unchanged | No hyperscaler capex cancellation had yet appeared.3 The risk the report identifies is economic: if AI demand disappoints and rapidly obsolete equipment has to be sold into a distressed market, the guarantees move losses back to technology companies precisely when their earnings and valuations are already under pressure. |
3Change since last reportWhat does this mean?
Compared against the 19 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 (19 Sep) | This report | Change | |
|---|---|---|---|
| Stage | 3 (evidence 11 Sep) | 3 (evidence 11 Sep) | Unchanged. No new forced-selling evidence; the 11 September entry evidence still holds, so the earliest downgrade is 11 October. |
| AI economics | Watch | Watch | Unchanged. No new capex or cash-flow evidence. The guarantee disclosure shows more of the downside sits with the hyperscalers than their balance sheets suggest, but nothing has been cancelled.1 |
| AI equities | Watch | Watch | Unchanged. No new evidence; same close as the last report. |
| Rates and macro | Elevated | Elevated | Unchanged. No new evidence; same close as the last report. |
| Broad credit | Calm | Calm | Unchanged. No broad corporate-credit dislocation; same closing spreads as the last report.3 |
| AI-specific credit | Elevated | Elevated | Unchanged. No new pricing; Project Jupiter’s debt is still materially below par.2 |
| AI financing flow | Elevated | Elevated | Unchanged. Worse inside it: up to roughly $300bn of AI infrastructure exposure is reported to sit in guarantee and SPV structures, on top of the Jupiter syndication failure in the last report.12 |
| Leverage and liquidity | Elevated | Elevated | Unchanged. No widespread margin calls.3 |
| Private credit | Elevated | Elevated | Unchanged. No new evidence. |
Net: no status moved. The report adds structural fragility rather than a new market event: the $300bn of contingent exposure builds inside an unchanged Elevated status for AI financing flow, and the market data are the same closes as the last report. The firewall held.
4Transmission chainWhat does this mean?
- 1 AI equity correction Confirmed
- 2 AI-specific credit stress Partial evidence
- 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 | Confirmed. Tripwire met on 3 September and not yet cleared. The index is back within 20% of its high (18.5% below on 18 September) but has not yet closed inside the line for five consecutive days.4 |
| 2 — AI-specific credit stress | Partial. Project Jupiter’s debt trading materially below par, with banks unable to distribute it, is still the one independent sign.2 The $300bn of guarantees is a financing structure, not credit stress in itself, so the two-sign test is not yet met.1 |
| 3 — Forced deleveraging | Confirmed. Entry evidence confirmed from the 13 September report: JPMorgan curbed lending to Situational Awareness after its margin call and forced liquidation, reported 11 September. No widespread margin calls in this report and no new forced selling; earliest downgrade 11 October.3 |
| 4 — Firewall breached | Not present. No broad corporate-credit dislocation: investment-grade 77bp and high-yield 268bp, far below the 130bp and 450bp tripwires.312 |
| 5 — Systemic liquidity event | Not present. No funding-market malfunction reported. |
5JudgementWhat does this mean?
This was a meaningful increase in structural fragility: contingent and off-balance-sheet AI financing exposure is larger than previously appreciated. Up to roughly $300bn of AI infrastructure exposure supported through residual-value guarantees and special-purpose structures changes the transmission mechanism from simply:
Big Tech cash → data centre
toward:
SPV/debt investors → AI infrastructure → Big Tech guarantee → apparently limited balance-sheet liability.1
If accelerator and data-centre residual values remain high, this works. If AI demand disappoints and rapidly obsolete equipment has to be sold into a distressed market, the guarantees can move losses back toward technology companies precisely when their earnings and valuations are already under pressure. The timing makes this harder to wave away: it came immediately after the $18bn Project Jupiter financing problem, where debt was trading materially below par and banks were unable to distribute as much as intended.2
The important firewall was still holding: no broad corporate-credit dislocation, widespread margin calls or hyperscaler capex cancellation had yet appeared.3 The combination of $300bn of contingent AI infrastructure exposure + an actual troubled AI financing + ~5% Treasury yields materially increased structural fragility, but still did not amount to a general credit crisis.
