My mother asked me last week what I thought about an AI crash. I suspect she's not the only one asking.
Whenever a transformative technology arrives, markets have a habit of getting ahead of themselves. The excitement, the possibilities, and the tendency of investors to price in a perfect future before it has fully arrived — it's a pattern that repeats. I started my career as a financial advisor in 1998, right at the peak of the dot-com bubble — a period that ultimately crashed and ushered in what many call the "lost decade" for U.S. equities.
Here's a number that might scare you: if you had invested $1 in the S&P 500 at the market peak in March 2000 and tracked only the price index without reinvesting dividends, it would have been worth roughly $0.93 by March 2009. Even with dividends reinvested, you'd have ended that decade with somewhere between $1.09 and $1.15 — less than 1% annualized for ten years of risk. And along the way you would have seen that dollar fall to roughly $0.51 at the market trough in October 2002, when the S&P 500 had dropped approximately 49% from its peak.
A lot of people are feeling a sense of déjà vu for 1999 right now. I understand why. But I want to walk through what the valuation gauges are saying, where this moment resembles that one, and where it critically doesn't — and then talk about what actually matters for your portfolio if things do go sideways.
What the Valuation Gauges Are Saying
Professional investors use several tools to assess whether markets are expensive. Most of them are currently flashing yellow — or red. Here are four of the most important, and what each one is telling us.
The Shiller CAPE ratio (Cyclically Adjusted Price-to-Earnings), developed by Nobel laureate Robert Shiller, smooths earnings over 10 inflation-adjusted years to remove the distortion of any single boom or recession. The long-run average is about 17×. We're currently at approximately 41× — the second-highest reading in history. The only time it was higher was at the peak of the dot-com bubble in late 1999, when it briefly touched 44×.
That warrants sobriety. But here's the important distinction from 1999: back then, companies like Pets.com and Webvan were trading at enormous multiples with little or no actual earnings. Today's elevated CAPE is driven substantially by companies — Nvidia, Microsoft, Meta, Alphabet — generating extraordinary profits. The price is high, but so is the earnings power behind it. That's a different situation, even if it doesn't make the valuation comfortable.
This is the metric that tends to make serious bears most nervous, and for good reason: you can manage earnings with accounting choices, but revenue is hard to fake. The S&P 500's price-to-sales ratio — what investors are paying for each dollar of corporate revenue — currently sits at roughly 3.3×, about double its post-2000 average of ~1.7×.
This level of premium is only sustainable if corporate profit margins remain historically elevated. Right now, they are — and that's an important part of the bull case. But margin compression is the hidden risk embedded in this number. Higher wages, tariffs, increased competition, or an AI cost cycle that benefits buyers more than sellers could all squeeze margins, making today's price-to-sales ratio look far less justified in retrospect. One honest caveat: the S&P 500 has shifted dramatically toward high-margin software and technology businesses over the past two decades, which does structurally justify a higher P/S than older historical averages. A dollar of Microsoft's revenue is worth more than a dollar of a steel company's revenue. So "double the historical average" is a useful flag, not a precise alarm.
This chart is perhaps the most important for understanding why the current moment feels different from the post-2008 era, when valuations were also elevated but few people worried about it. The equity risk premium is what stocks pay you above what you'd earn risk-free in U.S. Treasury bonds — the extra compensation investors demand for bearing the uncertainty of equities.
In the decade after the financial crisis, with interest rates near zero, that premium was generous: stocks offered 3–4% more in earnings yield than bonds. It made sense to pay up for equities when the alternative was a Treasury yielding almost nothing. Today, with the 10-year Treasury at approximately 4.3% and the S&P 500's earnings yield around 3.2%, stocks are paying you less than risk-free bonds. That has historically been rare and associated with elevated vulnerability. Notice the red bars. The last sustained period like this? 1999 and 2000. Worth noting: more sophisticated versions of this calculation — which factor in expected earnings growth rather than just trailing earnings — paint a less alarming picture. If you believe analyst forecasts for AI-driven earnings growth, stocks look more competitive against bonds than this simple version suggests. Which version you trust depends on how much faith you put in those forecasts.
One additional metric worth noting, particularly for factor-based investors: price-to-book, which compares a company's market value to the accounting value of its assets. It's the foundational signal in value investing and in factor-based portfolio construction. On this measure, the U.S. market is also elevated — currently around 4.5–5×, well above its long-run average of roughly 2.5–3×. The important caveat is that book value has become a less reliable anchor as the economy has shifted toward intangible-heavy businesses: software, intellectual property, and brand equity don't show up on balance sheets, which means book value systematically understates the true worth of today's dominant companies. Where price-to-book is most useful right now isn't as a market-timing tool but as a geographic comparison — developed international and emerging markets are trading at a substantial discount to the U.S. on this measure, which is part of the long-run case for maintaining diversified global exposure.
The argument that "it's different this time" is always at least partially true. The world is constantly evolving, and valuation metrics have to evolve with it. The shift toward intangible assets changes what book value means. The globalization of corporate revenues changes what GDP comparisons mean. The rise of high-margin software businesses changes what price-to-sales means. None of these metrics is a static truth — they're imperfect tools, all of them, and serious investors know that.
More importantly: if any of these indicators were a reliable signal that a crash was imminent, the market would have already corrected. That's how markets work. The fact that valuations are elevated and the market hasn't crashed isn't denial — it's millions of investors collectively weighing the same data and concluding that the earnings story, or the growth trajectory, or the interest rate path, still justifies current prices. They may be wrong. But they may not be wrong yet — and trying to time an exit before a potential pullback has its own significant cost. The market could go up another 50% before it pulls back 30%. An investor who got out today, waiting for the crash that may or may not come, would need to be right twice: once getting out, and once getting back in. Very few people manage that. The evidence on market timing is brutal and consistent — it destroys more wealth than it protects.
