AI Bubble Meter — how the breadth gauge is built
The AI Bubble Meter answers one question: how few names are doing the work? Near the dot-com top the headline index kept climbing while fewer and fewer stocks actually participated — narrowing breadth was the real warning sign. The meter measures that breadth across the whole S&P 500 and rolls it into a single 0–100 reading. It is a fragility gauge, not a crash predictor.
What it reads
Every month it scores the full S&P 500 (~500 companies) on four signals, using month-end closing prices:
- Concentration — the combined market-cap weight of the 10 largest companies as a share of the whole index. The higher their share, the more the index leans on a handful of mega-caps. (The card also shows the top-5 weight.)
- Narrow leadership — the gap between the cap-weighted S&P 500's trailing 12-month return and the average (equal-weight) stock's 12-month return. A wide positive gap means the index is being carried by a few giants while the typical stock lags — exactly the dot-com pattern.
- Weak participation — the share of S&P 500 stocks trading below their own 10-month trend (their 10-month average price). When fewer stocks are in uptrends, participation is thinning under the surface.
- Leader valuation — the earnings yield of the mega-cap leaders (net income ÷ market cap for the top 10). A lower yield means pricier leaders. This signal needs company fundamentals, which begin around 2022 (see below).
How the score is calculated
Each signal is turned into a percentile of its own history over the period the meter covers (2023 to today). For example, if today's top-10 concentration is higher than it was in 90% of all those months, that signal scores 90. The meter then averages the four percentiles into the headline 0–100 composite.
Scoring against each signal's own history (rather than fixed thresholds) keeps the gauge self-calibrating and neutral: a reading of 70 means "more stretched than 70% of the last decade," not a hard-coded verdict that prices are wrong.
The reading falls into three bands:
- 0–45 — Broad. Participation is wide; the rally rests on many stocks.
- 45–70 — Mixed. Leadership is somewhat concentrated.
- 70–100 — Narrow. A few names are doing the work; the market is more fragile to a stumble in the leaders.
Why the meter starts in 2023
The first three signals come from prices and shares, which reach back many years. The fourth — leader valuation — needs company fundamentals (net income), which begin around 2022 and are applied with a one-year reporting lag (each month uses the prior fiscal year, to avoid look-ahead). That pushes the first fully-scored month to 2023. Rather than have the gauge change definition partway through its history, the meter covers 2023 to today, so every month carries all four signals and the composite is computed the same way throughout. The exact start shifts as fundamentals data is refreshed.
How to read it
The needle shows the latest month by default. Hover the history line to move the gauge to any past month and read the signals as they stood then — useful for comparing today against, say, late 2021 or earlier peaks. A high reading is a caution flag about fragility, not a sell signal: markets can stay narrow for a long time. The reassuring counter-signal is the opposite — a rally that broadens out, with more stocks joining, tends to be healthier and last longer.
Important: breadth narrows at frothy tops and at panic bottoms (when only a few defensive names hold up). So a market low can also read "narrow." The meter measures participation and concentration — it does not forecast crashes or tops.
The thinking behind the gauge is in the analysis post AI Boom or Dot-Com Repeat?