On the surface, the US economy looks strong, and financial markets have rebounded from the Trump-induced turmoil earlier this year. Yet beneath that strength, doubts linger. A booming stock market and solid growth are increasingly driven by a handful of AI-focused companies — raising the question: what happens if the AI hype fades? Could it trigger a recession, and if so, how deep might it run? In this analysis, I examine whether an AI bubble actually exists. In my next piece, I will explore what could happen if the market experiences a downturn.
The US stock market has surged in recent years, lifting global equity markets along with it. Or, more precisely, a handful of large US technology stocks have surged — and pulled the indices up with them.
The question on everyone’s mind these days is whether a bubble has formed, and if so, when (and how) it might burst. Nobody knows for sure, but we are clearly at a point where the question is worth asking. There are two dimensions to this discussion:
- Are we currently in a bubble?
- What would happen if it bursts?
This analysis focuses on the first question; my next analysis will address the second.
What is a stock-market bubble?
A stock-market bubble occurs when prices are driven primarily by expectations of further price increases rather than by fundamentals — in other words, when valuations can no longer be justified by rational asset pricing. When stock prices exceed their fundamental value, any further increase must stem from investors believing that there will be others willing to buy the stock later at an even higher, more overvalued price. That is what defines a bubble.
We cannot observe investors’ belief formation directly. Do they buy because they expect stocks to continue rising, even though they are already overvalued, i.e., a bubble, or because they believe current prices reflect fair, risk-adjusted valuations? We simply don’t know. As a result, we can never state with certainty that “we are in a bubble.” So – please – don’t expect that conclusion from me here.
What we can do, however, is evaluate whether stock prices appear high relative to their fundamentals. We cannot with certainty declare them “too high,” but we can observe when valuations are elevated. And right now, they are.
Enter the CAPE ratio
Let’s start with an old friend: the CAPE ratio – the cyclically adjusted price-earnings ratio developed by Yale Professor and Nobel Laureate Robert Shiller. CAPE has earned recognition for identifying major stock market overvaluations, notably during the run-up to the Great Depression in the late 1920s and the dot-com bubble at the turn of the millennium.
CAPE is constructed by dividing the current real, i.e., inflation-adjusted, stock price (the current real value of the S&P 500) by the average of the past ten years of inflation-adjusted earnings per share for the S&P 500. The measure extends back over 140 years, to 1881. Figure 1 shows how it has developed over time and where we are today.
Figure 1. CAPE (cyclically adjusted price-earnings ratio). Monthly data, Jan. 1881 – Sept. 2025. Source: R.Shiller’s webpage and J.Rangvid.
As you can see, we are today very close to an all-time high. The CAPE ratio stands at 40. This means that investors are paying forty times the average real earnings of the past decade to own the S&P 500. The all-time high was 44.2, reached in December 1999, right before the dot-com bubble burst. Today’s level is even higher than the peak observed at the start of the Great Depression in 1929. In terms of statistics, the current CAPE value is at the 1st percentile of its historical distribution—meaning it has been lower in 99% of months between 1881 and today. The only period with a higher stock-market valuation than today was from January 1999 to September 2000, after which the dot.com bubble burst.
You might object to using the past ten years of earnings as a measure of “cyclically adjusted” profits, arguing that a decade is too long a period. It’s often pointed out that a key difference between today’s market and the dot-com bubble of the late 1990s is that the tech companies driving the current surge actually earn substantial profits—unlike many tech firms back then. That’s true. So, when we divide today’s rapidly rising stock prices by slowly changing long-term earnings (rather than by more responsive, current earnings), we risk misrepresenting the situation. Fair enough. Let’s then use just last year’s earnings instead. That gives us a standard price-earnings ratio rather than the Shiller “cyclically adjusted” one. The result is shown in Figure 2.