6Key indicatorsWhat does this mean?
Market data as at the close on 18 September 2026, compared with the close used by the 19 September 2026 report (18 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) | 11,922; 18.5% below its 52-week high (22 Jun) | 11,922 | ±0.0% | Stage 1 when more than 20% below its 52-week high for five closes | Confirmed, Yahoo Finance ^SOX4 |
| Nasdaq 100 | 29,644 | 29,644 | ±0.0% | Context | Confirmed, Yahoo Finance ^NDX5 |
| KOSPI | 6,894 | 6,894 | ±0.0% | Context | Confirmed, Yahoo Finance ^KS116 |
| Rates and macro | |||||
| US 10-year Treasury | 5.01% | 5.01% | ±0.00pp | Elevated 4.75–5.25%; Stressed above 5.25%, each for five closes | Confirmed, FRED DGS107 |
| US 30-year Treasury | 5.34% | 5.34% | ±0.00pp | Context | Confirmed, FRED DGS308 |
| Brent crude | $119.66 | $119.66 | ±0.00 | Context | Confirmed, FRED DCOILBRENTEU9 |
| US dollar index (broad) | 119.51 | 119.51 | ±0.0% | Context | Confirmed, FRED DTWEXBGS10 |
| Yen per US dollar | ¥156.87 | ¥156.87 | ±0.00 | Context | Confirmed, FRED DEXJPUS11 |
| Broad credit | |||||
| ICE BofA US investment-grade spread | 77bp | 77bp | ±0bp | Watch at 100bp; Stage 4 above 130bp for five days | Confirmed, FRED BAMLC0A0CM3 |
| ICE BofA US high-yield spread | 268bp | 268bp | ±0bp | Watch at 350bp; Stage 4 above 450bp for five days | Confirmed, FRED BAMLH0A0HYM212 |
| ICE BofA CCC & lower spread | 1,083bp | 1,083bp | ±0bp | Early warning only | Confirmed, FRED BAMLH0A3HYC13 |
| ICE BofA CCC & lower effective yield | 15.59% | 15.59% | ±0.00pp | Early warning only | Confirmed, FRED BAMLH0A3HYCEY14 |
| AI-specific credit | |||||
| AI investment-grade premium over the broad index | Not yet reported in the series | — | — | n/a | — |
| 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 | Not yet reported in the series | — | — | n/a | — |
| 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 18 Sep.
FRED DGS10, daily closes to 18 Sep.
FRED BAMLC0A0CM, daily closes to 18 Sep.
FRED BAMLH0A0HYM2, daily closes to 18 Sep.
Faint vertical lines mark each of the last 9 reports. Dashed red lines are the published thresholds.
What would change our view
Would make it worse (up toward Stage 4): A second, independent sign of AI-specific credit stress within a month, such as a persistent premium on AI borrowers’ bonds or another AI deal failing to syndicate, would confirm Stage 2 beneath the headline. A fall in accelerator resale values that triggers payments under the guarantees would bring the losses back onto Big Tech balance sheets. Broad spreads above the Stage 4 tripwires for five consecutive trading days would breach the firewall.
Would make it better (down to Stage 2): A month with no new forced-selling evidence takes the stage off 3; the clock runs from 11 September, so the earliest that can happen is 11 October. Project Jupiter’s debt being placed with investors or trading back above 90 cents would ease the AI-credit picture.
Would move the AI-economics axis: AI demand disappointing badly enough to push down residual values on accelerators and data centres, or any hyperscaler capex cancellation, would move it worse; residual values holding up would keep the guarantees theoretical.
7SourcesWhat does this mean?
- Financial Times — Big Tech uses guarantees to keep $300bn of AI exposure off balance sheets (2026-09-20) —
reconstructed-match - Reuters — Oracle’s $18 billion data center debt under pressure, FT reports (2026-09-18) —
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.