The Convergence

Hub AI & Robotics Health Sponsors Jump to a planet… AI & Robotics Health & Wellness Sports Entertainment Breaking News Celebrities / Pop Culture (soon) News / Politics (soon) Weather (soon) Gaming Food & Recipes (soon) Finance & Investing Travel Fashion / Beauty (soon) Automotive Shopping Business Archive Real Estate & Housing Relationships / Dating (soon) Education / How-To (soon) True Crime Alternative & Natural Medicine Nanotechnology Science (soon) Parenting / Family (soon)

Home / Valuation

Jeremy Grantham's CAPE Framework, Applied to the AI Trade

He called the dot-com crash and the housing bubble using the same lens. Here's what that lens says about markets today — and where the numbers actually stand.

PUBLISHED 2026-07-26 · VAL_03

As of early 2026, the Shiller CAPE ratio for the S&P 500 climbed above 40 — a level seen only once before in over 150 years of market history, the December 1999 dot-com peak of roughly 44.2. Jeremy Grantham, who correctly called both the dot-com and housing bubbles using a framework based on statistical price deviation from long-term trend, has said there is "slim to none" chance the current AI-driven rally avoids ending in a bust.

Grantham's framework, in plain terms

Grantham defines a bubble as a two-standard-deviation divergence of an asset's price above its long-term real price trend — a statistical measure, not a gut feeling. His central historical claim is that every prior two-sigma bubble in large developed equity markets has eventually broken and retraced all the way back down to that pre-existing trend line. Not partway. All the way.

Where CAPE stands today vs. past peaks

PeriodApprox. CAPE levelOutcome Long-term historical average~16–17Baseline 1929 peak (pre-crash)~~32.5Market crash 2021 post-pandemic high~~38.5Subsequent correction Dot-com peak (Dec. 1999)~~44.2~~2-year bear market Early 2026 (current)~40–40.6Unresolved

The current reading has been above 30 for an extended period already — and history shows that every time CAPE has stayed elevated above 30 for a sustained stretch, the market has eventually seen a decline of 20% or more. A reading above 40 has happened only twice: today, and in the run-up to the dot-com crash.

The case for "this time is different" — and its limits

Unlike the dot-com era, when many highly valued companies had little or no revenue, today's AI leaders are largely profitable, with real earnings growth and measurable productivity gains in sectors like software engineering. Hyperscaler capital expenditure — the combined infrastructure spending of the largest tech companies — is approaching $660–690 billion in 2026 alone, arguably the largest corporate investment program in history outside wartime mobilization. That's a real difference from 2000.

But Grantham's own point is that profitability doesn't inoculate a market against being priced for perfection. Market concentration today has exceeded even dot-com-era levels, meaning a small handful of companies carry an outsized share of index performance — so the risk isn't only "do these companies make money," it's "what happens to the index if their growth merely slows rather than fails."

What this means for investors

Grantham's own stated response to this setup isn't to go to cash — it's diversification: holding meaningful exposure outside U.S. equities, along with precious metals and bonds, rather than concentrating in the names driving the current rally. Whether or not his bubble call proves right on timing, the underlying valuation math — a CAPE ratio that has only been this high once before in 155 years — is not in dispute. What's in dispute is how much the current earnings quality changes what happens next.

Jeremy Grantham's CAPE Framework, Applied to the AI Trade

Grantham's framework, in plain terms

Where CAPE stands today vs. past peaks

The case for "this time is different" — and its limits

What this means for investors

Related reading