Analysis · Tool

AI Bubble Meter

S&P 500 breadth · updated weekly (Saturdays) · latest data loading…

Is this another dot-com? The real question isn't whether prices are "too high" — it's how few names are doing the work. This gauge measures that breadth live across the full S&P 500.

How it's built

The meter summarizes four measurements from its loaded dataset. Each component is ranked against its own history and those four percentile ranks are averaged. It describes the selected data; it has not been validated as a forecast or a sell signal.

• Concentration — the combined market-cap weight of the 10 largest names. The higher their share, the more the index rests on a handful of stocks.
• Narrow leadership — the trailing-12-month gap between the cap-weighted S&P 500 and the average (equal-weight) stock. A wide gap means the index is rising on a few giants while the typical stock lags.
• Weak participation — the share of S&P 500 stocks trading below their own 10-month trend. 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 of the top 10); a lower yield means pricier leaders. This signal needs company fundamentals (available from about 2022, applied with a one-year reporting lag to avoid look-ahead), so the scored history begins in 2023 — the meter covers the last few years to today.

A high reading means breadth is narrow and fragile; a low reading means the rally is broad and healthier. The reassuring counter-signal is a rally that broadens out — more stocks, sectors, and countries joining — which tends to last longer.

How to read the needle — and how not to

A high reading means several components are high relative to their loaded historical distributions. It does not estimate a probability of a crash. Inspect the raw components and your portfolio concentration before attaching an economic explanation to the dial.

Changes in the score can come from different components. Read the underlying values to distinguish a broader rally from changing leader valuations. The data begins much later than the dot-com period, so the historical chart cannot validate claims about 1999 or 2000.

Percentile ranks use the full loaded sample, including later observations when scoring earlier months. Adding data can therefore revise older readings. Current index membership and share counts are applied backward, and fundamentals have reporting-lag approximations. These features limit any interpretation as a signal available in real time.

The average of component percentiles is not the percentile of the combined score. A score of 70 does not mean it exceeds exactly 70% of historical composite readings. See methodology and the worked breadth example for the calculation and its limits.

Educational content only — not investment advice, a recommendation, or a forecast. Built from historical S&P 500 data using the current index membership and current shares applied back through history (a survivorship/share approximation). It does not predict crashes. See our editorial policy.