AI Bubble or Bull Run? What Investors Need to Know
Quick Summary
36% of the S&P 500 is concentrated in 7 stocks. Is the AI bubble real? Here's what history, data, and smart diversification tell us about managing the risk.
In This Article
The Numbers That Should Get Your Attention
Broad diversification is the foundational promise of index investing. Right now, that promise is under strain. Seven stocks account for 36% of the S&P 500 and 32% of the total US market — the highest concentration level recorded in US market history going back to 1927. At the same time, US stock valuations are approaching their 1999 peaks, the same peaks that preceded a decade of flat-to-negative returns for American equities.
So is this an AI bubble? And if it is, what should investors actually do about it?
The honest answer is that nobody knows whether we're in a bubble until it's already over. What we do know is that AI-related stocks have driven 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in November 2022, according to a September 2025 JP Morgan report. That is not noise. It's a structural shift in how the US market generates returns — and it demands serious analysis, not reflexive panic or uncritical optimism.
This article breaks down what concentration risk and elevated valuations actually mean for your portfolio, what history tells us about both, and where smart money tends to hide when crowded markets eventually correct.
What Market Concentration Actually Tells Us (And What It Doesn't)
Market concentration measures how much of a stock index's total value sits in a small number of companies. At current levels, the US market's top seven stocks carry more weight than at any point in nearly a century of data. That sounds alarming. The historical evidence, however, is more nuanced.
Analysis of US market data from 1926 to 2024 shows only a weak, statistically insignificant negative correlation between market concentration and subsequent 10-year returns. In plain terms: high concentration has not reliably predicted poor returns. The relationship is too noisy to act on alone.
Look beyond the US and the picture gets even more complicated. Across the 10 largest non-US developed stock markets, the average weight of the top seven stocks was 40.94% in November 2015 — higher than today's US figure. Switzerland sat at 60.11% concentration; Japan at 16.91%. Over the following decade, these markets delivered an average USD return of 8.44% annually. Taiwan, one of the most concentrated markets in that sample, actually outperformed the US over that period.
The takeaway: concentration alone is not a reliable signal to de-risk your portfolio. It raises valid questions, but it does not reliably predict crashes.
Key data points:
- Top 7 stocks = 36% of S&P 500 (highest since 1927)
- Non-US developed markets averaged 40.94% top-7 concentration in 2015
- Correlation between US concentration and 10-year returns: weak and statistically insignificant
Valuations Are the More Important Signal
While concentration grabs headlines, market valuations carry more predictive weight — and here the US market's current position is harder to dismiss.
The Cyclically Adjusted Price-Earnings ratio, or CAPE, measures stock prices against 10-year average real earnings. It smooths out short-term profit volatility to give a cleaner picture of whether markets are expensive relative to their long-run fundamentals. A high CAPE means investors are paying more for each unit of future earnings, which historically implies lower future returns.
Analysis of rolling 10-year return periods across the 10 largest developed markets going back to 1982 shows a clear monotonic relationship: higher starting CAPE ratios consistently correspond to lower subsequent returns. This relationship holds economically even if it struggles with statistical significance due to limited independent data samples.
US CAPE ratios are currently near their 1999 peaks. That alone does not mean a crash is imminent — valuations can stay elevated for extended periods, and high valuations have been followed by continued strong returns before, including in the US market between 2021 and 2025. But the data suggests investors should meaningfully moderate their return expectations for US equities from here.
What this means practically:
- High CAPE = lower expected future returns, not guaranteed losses
- The US market in 2021 had elevated valuations; it still delivered strong returns through 2025
- Valuation-based signals are most useful for long-run expectation-setting, not short-term timing
Three Historical Warnings Worth Studying
Three historical episodes provide the clearest lens through which to examine today's US market conditions.
Nortel and the Canadian Market (1999–2005)
Nortel Networks peaked at over 36% of the TSE 300, Canada's benchmark index, in 2000. Its CAPE ratio hit 60.6 — far exceeding the US market's own dot-com peak. When Nortel collapsed following a series of ill-advised acquisitions and the broader tech bust, the Canadian index fell 43% between September 2000 and September 2002. Brutal. But the Canadian market recovered fully by July 2005, less than three years later. And critically, Canadian value stocks — those with low prices relative to fundamentals — did not crash with the market and delivered even stronger returns through the recovery.
