Everyone's Right That Stocks Keep Rising. That's Exactly Why the Margin of Safety Is So Thin.

The consensus has the receipts. The S&P 500 earned roughly an 18% total return in 2025 and is up double digits again in 2026, and Ainvest data shows its proxy, the SPY ETF, higher by 12.08% year to date as of mid-September. It set a fresh all-time high around 7,814 in August 2026. Is there a safer-sounding sentence on Wall Street right now than "US stocks keep going up"? Here is the part that sentence leaves out. A belief becomes financially interesting only once people have paid for it, and what people are paying for this one is a price so high that the modern market has essentially walked this ground once before - in the run-up to the dot-com crash that erased roughly half the index. The rally is not the argument against the market. The rally is the reason the safety has been priced out of it. ## Only One Other Time Was It This Expensive The cleanest way to see the price is the Shiller CAPE - the cyclically adjusted price-to-earnings ratio, which divides the index's price by its average inflation-adjusted earnings over the prior decade to smooth out the boom and bust in profits. As of September 2026 it stands at about 40.6, up from 38.6 a year earlier.

S&P 500 Shiller CapE trend (Sept 2025 - Sept 2026)Cyclically adjusted price-to-earnings ratio

Shiller CapE rose from 38.6 in Sept 2025 to a peak of 41.1 in Aug 2026 before easing to 40.6 in Sept 2026, a level historically associated with stretched valuations.

Period	Shiller CapE
Sept 2025	38.58
Aug 2026	41.12
Sept 2026	40.58

A reading in the low 40s means investors are paying more than forty years of the decade's average earnings for the market. The striking part is not the number itself but where it sits in the record: the modern era has posted a CAPE this high only in the run-up to the dot-com crash of 2000, after which the S&P 500 lost about 49% of its value. That is effectively the one precedent the price has - and it is not a reassuring one. ## Diversified By Count, Concentrated By Cause The usual objection to that is fair: valuation alone never killed an index, the market keeps making highs, and the S&P 500 holds about five hundred companies, so the danger is diffuse. It is not, and the reason is the metric investors generously call breadth. The top ten stocks now account for roughly 36% of the S&P 500's total weight, up from about 30% in mid-2023. That figure carries a weaker source than the rest of the story, so treat it as supporting detail, not the crux - but the shape it describes is real.

The index looks like five hundred independent bets. A large share of its value is actually one bet, concentrated in a handful of AI-linked mega-caps, and the plumbing runs like this: those names are so heavy in the index that when a catalyst hits them - a rate shock, an AI-capital-spending disappointment, a demand reversal - a narrow correction in a few stocks moves the entire index, because their weight is what the index has become. The breadth of a return to mean is much narrower than the index's apparent diversification suggests. ## The Price, the Age, and the Base Rate This rally is also old. From the October 2022 low, the bull market had reached about 1,314 days by May 2026, the tenth-longest of 27 on record since 1928, against an average of about 1,022 days and a median of 651. And the historical record is blunt about what tends to happen: since 1926 there have been about 24 declines of 20% or more, on average one every 3.5 years, with an average peak-to-trough fall of about 36%. The most recent 20%-plus drawdown was 2022. Now the honesty a statistic like that demands. None of this is a crash forecast, and the entire point of a margin of safety is that you should not have to predict a date to respect the size of a possible loss. The every-3.5-years figure is a roughly one-hundred-year average frequency, not a schedule - markets do not bottom on a clock. CAPE is one lens, not a verdict: on forward earnings, some screens show the index looking relatively cheap after July's selloff. Hold all of that, and the picture still changes where you are standing. You can hold the consensus - that stocks keep rising - and simultaneously accept that the current price carries a thinner buffer against a historically large drawdown than almost any other moment of the modern record, because the two facts are not actually in tension. The second is simply the price of the first. The question is not whether the market can keep going up; it can, and nothing here says it stops this quarter or next. The question is how much downside you are volunteering to hold for the privilege of being right about the upside, when the top of the index is one concentrated bet and the price you are paying is a level that, in the whole modern record, has preceded a roughly 49% loss before. If you want that trade, take it knowing what it is. If you do not, the crowding gets expensive now, when everyone already owns it - not later.