Datasets › Market signals
13F.ai AI Bubble Index
A daily reading, from 0 to 100, of how likely it is that the AI investment boom is about to break. It combines how much has been built up with how far the supports have failed, from public market and filing data.
Latest reading
At the close of Nov 20, 2026 Live
+1.4 since the previous trading day+5.3 since a week ago
Live: computed once, at this close, and never restated. Values before November 20, 2026 are reconstructed: worked out after the fact, with the weights chosen in November 2026. 39 of 41 inputs had a value on this day. Information, not a forecast and not investment advice.
- Dormant 0 to 25
- Inflating 25 to 45
- Stretched 45 to 60
- Cracking 60 to 75
- Breaking 75 to 90
- Burst under way 90 and up
Since May 2023
Show the chart as a table (the last trading day of each month)
| Date | Score | Band | Pressure | Puncture | State |
|---|---|---|---|---|---|
| Nov 20, 2026 | 69.4 | Cracking | 80.6 | 62.8 | Live |
| Oct 30, 2026 | 58.8 | Stretched | 82.9 | 46.8 | Reconstructed |
| Sep 30, 2026 | 46.0 | Stretched | 80.5 | 31.6 | Reconstructed |
| Aug 31, 2026 | 43.3 | Inflating | 79.9 | 28.8 | Reconstructed |
| Jul 31, 2026 | 46.4 | Stretched | 79.2 | 32.5 | Reconstructed |
| Jun 30, 2026 | 39.7 | Inflating | 75.3 | 25.9 | Reconstructed |
| May 29, 2026 | 32.6 | Inflating | 76.4 | 18.5 | Reconstructed |
| Apr 30, 2026 | 31.8 | Inflating | 74.4 | 18.0 | Reconstructed |
| Mar 31, 2026 | 43.7 | Inflating | 62.5 | 34.4 | Reconstructed |
| Feb 27, 2026 | 41.0 | Inflating | 63.9 | 30.5 | Reconstructed |
| Jan 30, 2026 | 32.6 | Inflating | 64.5 | 20.7 | Reconstructed |
| Dec 31, 2025 | 34.7 | Inflating | 60.2 | 24.0 | Reconstructed |
| Nov 28, 2025 | 31.4 | Inflating | 63.5 | 19.6 | Reconstructed |
| Oct 31, 2025 | 31.0 | Inflating | 62.6 | 19.4 | Reconstructed |
| Sep 30, 2025 | 28.6 | Inflating | 61.2 | 17.3 | Reconstructed |
| Aug 29, 2025 | 29.3 | Inflating | 62.2 | 17.7 | Reconstructed |
| Jul 31, 2025 | 28.0 | Inflating | 61.3 | 16.6 | Reconstructed |
| Jun 30, 2025 | 32.8 | Inflating | 56.8 | 22.7 | Reconstructed |
| May 30, 2025 | 39.0 | Inflating | 56.3 | 30.5 | Reconstructed |
| Apr 30, 2025 | 53.4 | Stretched | 54.3 | 52.8 | Reconstructed |
| Mar 31, 2025 | 53.8 | Stretched | 57.7 | 51.4 | Reconstructed |
| Feb 28, 2025 | 42.0 | Inflating | 60.5 | 33.0 | Reconstructed |
| Jan 31, 2025 | 34.6 | Inflating | 60.8 | 23.8 | Reconstructed |
| Dec 31, 2024 | 31.8 | Inflating | 57.9 | 21.3 | Reconstructed |
| Nov 29, 2024 | 23.3 | Dormant | 56.6 | 12.9 | Reconstructed |
| Oct 31, 2024 | 30.7 | Inflating | 54.1 | 21.1 | Reconstructed |
| Sep 30, 2024 | 32.7 | Inflating | 53.2 | 23.7 | Reconstructed |
| Aug 30, 2024 | 37.3 | Inflating | 53.3 | 29.4 | Reconstructed |
| Jul 31, 2024 | 37.0 | Inflating | 54.5 | 28.6 | Reconstructed |
| Jun 28, 2024 | 34.8 | Inflating | 53.9 | 25.9 | Reconstructed |
| May 31, 2024 | 32.2 | Inflating | 50.6 | 23.8 | Reconstructed |
| Apr 30, 2024 | 35.0 | Inflating | 45.7 | 29.3 | Reconstructed |
| Mar 28, 2024 | 30.5 | Inflating | 45.8 | 23.3 | Reconstructed |
| Feb 29, 2024 | 31.5 | Inflating | 42.3 | 25.9 | Reconstructed |
| Jan 31, 2024 | 33.1 | Inflating | 40.6 | 28.8 | Reconstructed |
| Dec 29, 2023 | 27.8 | Inflating | 29.6 | 26.7 | Reconstructed |
| Nov 30, 2023 | 30.7 | Inflating | 30.9 | 30.6 | Reconstructed |
| Oct 31, 2023 | 34.7 | Inflating | 32.4 | 36.4 | Reconstructed |