The US Dot-Com Bust (2000–2013)
The US market, despite being less concentrated than Canada, suffered a longer and arguably more damaging bust. Measured in Canadian dollar terms, the US market remained flat or below its March 2000 peak until July 2013 — over 13 years. Part of this was compounded by the 2008 financial crisis arriving just as prices were recovering. US small-cap value stocks, however, delivered positive returns throughout this period, while the broad market went nowhere.
Japan's Lost Decades (1990–present)
At the end of 1989, Japan was the world's largest stock market by capitalisation. Valuations had reached historic extremes. The crash that followed has never fully reversed in real, inflation-adjusted terms — over 35 years later, Japanese investors who bought at the peak have still not recovered their purchasing power. Two things would have protected them: geographic diversification into other global markets, and diversification within Japan itself into value and small-cap value stocks, which held up far better than the index.
The common thread across all three: broad diversification and a tilt toward value stocks provided meaningful protection in each of these episodes. This is not coincidence.
Why Bubbles Aren't Entirely Bad News
Before the doom narrative takes full hold, it's worth recognising that speculative technology bubbles have historically served a productive economic function — even when they destroy investor wealth.
The railway bubble of the 1840s funded the installation of vast rail networks across Britain and the US. The dot-com bubble of the late 1990s funded the construction of fiber optic infrastructure that now underpins the modern internet. In both cases, the companies building the infrastructure collapsed. But the infrastructure itself remained, enabling transformational economic growth that followed.
AI infrastructure spending today follows a recognisable pattern. Companies are raising enormous capital and deploying it into data centres, chips, and compute power at a blistering rate. Whether or not current stock prices are justified, the physical and digital infrastructure being built may well deliver economic value long after the speculative froth has cleared.
The implication for investors is not that AI is overhyped — it's that the economic benefits of a technology and the investment returns from the companies building it are two very different things.
How to Position a Portfolio Against Concentration and Valuation Risk
None of the historical evidence reviewed here suggests investors should exit the US market or abandon equity investing. What it does suggest is that a few structural adjustments can materially reduce tail risk without sacrificing long-run return potential.
Free Weekly Newsletter
Enjoying this guide?
Get the best articles like this one delivered to your inbox every week. No spam.
1. Geographic diversification The US makes up roughly 65–70% of global market cap indices. Investors overweighted to the US relative to a global benchmark are taking on concentrated valuation risk that history suggests may not be rewarded proportionally. International developed and emerging markets offer exposure to different economic cycles, valuation profiles, and currency dynamics.
2. Factor diversification — particularly value In each of the three historical episodes examined — Canada post-Nortel, the US post-dot-com, and Japan post-1989 — value stocks outperformed the broader market significantly. Value investing is not a guarantee, and it carries its own periods of underperformance. But as a structural complement to a market-cap-weighted index, it has a long record of reducing vulnerability during concentrated market selloffs.
3. Recalibrate return expectations The data on CAPE ratios and future returns is clear enough to warrant adjusting forward return assumptions for US equities. Financial plans built on the assumption that US equities will continue delivering 10–12% annual returns indefinitely may need stress-testing against scenarios of flat or modestly positive returns over the next decade.
4. Resist the behavioural trap Diversification's biggest enemy is not theory — it's psychology. Holding assets that are underperforming while your neighbour brags about Nvidia is genuinely difficult. The evidence is clear that diversification works, but it only works if you stay diversified through the periods when it feels pointless.
The Bottom Line on the AI Bubble
The AI bubble debate is, ultimately, unanswerable in real time. What is answerable is how to build a portfolio resilient enough to withstand a range of outcomes — including ones where the most expensive, most concentrated market in US history eventually reverts toward historical norms.
Concentration alone has not reliably predicted poor returns. Valuations have — modestly, imperfectly, but consistently enough to take seriously. And across every major historical episode of high valuations and concentration, from Japan to Canada to the US dot-com era, the investors who fared best were those who held diversified portfolios with meaningful exposure to value stocks and international markets.
The data does not say sell everything. It says: diversify deliberately, moderate your expectations, and hold the course — especially when it's uncomfortable.
This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial professional before making investment decisions.