Figure 2. Stock price-earnings ratio of the S&P 500. Monthly data, Jan. 1881 – Sept. 2025. June 2009 – May 2010 left out to enhance readability. Source: R.Shiller’s webpage and J.Rangvid.
This adjustment helps somewhat. The earnings of today’s fast-growing companies have also been rising rapidly. The market therefore appears less extreme using this measure. Still, while not as overstretched, as when measured via CAPE, the market remains highly valued: the current price-earnings ratio sits around the 95th percentile of its historical distribution.
Looking at other valuation metrics leads to the same conclusion. The U.S. stock market is expensive—perhaps not as extreme as during the dot.com bubble, but still elevated by historical standards. I’ll leave it to you to decide whether that distinction between the dot.com bubble and today feels reassuring or not.
Risk Compensation
Another way to gauge investors’ exuberance is to look at how much they are being compensated for taking stock-market risk. When investors accept low compensation for risk, it can signal a certain frothiness — a sense of “I really need to get on that stock-market wagon before it’s too late,” regardless of the risks involved.
We can assess this through the equity risk premium — the extra return investors demand for holding equities rather than safe assets. And here, too, the picture is telling: estimated risk premiums for the US stock market are low, around 2%.
I came across some interesting estimates of ex-ante risk premia from Absolute Strategy Research, which provide comparable figures for other regions as well. Figure 3 shows their estimates for the euro area and the United States.
The equity premium for the US is low in historical terms, and considerably lower than in the euro area. True, the equity premium was even lower just before the dot-com crash at the turn of the millennium, but – again – that is hardly something to aspire to.
Figure 3, together with Figures 1 and 2, imply that investors are buying US stocks at near record-high valuations even though they cannot reasonably expect to be well compensated for the risk that the market might fall.
Figure 3. Ex ante equity risk premium for the euro area and the US. Monthly data, Jan. 1990 – Sept. 2025. Source: Absolute Strategy Research via Datastream and J.Rangvid.
Of course, valuations and risk premiums are mechanically linked, at least under “all-else-equal” assumptions — when prices go up, expected returns go down, all else equal. Still, the point remains: investors today are willing to buy at high prices and accept low expected compensation for taking risk.
Robert Shiller’s own estimates of the US equity risk premium tell a similar story. His series also shows that the current premium is low — though, it has been even lower at other points in history, notably in 1929 and 2000.
Concentration risk
So, US stock-market valuations are high, and the expected risk premium is low. Hmmm…
Let’s add another piece of evidence suggesting that investors are taking on substantial risk at the moment: concentration risk.
A market rally is more fragile when it rests on the performance of a few large companies in one sector, rather than on broad gains across many industries. And, unfortunately, we are in the former situation today.
A few years ago, I wrote a piece showing how the performance of all 500 stocks in the S&P 500 was being driven by a handful of stocks (link) — the so-called “Magnificent Seven.”
Figure 4 shows how the combined market value of these seven giants has evolved since 2023, compared with the remaining 493 companies in the S&P 500, bringing the analysis up to the present day.
Figure 4. Total market value of Magnificent 7 companies and the remaining 493 companies in the S&P 500, both normalized to “1” on January 1, 2023. Daily data, Jan. 2023 – Nov. 2025. Source: Datastream via Refinitiv and J.Rangvid.
The 493 “other” stocks have done quite well: their total market value has increased from about USD 26 trillion in January 2023 to USD 39 trillion today — a gain of roughly 50%. A 50% rise in three years is impressive by any standard.
But the Magnificent Seven have been in a league of their own. Their combined market value has surged from USD 6.25 trillion in 2023 to USD 20 trillion today — an extraordinary 220% increase in less than three years.
This spectacular run means that these seven firms now account for more than one-third of the entire S&P 500, as shown in Figure 5.
Figure 5. Total market value of Magnificent 7 companies divided by total value of all companies in the S&P 500. Daily data, Jan. 2013 – Nov. 2025. Source: Datastream via Refinitiv and J.Rangvid.