| Sep 29, 2023 | 35.3 | Inflating | 30.6 | 38.8 | Reconstructed |
| Aug 31, 2023 | 32.0 | Inflating | 34.4 | 30.5 | Reconstructed |
| Jul 31, 2023 | 33.1 | Inflating | 37.7 | 30.3 | Reconstructed |
| Jun 30, 2023 | 36.8 | Inflating | 37.1 | 36.5 | Reconstructed |
| May 31, 2023 | 42.7 | Inflating | 38.9 | 45.5 | Reconstructed |
Pressure and Puncture
The score is high only when both readings are high.
Pressure
80.6How much has been built up: valuations, how the build-out is financed, and the leverage and speculation riding on it.
Puncture
62.8How far the supports have failed: credit, price action, interest rates, the supply of new stock, demand for AI and confidence.
The nine pillars
Each pillar is scored from 0 to 100 from the inputs named beside it.
| Pillar | Score | Bar | Inputs |
|---|---|---|---|
| ValuationHow far prices have run and how expensive the market is. | 92.1 | AI stocks: two-year gain; Chip stocks: two-year gain; Shiller CAPE; Big stocks against the average stock | |
| FinancingHow the largest spenders pay for the build-out. | 76.7 | Spenders: capital spending against cash flow; Spenders: new stock against cash flow; Nvidia: days to collect payment | |
| SpeculationBorrowed money and leveraged trading riding on the theme. | 74.3 | Margin debt: growth over a year; Trading in leveraged funds; Money raised by AI companies selling stock (not yet included) |
| Pillar | Score | Bar | Inputs |
|---|---|---|---|
| CreditStrain among the borrowers and lenders behind the build-out. | 75.9 | Borrowers' shares: fall from high; High-yield bonds against Treasuries; Private-credit lenders: fall from high; CoreWeave bond yield (not yet included) | |
| Price actionHow AI stocks are trading: falls from highs, breadth and swings. | 96.3 | AI stocks: fall from high; AI stocks below their 200-day average; AI stocks down 40% or more; Seven largest AI companies: fall from high; Chip stocks and Korea: fall from high; AI stocks: how much prices swing; Days since the AI stocks' high; Nasdaq 100 volatility (VXN); Reaction to results | |
| LiquidityInterest rates and near-term fear in options prices. | 76.6 | Short-term interest rates: six-month change; Ten-year Treasury yield: six-month change; Near-term against three-month volatility | |
| Supply of stockInsiders, funds and new issuers heading for the exit. | 35.4 | Insiders selling against buying; Funds closing against opening positions; Planned insider sales; Withdrawn IPOs | |
| Demand for AIWhether businesses keep adopting and paying for AI. | 24.2 | Firms using AI; AI job postings; Business spending on AI tools; Nvidia revenue growth: speeding up or slowing; Spenders: growth in capital spending | |
| ConfidenceSigns that investors' conviction is cracking. | 64.8 | Leveraged AI funds: fall from high; Margin debt against its peak; AI stocks moving together; Attention: Nvidia on Wikipedia; Officer changes; Short selling of the borrowers |
What moved
- Seven largest AI companies: fall from high Price actionUp: toward more risk
- Nasdaq 100 volatility (VXN) Price actionUp: toward more risk
- AI stocks: fall from high Price actionUp: toward more risk
- AI stocks: how much prices swing Price actionUp: toward more risk
- Near-term against three-month volatility LiquidityUp: toward more risk
The inputs whose readings changed most between the previous trading day and Nov 20, 2026, largest first. How much each one moved is not published.