Frequently Asked Questions
Is the US stock market currently in an AI bubble? No one can definitively answer this in real time — bubbles are only confirmed in hindsight. What is observable is that AI-related stocks have driven 75% of S&P 500 returns since November 2022, that the top seven stocks account for 36% of the index (the highest concentration since 1927), and that US CAPE valuations are near their 1999 peaks. Whether this constitutes a bubble depends on whether current valuations are ultimately justified by earnings. That question won't be answered until later.
Does high stock market concentration predict poor returns? The historical evidence suggests it does not — at least not reliably. Analysis of US data from 1926 to 2024 shows only a weak, statistically insignificant negative correlation between the concentration of the top seven stocks and subsequent 10-year returns. Globally, highly concentrated markets like Taiwan outperformed the US over a recent 10-year period. Concentration is a risk factor worth monitoring, but it is not a dependable predictor of crashes on its own.
What investment strategies have historically protected against bubble risk? Across three major historical episodes — Canada's Nortel crash, the US dot-com bust, and Japan's post-1989 collapse — two strategies consistently provided protection: geographic diversification and exposure to value stocks (companies with low prices relative to their fundamentals). In each case, investors in diversified global portfolios with value tilts significantly outperformed those holding concentrated market-cap-weighted positions. These are structural strategies, not timing calls.
Should I reduce my US stock market exposure right now? This article does not make investment recommendations, and neither should any single data point. What the evidence does suggest is that investors with very high US equity concentration relative to global markets may benefit from reviewing their geographic diversification. High CAPE valuations have historically been associated with lower subsequent 10-year returns — not crashes, but moderated expectations. Any changes to portfolio allocation should be made in consultation with a qualified financial adviser and in the context of your individual risk tolerance and time horizon.
Frequently Asked Questions
The Numbers That Should Get Your Attention
Broad diversification is the foundational promise of index investing. Right now, that promise is under strain. Seven stocks account for 36% of the S&P 500 and 32% of the total US market — the highest concentration level recorded in US market history going back to 1927. At the same time, US stock valuations are approaching their 1999 peaks, the same peaks that preceded a decade of flat-to-negative returns for American equities.
So is this an AI bubble? And if it is, what should investors actually do about it?
The honest answer is that nobody knows whether we're in a bubble until it's already over. What we do know is that AI-related stocks have driven 75% of S&P 500 returns, 80% of earnings growth, and 90% of capital spending growth since ChatGPT launched in November 2022, according to a September 2025 JP Morgan report. That is not noise. It's a structural shift in how the US market generates returns — and it demands serious analysis, not reflexive panic or uncritical optimism.
This article breaks down what concentration risk and elevated valuations actually mean for your portfolio, what history tells us about both, and where smart money tends to hide when crowded markets eventually correct.
What Market Concentration Actually Tells Us (And What It Doesn't)
Market concentration measures how much of a stock index's total value sits in a small number of companies. At current levels, the US market's top seven stocks carry more weight than at any point in nearly a century of data. That sounds alarming. The historical evidence, however, is more nuanced.
Analysis of US market data from 1926 to 2024 shows only a weak, statistically insignificant negative correlation between market concentration and subsequent 10-year returns. In plain terms: high concentration has not reliably predicted poor returns. The relationship is too noisy to act on alone.
Look beyond the US and the picture gets even more complicated. Across the 10 largest non-US developed stock markets, the average weight of the top seven stocks was 40.94% in November 2015 — higher than today's US figure. Switzerland sat at 60.11% concentration; Japan at 16.91%. Over the following decade, these markets delivered an average USD return of 8.44% annually. Taiwan, one of the most concentrated markets in that sample, actually outperformed the US over that period.
The takeaway: concentration alone is not a reliable signal to de-risk your portfolio. It raises valid questions, but it does not reliably predict crashes.
Key data points:
- Top 7 stocks = 36% of S&P 500 (highest since 1927)
- Non-US developed markets averaged 40.94% top-7 concentration in 2015
- Correlation between US concentration and 10-year returns: weak and statistically insignificant
Valuations Are the More Important Signal
While concentration grabs headlines, market valuations carry more predictive weight — and here the US market's current position is harder to dismiss.
