Methodology
The AI Capital Stress Score runs from 0 (low stress) to 100 (high stress). It combines seven Layer Scores, each built from many indicators of the AI capital cycle, and is read in five bands: Low, Moderate, Elevated, High and Crisis.
The seven Layers: AI Fundamentals (is AI revenue growing fast enough to justify the capital?), Physical AI (is compute still scarce?), AI Credit (is the marginal dollar of AI capital getting more expensive?), Rates & Sovereign, Liquidity, Capital Crowding (is AI riding broad liquidity or absorbing capital from everything else?) and Policy & Geopolitics.
A Regime names the state of the cycle, such as Crowded Expansion or Funding Stress. A new Regime first appears as Pending: its conditions must hold for several consecutive closes before it replaces the current one. The site names the current Regime and any Pending one, not the values behind their conditions.
A Divergence flags two markets that normally move together drifting apart, for example AI equities rising while credit weakens. Each has an intensity from 0 to 1 and turns Active when it stays high for several closes.
Public data are shown with their value, change and source time; their publishers are credited on the Data sources page. Licensed market data feed the calculation but are never shown as a level, change or chart: only in the treated forms listed below. A new bond's AI credit premium is shown only as a range (such as 20–40 bp) and a single licensed price only by its 20-day direction. Raw licensed rows are not published.
The page is rebuilt twice each US trading day, before the open and after the close; each value shows its source time on hover. If an update is missed, the page keeps the previous update's data.
This page is general information about market conditions, not investment advice. It is not tailored to anyone's circumstances and is not an offer or solicitation to buy or sell any security.
How each value is shown
- Public value
- Series whose publisher allows reuse (U.S. Treasury and Federal Reserve data, central banks, EIA, SEC filings, CFTC): the value, its change and its source time.
- Aggregate
- Cloud providers' GPU and model prices: medians, indices and availability scores across many quotes, never a single zone's or region's price.
- Basket statistic
- Statistics over the monitor's own baskets of AI companies, such as breadth, concentration or the cap-weighted index, from which no single company's price can be read.
- Normalized
- Licensed market data the scores use (equity and option prices, credit spread indices, volatility indices): its Metric Stress from 0 to 100 and band, its two-year percentile and 20-day direction, with the source time; no level, change or chart.
- Percentile
- Licensed series shown for context but not scored (an issuer's distance to default, fund flows, Bitcoin): only the two-year percentile and 20-day direction, with an issuer's rank among the issuers or whether a flow is an inflow or an outflow.
What each page shows
Overview — The Stress Score and its trend, a short analysis by topic, the seven Layer Scores, seven summary cards, the Divergence monitor, the Themes and what changed in the last seven days; a ticker of public values.
Credit — The Layer Score and its drivers; the AI credit premium, distance to default and credit spreads as stress; AI bond issuance and debt due; each issuer's credit and the newest bond terms.
Physical AI — The Layer Score and its drivers; GPU rental prices and availability as aggregates across many quotes; data-centre and power events.
Rates & Liquidity — The Rates & Sovereign and Liquidity Layer Scores; US yields and the curve, global sovereign yields, the dollar, funding, net liquidity and Treasury auctions.
Crowding — The Layer Score and its drivers; breadth and concentration of the AI baskets, suppliers against buyers, AI against other risk assets, and fund flows as direction.
Policy & Geo — The Layer Score; policy and geopolitical events classified by theme and severity, the AI social license by state, the Themes and the event log.
Fundamentals — The Layer Score and its drivers; the investment gap, capex against operating cash flow and cloud/AI revenue, token prices and hyperscaler economics.