US AI Cycle Monitor — text / AI audit

Machine- and human-readable snapshot of every dashboard metric. No JavaScript; safe to read with curl. Structured JSON twin: /api/ai-audit. Interactive dashboard: /.

Last updated: 2026-07-08T23:20:06.050Z · Data through: 2026-07-08 US close · Snapshot provenance: live · Audit generated: 2026-07-09T13:05:32.275Z

1. Current cycle regime

84.2/100 composite fragility

RED · Stage 3 — Distribution / stress-test (held 36 sessions) · confidence high

Distribution / stress-test: composite fragility 84.2/100 (red), easing over the past month.

The composite fragility score stands at 84.2 — distribution-risk territory (-5.1 points over 20 sessions). That includes a +12.0 synchronization add-on — multiple warnings are lining up at once, which is the pattern that matters more than any single signal. The cycle read is Stage 3 — Distribution / stress-test (held 36 sessions): Tape distribution elevated (tape 72) with composite 84 — selling pressure into a still-standing trend. The equal-weight AI basket sits 8.9% below its one-year high.

What is driving it: issuance activity is heavy: 59 weighted registration/offering filings across the universe in the past quarter (score 100). Also, leverage and speculation are elevated: FINRA margin debt is up 54% year-over-year (monthly, ~4-week publication lag); universe dollar-volume runs 62% above its one-year norm (score 96). Also, trend extension is stretched — the AI basket trades 32.9% above its 200-day average, with a 103% twelve-month return (score 92).

What is not confirming: rates and credit are not the problem — spreads and yields sit inside normal ranges (40).

What would change this read:

2. Composite fragility score

Composite score84.2 (RED)
Base (before synchronization add-on)72.2
Synchronization add-on applied+12.0
Bandsgreen <35 · amber 35–65 · red >65
Pillars available6 / 6
Pillar weightsExtension 10% · Tape: distribution & breadth 25% · Leverage & speculation 15% · Rates & credit 20% · Earnings reaction 20% · Supply & funding 10%
FormulaAvailability-reweighted weighted mean of the six pillar scores (weights below), plus a synchronization add-on: +8 if ≥3 pillars score ≥65, +4 if ≥4 pillars score ≥50, capped at 100. Bands: green <35 · amber 35–65 · red >65.

3. Pillar scores (summary table of all dashboard cards)

PillarScoreBandWeightSourceData throughStatus
Extension92.1RED10%Yahoo Finance — daily adjusted prices (equal-weight AI basket)2026-07-08live
Tape: distribution & breadth72.0RED25%Yahoo Finance — daily prices & volume (SOXX, QQQ, SMH, universe members)2026-07-08live
Leverage & speculation95.5RED15%FINRA margin statistics (monthly, ~4-week lag) + Yahoo Finance dollar-volume2026-05-01live
Rates & credit39.8AMBER20%FRED (DGS10, DFII10, ICE BofA HY & IG OAS) + Yahoo ^TNX for the live 10-year yield2026-07-08live
Earnings reaction63.7AMBER20%SEC EDGAR (8-K item 2.02 results filings) + Yahoo Finance next-session returns2026-06-24live
Supply & funding99.6RED10%SEC EDGAR (S-1 / S-3 / 424B* / FWP issuance filings)2026-06-24live

4. Every metric — raw value, score, formula

Extension — 92.1 (RED)

Weight 10% · Source: Yahoo Finance — daily adjusted prices (equal-weight AI basket) · Data through: 2026-07-08
How stretched the AI trade is versus its own trend.

IndicatorRaw valueScore (pctl 0–100)CalculationNote
AI basket vs 200-day average32.9%90basket ÷ 200-day average − 1, then percentile-ranked against its own history up to that date
12-month basket return102.8%93252-session total return of the equal-weight basket, percentile-ranked
3-month trend slope0.005495OLS slope of the basket over 63 sessions ÷ mean level, percentile-ranked

Caveats: Equal-weighted basket of the designated universe; SMH/SOXX act as cap-weighted proxies.

Tape: distribution & breadth — 72.0 (RED)

Weight 25% · Source: Yahoo Finance — daily prices & volume (SOXX, QQQ, SMH, universe members) · Data through: 2026-07-08
Selling pressure and breadth in the tape. This is a price/volume proxy — it does NOT observe who is selling (no institutional-flow feed connected).

IndicatorRaw valueScore (pctl 0–100)CalculationNote
Distribution days (SOXX+QQQ, 25 sessions)1584count of sessions in the last 25 where SOXX or QQQ fell >0.2% on higher volume than the prior day (split-day-guarded), percentile-ranked
Near-high breadth divergence0.0047when the basket is within 2% of its 252-day high: (0.7 − share of members above their 50-day average) ÷ 0.7, else 0; percentile-ranked0 = none, 1 = severe
Equal-weight AI vs SMH, 3-month slope-0.000581OLS slope of (equal-weight AI ÷ SMH) over 63 sessions, inverted so a falling ratio (narrowing leadership) = more fragile
% of universe above 200-day average67.5%70share of universe members above their own 200-day average, inverted (weak breadth = more fragile)

Caveats: Measures selling pressure in the tape; does not observe who is selling. · Institutional flow (13F aggregation / prime-broker / fund flows): no feed connected.

Leverage & speculation — 95.5 (RED)

Weight 15% · Source: FINRA margin statistics (monthly, ~4-week lag) + Yahoo Finance dollar-volume · Data through: 2026-05-01
Broad-market leverage growth and universe turnover intensity.