Had the rally been broad-based, we probably wouldn’t be talking about a bubble at all. But a handful of firms have done so well that they have lifted the entire market’s valuation to near-record levels.
This concentration is, undeniably, a significant risk. If just one, two, or three of these mega-stocks were to disappoint, the consequences for the broader market could be significant.
As a simple illustration – and it is just an illustration – imagine Nvidia, with its nearly USD 5 trillion valuation, were to lose half its value. Suppose Apple and Microsoft, each worth roughly USD 4 trillion, did the same. That alone would wipe out about USD 6.5 trillion in market value. The total US stock market is worth around USD 58 trillion, so this would amount to a decline of more than 10% for the S&P 500 – from just three companies.
And, of course, such a correction would not stay contained. A sharp drop in the “Magnificent Seven” would almost certainly drag the rest of the market down as well.
You might think a 50% decline sounds extreme, but it is exactly what happened after the dot.com bubble burst.
The difficult thing in all this
Many indicators suggest that the market looks ripe for a correction – perhaps even a large one. The question, of course, is when.
You would have looked foolish if you had sold your stocks in early 2024 because you thought the market looked frothy and dangerously concentrated, as it did. The small group of leading stocks continued to surge, however, pushing valuations and market concentration to even more extreme levels. So is now really the right time to sell?
The same would have been true if you had sold in September 1998, after markets had dropped more than 10% in two months. At the time, such caution would have seemed entirely reasonable: the market had surged 150% in just over three years and appeared “expensive.” You might well have thought the correction signalled worse to come.
Instead, the market kept rising. Between September 1998 and September 2000, it gained another 44%.
So yes — timing is not everything, but it is a lot. And it takes considerable conviction (and patience) to stay out of a market that keeps defying gravity.
Prices rose more the dot.com era
One somewhat reassuring aspect of the current situation is that stock prices have not yet risen as dramatically as they did during the dot-com bubble. Figure 6 illustrates the performance of the S&P 500, measured in months since the beginning of the dot-com rally in January 1995 and since the onset of the current rally. I define the start of the current rally as October 2022, when stocks reached their post-pandemic low following a series of interest rate hikes. Choosing an earlier date—such as early 2020—would lead to essentially the same conclusion.
Figure 6. Development in the S&P 500 since January 1995 for the dot.com period and since October 2022 for the current period, both normalised to “1” in the first month. Monthly data. Source: Source: R.Shiller’s webpage and J.Rangvid.
As illustrated, the current S&P 500 trajectory has so far mirrored the pattern observed during the dot-com era, albeit over a shorter period. It remains to be seen whether this rally will continue along a similar path.
Finally, you might also argue that although valuations are close to an all-time high (Figure 1), the pace of the increase has been more moderate than during the dot-com era. In 1995, the CAPE ratio was around 20 and peaked at 45 in 2000—an increase of more than 100%. Five years ago, in 2020, it was about 25, and today it stands near 40—an increase of roughly 60%. On the other hand, as mentioned and unlike in 2000, today’s market is far more concentrated.
Conclusion
Could this market continue to rally? Who knows — maybe. At least for a while. Optimists may note that stock prices rose even more sharply during the dot-com era than they have so far in the current cycle.
But remember: the US market is trading at close to all-time-high valuations; the compensation for taking on risk is low; and the rally is driven by a very small number of stocks, making it fragile. One day, valuations will come down again.
They can do so in two ways — or a combination of both: earnings can rise, or stock prices can fall. Decades of academic research show that, for the US market at least, valuations usually come down because prices fall.
Perhaps this time will be different. Perhaps this handful of companies really will deliver on investors’ lofty expectations. But the phrase “this time is different” is a dangerous one. It was also said in the late 1990s. And, as history has taught us, “this time” often turns out not to be so different after all.
In my next post, I will explore what could happen if the market does correct — and what such a downturn could mean for investors and the global economy.