How it works
The index asks one question: how likely is it that the AI investment boom is about to break? It combines two readings.
Pressure measures how much has been built up: prices and valuations, how the build-out is financed, and how much leverage and speculation ride on it. Puncture measures how far the supports have failed: credit, price action, interest rates, the supply of new stock, demand for AI, and confidence. The score is high only when both are high.
Each reading is built from public data: market prices, SEC filings, FINRA, Treasury, Census and other sources listed on this page. Inputs are scored against fixed reference levels drawn from past booms and from this cycle.
History before November 20, 2026 is reconstructed with weights chosen in November 2026 and is labelled as such; values published since are never restated.
The index is information, not a forecast and not investment advice.
Known limits
- The weights and reference levels are judgement, not fitted to data: there are too few past bubbles to fit them.
- April 2025 read high during the tariff sell-off, and no break followed. That was a false alarm.
- The basket of companies was chosen in November 2026, so its history has hindsight in it.
- Two inputs are not yet included (listed below).
What goes into it
Valuation Pressure
- AI stocks: two-year gain. How far a fixed basket of AI-related stocks (chip makers, cloud and software companies, data-centre builders and power suppliers) has risen over two years, taking the highest reading of the past year. Source: Tiingo daily prices. Daily.
- Chip stocks: two-year gain. The same two-year gain for semiconductor stocks, measured with the iShares Semiconductor ETF (SOXX). Source: Tiingo daily prices. Daily.
- Shiller CAPE. The cyclically adjusted price-to-earnings ratio of the S&P 500: prices against ten years of inflation-adjusted earnings. Source: Robert Shiller's online data (shillerdata.com). Monthly.
- Big stocks against the average stock. How much better the cap-weighted S&P 500 (SPY) has done than the equal-weighted S&P 500 (RSP) over two years, taking the highest reading of the past year. A wide gap means a few large companies carry the market. Source: Tiingo daily prices. Daily.
Financing Pressure
- Spenders: capital spending against cash flow. The five largest AI spenders (Microsoft, Alphabet, Amazon, Meta and Oracle): capital spending in their latest quarter as a share of the cash their operations brought in. Source: SEC financial statement data (XBRL company facts). Quarterly, as each 10-Q and 10-K is filed.
- Spenders: new stock against cash flow. The same five companies: stock issued minus stock bought back over the last four quarters, as a share of their operating cash flow. Source: SEC financial statement data (XBRL company facts). Quarterly, as each 10-Q and 10-K is filed.
- Nvidia: days to collect payment. Nvidia's receivables against its sales: how many days of sales its customers still owe. Source: SEC financial statement data (XBRL company facts). Quarterly, as each 10-Q and 10-K is filed.
Speculation Pressure
- Margin debt: growth over a year. How much more investors have borrowed against their brokerage accounts than a year earlier. Source: FINRA margin statistics. Monthly.
- Trading in leveraged funds. Dollar trading in two leveraged chip and Nasdaq funds (SOXL and TQQQ) against trading in the S&P 500 fund SPY, over 20 days, ranked against its own history since 2012. Source: Tiingo daily prices. Daily.
