Everyone Agrees the Market Is Mad. The Agreement Is the Risk.

The consensus is right: the market is flashing signs of madness. Stocks trade around 40 times a decade of inflation-adjusted earnings, a level touched only once in 150 years, at the top of the dot-com bubble. Ten companies now carry roughly 40 percent of the S&P 500. Margin debt just spent months at record highs. The people telling you a crash is coming have been telling you that for two years, while the S&P 500 rose about 12 percent this year and sat near its all-time high.

None of that is wrong. It is also not the reason to be afraid. The mistake is not in the facts. It is in believing that a list of expensive-looking numbers tells you when the market will break.

The uncomfortable truth about the "many signs of madness" is that almost all of them are receipts, not warnings. A valuation is the price someone already paid. When a number becomes famous for being high — the Shiller CAPE, price-to-sales, the ratio of market value to GDP — it has usually been high for years and low for nothing. Every one of these gauges was lit in 2016, 2019, and 2021, long before the market actually stumbled, and standing outside and watching them glow protected no one.

A level has no marginal buyer left in it. The crowd already owns the price. That is the hidden distinction the catalogues blur: expensive signs they can see, and crowded positions that can be forced to move.

Diversified by Count, Concentrated by Cause

The catalogue loves to cite the market's height. It tends to bury the market's shape.

Ten names are close to 40 percent of the index, a concentration the market has not carried before. That is not a headline number that crawls higher and then stops mattering. It is a structural fact about how this market falls. When a handful of companies are two-fifths of your index, an ordinary stumble in one of them is no longer a single-company event — it is a large, immediate, un-diversifiable hit to everyone who holds the index at all. The S&P 500 feels like a broad portfolio because it owns hundreds of securities. It is in fact concentrated because most of those names are asked to deliver the same future: more AI spending, the same generous buyers, the same five lenders staying open.

Security count is not diversification when the future has only one script.

The Leverage That Has to Sell

The second fact the catalogues record rather than interpret is the debt. Margin balances reached about $1.42 trillion mid-year, up more than half from a year earlier, before a small July pullback — and most of that money is not held by people who can wait out a dip. It is held by people who are sold off by a broker if the price drops far enough.

That is the difference between an expensive market and a falling one. Expensiveness is a photograph. Leverage is a mechanism. When an investor buys on margin, a decline is not a paper loss they can ignore; it becomes a call to post more cash or be sold out, and the forced sale knocks the price down further, which triggers the next call. A market with record margin debt does not quietly deflate. It has a reflexive gear that turns a routine pullback into a margin-call cascade.

The same logic hides under the boom in AI. The build-out is not being financed only, or even mostly, by cash. AI-linked debt issuance roughly quadrupled from about $166 billion in 2023 to an estimated $625 billion in 2025, with roughly another half-trillion dollars expected this year. Hyperscalers spent on the order of $300 billion on capital investment in 2025 against AI revenue that still measures in the tens of billions. The difference between the celebrated numerator — the wonder of the technology — and the cash required to keep it running is where the leverage has moved in.

The Crowd Selling You the Future

Watch what insiders do, not what they announce. A wave of private giants — OpenAI, SpaceX, Anthropic — is said to be rushing toward public listings at valuations near a trillion dollars each. Every valuation "near a trillion" is a seller saying the number feels attainable now and may not be in three years. When the people who know a business best choose to sell you their future at what amounts to a party price, the madness is not that the public is buying. It is that the smartest cash is quietly offering the shares.

This is the deepest irony of the "many signs" story. The sign the crowd treats as safety — its own worry — is the one that matters least. A forewarned, widely watched market is not obviously safer. But a crash does not arrive because people are afraid; it arrives when a position that must be sold meets a price that can no longer absorb it. The crowd is photographing the levels. The forced sellers live in the leverage and the concentration.

What Would Prove the Consensus Wrong

Which means the bear story is testable, and the tests cut both ways.

The inversion fails — the mad-consensus is wrong and the market keeps climbing — if the market broadens and delevers at the same time it grinds higher: more of the index's 500 names participating, margin debt draining down without a price decline, and earnings catching up to the multiples. That combination says the froth is being worked off rather than built up.

The dangerous version is the opposite. It is not the CAPE crossing a prettier round number. It is margin debt breaking lower while prices fall — leverage forcing sellers — and the index's handful of giants leading the decline instead of cushioning it. That is the mechanism, and it does not announce itself in the catalogue of levels.

So keep the catalogue, but read it as what it is: a crowd of observers who have already paid for the future and now want someone else to validate their fear. Agreement feels like prudence. It is what a consensus looks like before it has to be sold. Being afraid in the same direction as everyone else protects a career, not a portfolio. The number to watch is not how expensive the market is. It is who is leveraged, how concentrated the index is, and what happens to that leverage on the first bad day.