Bubble watch
Is the AI Market Expensive? What CAPE and Valuation History Actually Say
"Expensive" is the most-thrown and least-useful word in markets. The cyclically-adjusted P/E — CAPE — is the most-cited way to put a number on it. Here's what it measures, what history says about high readings, and why it's a barometer, not a sell button.
What CAPE actually measures
The Shiller CAPE (cyclically-adjusted price-to-earnings, after economist Robert Shiller) divides a market's price by its average inflation-adjusted earnings over the past ten years, instead of the last twelve months. The ten-year smoothing is the whole point: single-year earnings swing wildly with the cycle, so a trailing P/E can look "cheap" at a profit peak and "expensive" at a trough. CAPE strips out that noise to compare valuations across decades on a level field.
What history says about a high CAPE
The robust finding across long datasets is narrow but real: elevated CAPE has historically lined up with lower average returns over the following decade — not the following month, quarter, or even year. As a long-run expectations dial it has signal. As a market-timing tool it is close to useless: valuations have stayed historically elevated for years at a stretch while prices kept rising. Anyone who shorted "expensive" on CAPE alone has a long history of getting run over.
That is the same lesson, from a different instrument, as why "expensive" doesn't mean "short."
Applying it to the AI trade
Two caveats matter for AI specifically. First, concentration: when a handful of mega-caps dominate the index, the headline market multiple is really a few companies' multiples wearing a trenchcoat — which is why our bubble-risk index scores concentration as its own pillar alongside valuation. Second, a market-level number can't tell you whether an individual name is stretched. For that, flip the question from "what's fair value?" to "what growth is the price assuming?" — the reverse-DCF read the fair-value tool prints for any ticker.
Treat CAPE the way you'd treat a barometer before a hike: useful context, not a forecast of the exact hour it rains. A high reading argues for humbler return expectations and tighter risk, not a short. Next, the same idea through a different lens: bubbles in real terms.
Santro AI is an informational tool. Market data is delayed ~15 minutes and provided for education, not as financial advice. We deliberately don't quote a live CAPE figure here — check a primary source for the current reading.