So Is This 1999?
Yes and no — and the distinction matters enormously for how you think about your portfolio.
| Factor | Dot-Com Peak (1999–2000) | Today (2026) |
|---|---|---|
| Shiller CAPE | 44× — all-time record | ~41× — 2nd highest ever |
| Earnings behind the rally | Largely absent — many leaders had no profits | Real and substantial — Nvidia, Meta, Microsoft generating extraordinary cash |
| Interest rates | ~6.5% — bonds were a genuine alternative | ~4.3% — bonds are a genuine alternative again |
| Equity risk premium | Negative — stocks expensive vs. bonds | Near zero / slightly negative |
| Breadth of rally | Narrow — internet stocks driving everything | Narrow — "Magnificent 7" dominating returns |
| Path to profitability | Speculative — "we'll figure it out" | Demonstrated — AI revenues and margins are here today |
The honest summary: the valuation picture is comparably concerning to 1999 on most measures. What's different — and it is a significant difference — is that the companies driving this market are generating genuine, extraordinary profits. This is not a bubble built on hope alone. It's a market that has priced in a very optimistic future for companies that are already succeeding. The risk isn't that the earnings are fake. The risk is that the market has priced in a rate of growth that is extremely difficult to sustain — and that any disappointment could be painful.
Why the Word "Crash" Is Doing Investors a Disservice
A market crash is technically defined as a decline of 20% or more in a major index over a short period. The word "crash" is doing investors a disservice — it sounds like total devastation, no survivors, no recovery, permanent ruin. Financial history tells a different and more useful story.
The dot-com crash was severe. The Nasdaq fell approximately 78% from peak to trough between March 2000 and October 2002. The S&P 500 fell roughly 49%. If you were concentrated in U.S. large-cap growth and technology and you panicked and sold near the bottom, the damage was deep and took many years to recover from.
But here's what happened if you were well-diversified.
Imagine a 70/30 portfolio — 70% stocks, 30% bonds — structured as follows: of your equity allocation, one-third in international developed markets (MSCI EAFE), one-third in U.S. small and mid-cap stocks, one-third in U.S. large cap. The bond allocation tracks the Bloomberg U.S. Aggregate Bond Index. That same $1 invested at the March 2000 market peak:
October 2002
same period
nine years later
same period
The diversified portfolio still fell. It was still uncomfortable. But it fell roughly half as far at the trough, recovered far more quickly, and ended the period well ahead — at a point when the concentrated S&P 500 investor, even with dividends reinvested, had earned less than 1% a year for nine years.
Why? International stocks held up meaningfully better. U.S. small-cap and value stocks outperformed dramatically in the early 2000s, precisely when tech was imploding — a reminder that diversification across factors and geographies isn't just defensive, it can be a genuine return-generating engine working in your favor when growth stocks fall. And bonds rallied sharply, cushioning the fall and providing stability when equity markets were in freefall.
Diversification didn't just reduce volatility through the dot-com crash. For patient investors who didn't sell, it generated positive returns over the full decade — at the same time that the undiversified U.S. equity investor had nothing to show for ten years of patience. The key condition: not being forced to sell at the worst moment.
What This Means for You Right Now
Markets are expensive. That is an honest statement, supported by nearly every long-run valuation metric we have. If you are heavily concentrated in U.S. large-cap growth — particularly if you've ridden the AI wave and haven't rebalanced in a while — now is a reasonable time to think carefully about your positioning.
But "expensive" is not the same as "about to crash." Expensive markets can stay expensive for a long time, especially when the underlying earnings story holds up. No valuation metric is a reliable short-term timing tool. The CAPE ratio has been elevated by historical standards since the early 1990s, and investors who exited on that signal alone have missed extraordinary returns.
The more useful questions to sit with are these:
How concentrated are you? If the bulk of your equity exposure is in U.S. large-cap growth — particularly the handful of mega-cap AI names — you are more exposed to a repricing of those specific valuations than your overall allocation percentage might suggest.
Do you have meaningful international exposure? As the charts above show, MSCI EAFE and Emerging Markets are trading at significantly lower multiples than U.S. equities. Historically, valuation gaps of this magnitude eventually narrow — one way or another.
What is your time horizon? If you don't need to access this money for ten or more years, the historical evidence is clear: diversified, patient investors have navigated every previous crash and come out ahead. The investors who were genuinely hurt were those who sold into the downturn — converting a temporary paper loss into a permanent one.
Can you actually stay the course? This is the one that matters most and gets talked about least. The best portfolio in the world is useless if you can't psychologically hold it through a 30% drawdown. If the current level of market exposure is keeping you up at night, that's information worth acting on — not because the market is about to crash, but because a portfolio you can stay in is worth more than an optimal one you'll abandon at the worst moment.
I don't know if an AI crash is coming. I don't. And neither does anyone else, whatever confidence they project. What I do know is that the investors who fared best through the last great technology bubble weren't the ones who called the top. They were the ones who were diversified broadly enough that when the crash came, it didn't define their outcome.
That's still the playbook. It just requires the patience — and the portfolio construction — to stay in it.
The valuation metrics shown here are long-run tools, not short-term timing signals. Elevated readings have persisted for years at a time without a correction. No single metric — or even all four together — can tell you when a correction will happen, only that the margin for error has narrowed. Use them as context, not as a trigger.