Credit Puncture
- Borrowers' shares: fall from high. The average fall from their two-year high of the shares of AI companies that borrow heavily to build data centres. A stand-in until their bond prices are included. Source: Tiingo daily prices. Daily.
- High-yield bonds against Treasuries. A high-yield bond fund (HYG) against a Treasury bond fund (IEF): how far the ratio has fallen from its one-year high. Source: Tiingo daily prices. Daily.
- Private-credit lenders: fall from high. How far business development companies, the listed private-credit lenders (BIZD), have fallen from their two-year high. Source: Tiingo daily prices. Daily.
Price action Puncture
- AI stocks: fall from high. How far the AI basket has fallen from its two-year high. Source: Tiingo daily prices. Daily.
- AI stocks below their 200-day average. The share of the AI basket trading below its 200-day average price. Source: Tiingo daily prices. Daily.
- AI stocks down 40% or more. The share of the AI basket at least 40% below its high. Source: Tiingo daily prices. Daily.
- Seven largest AI companies: fall from high. How far the seven largest AI companies, taken together, have fallen from their two-year high. Source: Tiingo daily prices. Daily.
- Chip stocks and Korea: fall from high. The average fall from their two-year highs of semiconductor stocks (SOXX) and South Korean stocks (EWY), home of the largest memory makers. Source: Tiingo daily prices. Daily.
- AI stocks: how much prices swing. The AI basket's price swings over 20 days against their typical level over the past year. Source: Tiingo daily prices. Daily.
- Days since the AI stocks' high. Trading days since the AI basket last set a one-year high. Source: Tiingo daily prices. Daily.
- Nasdaq 100 volatility (VXN). Cboe's index of the volatility options traders expect in the Nasdaq 100. Source: Cboe index history (VIX, VIX3M, VXN). Daily.
- Reaction to results. The average one-day share-price move of the seven largest AI companies after their quarterly results (8-K item 2.02), over the last 60 days. Source: SEC 8-K filings and Tiingo daily prices. As results are filed.
Liquidity Puncture
- Short-term interest rates: six-month change. The change in the three-month Treasury yield over six months. Source: US Treasury daily par yield curve rates. Daily.
- Ten-year Treasury yield: six-month change. The change in the ten-year Treasury yield over six months. Source: US Treasury daily par yield curve rates. Daily.
- Near-term against three-month volatility. Cboe's VIX against its three-month VIX. Above one, traders expect more turbulence soon than later. Source: Cboe index history (VIX, VIX3M, VXN). Daily.
Supply of stock Puncture
- Insiders selling against buying. In the AI basket, the number of insiders who sold shares on the open market against the number who bought, over 12 months (people, not dollars). Source: SEC Forms 4 (insider trades). As filed.
- Funds closing against opening positions. For the seven largest AI companies, the number of funds that closed a position against the number that opened one in the latest quarter, from 13F reports. Source: SEC Form 13F reports. Quarterly, after the filing deadline.
- Planned insider sales. Form 144 notices of planned insider sales in the AI basket over 30 days, against the usual 30-day count of the past year. Source: SEC Form 144 notices. As filed.
- Withdrawn IPOs. The share of initial public offerings that were withdrawn rather than priced over 90 days. Source: SEC registration statements and prospectuses. As filed.
Demand for AI Puncture
- Firms using AI. The change in the share of US firms that say they use AI, from the Census Bureau's business survey. Source: US Census Bureau, Business Trends and Outlook Survey. Every two weeks.
- AI job postings. The change over 90 days in the share of US job postings on Indeed that mention AI. Source: Indeed Hiring Lab (hiringlab.org), AI tracker, CC BY 4.0. Daily figures, published about monthly.
- Business spending on AI tools. Ramp's measure of how many US businesses pay for AI tools: the recent three-month gain against the best three-month gain of the past year. Source: Ramp AI Index. Monthly.