IndicatorRaw valueScore (pctl 0–100)CalculationNote
FINRA margin debt, YoY growth53.7%FINRA debit balances: latest month ÷ same month one year earlier − 1, percentile-rankedmonthly, ~4-week publication lag
FINRA margin debt, 6-month growth16.6%FINRA debit balances: latest month ÷ six months earlier − 1, percentile-ranked
Universe dollar-volume vs 1-year norm62.5%9621-session average universe dollar-volume ÷ 252-session average − 1, percentile-ranked

Caveats: Margin debt is market-wide, monthly, published ~4 weeks after month end — not AI-specific. · Options positioning (put/call, 0DTE): no reliable free feed — not scored.

Rates & credit — 39.8 (AMBER)

Weight 20% · Source: FRED (DGS10, DFII10, ICE BofA HY & IG OAS) + Yahoo ^TNX for the live 10-year yield · Data through: 2026-07-08
Discount-rate and funding-cost pressure on a long-duration, capital-intensive trade.

IndicatorRaw valueScore (pctl 0–100)CalculationNote
10y yield, 3-month change (pp)-388 bp010-year Treasury yield (DGS10) minus its level 63 sessions ago
10y real yield, 3-month change (pp)+26 bp7910-year real yield (DFII10) minus its level 63 sessions ago
HY spread, 1-month change (pp)-7 bp50ICE BofA US High-Yield OAS minus its level 21 sessions ago
HY spread within 1-year range0.0518(HY OAS − trailing-1-year min) ÷ (1-year max − min); 0 = tightest of the year, 1 = widest0 = tightest, 1 = widest
IG spread, 1-month change (pp)+2 bp74ICE BofA US Investment-Grade OAS minus its level 21 sessions ago

Caveats: FRED daily series applied with a T+1 knowability rule.

Earnings reaction — 63.7 (AMBER)

Weight 20% · Source: SEC EDGAR (8-K item 2.02 results filings) + Yahoo Finance next-session returns · Data through: 2026-06-24
Whether the market is selling good results — a 'good news sold' gauge. NOT consensus estimate revisions (no revisions feed connected).

IndicatorRaw valueScore (pctl 0–100)CalculationNote
Share of results sold next session (63-session window)56.8%77share of universe 8-K item 2.02 results in the last 63 sessions with a negative next-session return vs their layer benchmark (semis→SMH, software→IGV, else QQQ)
Mean abnormal reaction to results1.1%47mean next-session abnormal return across those results, inverted (more negative = more fragile)

Caveats: A reaction gauge, not consensus revisions — noisy by construction; reads best over full reporting seasons. · Excluded (6-K reporters, timing ambiguous): ARM, TSM, ASML, IREN, NBIS. · Consensus revision momentum: data unavailable — no estimates feed connected.

Supply & funding — 99.6 (RED)

Weight 10% · Source: SEC EDGAR (S-1 / S-3 / 424B* / FWP issuance filings) · Data through: 2026-06-24
Equity/credit issuance intensity within the universe. Count-based, not dollar-based.

IndicatorRaw valueScore (pctl 0–100)CalculationNote
Issuance filings, rolling quarter (weighted count)59100weighted count of S-1/S-3/424B/FWP filings in the last ~63 sessions (neocloud filings ×2), percentile-rankedneocloud filings ×2

Caveats: Counts, not dollars — a single mega-deal can matter more than many small filings. · Universe-only: does not capture market-wide AI issuance. · Deal sizes/pricing: data unavailable — no deals feed connected (deals.csv adapter ready).

5. Basket levels (indexed to 100 at 2022-11-30, ChatGPT release)

SeriesIndex level
Neoclouds (EW)5251.9
AI infrastructure (EW)1203.9
AI basket (equal-weight, all layers)825.3
Semis & hardware (EW)694.5
Power & grid (EW)608.8
SMH533.9
SOXX446.2
AI software (EW)337.0
Hyperscalers (EW)293.1
AIQ290.1
QQQ248.1
SPY192.1
IGV170.1
BOTZ169.7
IWM164.1

6. Missing / stale data warnings

Data availability by pillar: Caveats:

7. Stress episodes (22 detected since 2022-11-30)

Active episode since 2026-05-05 · peak 92.6 · AI-basket drawdown -12.4% · pillars: Extension, Leverage & speculation, Supply & funding, Earnings reaction, Rates & credit, Tape: distribution & breadth

PeriodPeakAI ddPillars flashedResolutionRecovered
2026-05-05 → ongoing92.6-12.4%Extension, Leverage & speculation, Supply & funding, Earnings reaction, Rates & credit, Tape: distribution & breadthongoing
2026-01-02 → 2026-04-0280.8-12.7%Extension, Tape: distribution & breadth, Supply & funding, Leverage & speculation, Rates & creditbreadth repaired, reaction recovered2026-04-02
2025-10-31 → 2025-12-2379.5-15.4%Extension, Leverage & speculation, Supply & funding, Tape: distribution & breadthrates relief, reaction recovered2026-01-06
2025-06-03 → 2025-08-2676.6-5.4%Extension, Earnings reaction, Supply & fundingrates relief, breadth repaired, reaction recovered2025-08-26
2025-02-24 → 2025-05-0283.1-27.1%Leverage & speculation, Earnings reaction, Supply & funding, Tape: distribution & breadth, Rates & creditrates relief, breadth repaired2025-05-12
2025-01-17 → 2025-01-2171.0-23.7%Extension, Rates & credit, Earnings reaction, Supply & fundingrates relief, breadth repaired2025-01-21

8. Underlying data endpoints

NameURL
auditJson/api/ai-audit
latest/api/latest
history/api/history
episodes/api/episodes
benchmarks/api/benchmarks
reports/api/reports
health/api/health
snapshotByDate/api/snapshot/{YYYY-MM-DD}