- Nvidia revenue growth: speeding up or slowing. Nvidia's quarter-on-quarter revenue growth minus the growth of the quarter before. Source: SEC financial statement data (XBRL company facts). Quarterly, as each 10-Q and 10-K is filed.
- Spenders: growth in capital spending. The quarter-on-quarter growth in capital spending of Microsoft, Alphabet, Amazon, Meta and Oracle together. Source: SEC financial statement data (XBRL company facts). Quarterly, as each 10-Q and 10-K is filed.
Confidence Puncture
- Leveraged AI funds: fall from high. The average fall from their two-year high of two leveraged funds that track chip stocks and Nvidia (SOXL and NVDL). Source: Tiingo daily prices. Daily.
- Margin debt against its peak. Margin debt against its highest level of the past 12 months. Source: FINRA margin statistics. Monthly.
- AI stocks moving together. How closely the stocks in the AI basket have moved together over 60 days. When everything moves as one, investors are trading the theme, not the companies. Source: Tiingo daily prices. Daily.
- Attention: Nvidia on Wikipedia. Views of Nvidia's English Wikipedia article over seven days against their typical level over the past year. Source: Wikimedia page views. Daily.
- Officer changes. 8-K filings on officer and director changes (item 5.02) in the AI basket over 30 days, against the usual 30-day count of the past year. Source: SEC 8-K filings. As filed.
- Short selling of the borrowers. The change in shares sold short of the AI companies that borrow heavily, over the last three reports. Source: FINRA equity short interest. Twice a month.
Not yet included
- Money raised by AI companies selling stock (Speculation). Not yet included. The filings we collect show only priced initial offerings, not follow-on offerings, convertible notes or private rounds, so the figure would read zero while shares are being sold. It will be added once follow-on offerings are collected.
- CoreWeave bond yield (Credit). Not yet included. It waits for a decision on the terms of use of FINRA's bond trade data.
Sources and credits
- Tiingo daily prices
- Robert Shiller's online data (shillerdata.com)
- SEC financial statement data (XBRL company facts)
- FINRA margin statistics
- SEC prospectuses (EDGAR)
- FINRA TRACE bond trade data
- Cboe index history (VIX, VIX3M, VXN)
- SEC 8-K filings and Tiingo daily prices
- US Treasury daily par yield curve rates
- SEC Forms 4 (insider trades)
- SEC Form 13F reports
- SEC Form 144 notices
- SEC registration statements and prospectuses
- US Census Bureau, Business Trends and Outlook Survey
- Indeed Hiring Lab (hiringlab.org), AI tracker, CC BY 4.0
- Ramp AI Index
- Wikimedia page views
- SEC 8-K filings
- FINRA equity short interest
Job postings data: Indeed Hiring Lab (hiringlab.org), AI tracker, licensed under CC BY 4.0. We transform the data: the index uses the change in the AI share of US postings, scored on our own scale.
The same data is in the REST API (GET /api/v1/ai-bubble-index) and in the connector (get_ai_bubble_index), free on every plan. About this data says when it was last updated.
Change log
Method version 1.0.0.
| Version | Date | What changed |
|---|---|---|
| 1.0.0 | Nov 20, 2026 | First published version. The index launches with 39 inputs in nine pillars, from market prices (Tiingo), SEC filings (company financial statements, Forms 4, 144 and 13F, 8-K items, registration statements), FINRA margin statistics and short interest, Treasury yields, Cboe volatility indexes, the Census Bureau's business survey, Indeed Hiring Lab, the Ramp AI Index, Wikipedia page views and Shiller's CAPE. Two inputs are not yet included: money raised by AI companies selling stock (the filings we collect show only priced initial offerings, not follow-on offerings, convertible notes or private rounds) and the yield on CoreWeave's bonds (it waits for a decision on the terms of use of FINRA's bond trade data). History from May 2023 to November 19, 2026 is reconstructed with the weights chosen in November 2026; values from November 20, 2026 on are computed once, on their day, and never restated. |