The Cyclically Adjusted Price-Earnings ratio, or CAPE, measures stock prices against 10-year average real earnings. It smooths out short-term profit volatility to give a cleaner picture of whether markets are expensive relative to their long-run fundamentals. A high CAPE means investors are paying more for each unit of future earnings, which historically implies lower future returns.
Analysis of rolling 10-year return periods across the 10 largest developed markets going back to 1982 shows a clear monotonic relationship: higher starting CAPE ratios consistently correspond to lower subsequent returns. This relationship holds economically even if it struggles with statistical significance due to limited independent data samples.
US CAPE ratios are currently near their 1999 peaks. That alone does not mean a crash is imminent — valuations can stay elevated for extended periods, and high valuations have been followed by continued strong returns before, including in the US market between 2021 and 2025. But the data suggests investors should meaningfully moderate their return expectations for US equities from here.
What this means practically:
- High CAPE = lower expected future returns, not guaranteed losses
- The US market in 2021 had elevated valuations; it still delivered strong returns through 2025
- Valuation-based signals are most useful for long-run expectation-setting, not short-term timing
Three Historical Warnings Worth Studying
Three historical episodes provide the clearest lens through which to examine today's US market conditions.
Nortel and the Canadian Market (1999–2005)
Nortel Networks peaked at over 36% of the TSE 300, Canada's benchmark index, in 2000. Its CAPE ratio hit 60.6 — far exceeding the US market's own dot-com peak. When Nortel collapsed following a series of ill-advised acquisitions and the broader tech bust, the Canadian index fell 43% between September 2000 and September 2002. Brutal. But the Canadian market recovered fully by July 2005, less than three years later. And critically, Canadian value stocks — those with low prices relative to fundamentals — did not crash with the market and delivered even stronger returns through the recovery.
The US Dot-Com Bust (2000–2013)
The US market, despite being less concentrated than Canada, suffered a longer and arguably more damaging bust. Measured in Canadian dollar terms, the US market remained flat or below its March 2000 peak until July 2013 — over 13 years. Part of this was compounded by the 2008 financial crisis arriving just as prices were recovering. US small-cap value stocks, however, delivered positive returns throughout this period, while the broad market went nowhere.
Japan's Lost Decades (1990–present)
At the end of 1989, Japan was the world's largest stock market by capitalisation. Valuations had reached historic extremes. The crash that followed has never fully reversed in real, inflation-adjusted terms — over 35 years later, Japanese investors who bought at the peak have still not recovered their purchasing power. Two things would have protected them: geographic diversification into other global markets, and diversification within Japan itself into value and small-cap value stocks, which held up far better than the index.
The common thread across all three: broad diversification and a tilt toward value stocks provided meaningful protection in each of these episodes. This is not coincidence.
Why Bubbles Aren't Entirely Bad News
Before the doom narrative takes full hold, it's worth recognising that speculative technology bubbles have historically served a productive economic function — even when they destroy investor wealth.
The railway bubble of the 1840s funded the installation of vast rail networks across Britain and the US. The dot-com bubble of the late 1990s funded the construction of fiber optic infrastructure that now underpins the modern internet. In both cases, the companies building the infrastructure collapsed. But the infrastructure itself remained, enabling transformational economic growth that followed.
AI infrastructure spending today follows a recognisable pattern. Companies are raising enormous capital and deploying it into data centres, chips, and compute power at a blistering rate. Whether or not current stock prices are justified, the physical and digital infrastructure being built may well deliver economic value long after the speculative froth has cleared.
The implication for investors is not that AI is overhyped — it's that the economic benefits of a technology and the investment returns from the companies building it are two very different things.
How to Position a Portfolio Against Concentration and Valuation Risk
None of the historical evidence reviewed here suggests investors should exit the US market or abandon equity investing. What it does suggest is that a few structural adjustments can materially reduce tail risk without sacrificing long-run return potential.
1. Geographic diversification The US makes up roughly 65–70% of global market cap indices. Investors overweighted to the US relative to a global benchmark are taking on concentrated valuation risk that history suggests may not be rewarded proportionally. International developed and emerging markets offer exposure to different economic cycles, valuation profiles, and currency dynamics.
2. Factor diversification — particularly value In each of the three historical episodes examined — Canada post-Nortel, the US post-dot-com, and Japan post-1989 — value stocks outperformed the broader market significantly. Value investing is not a guarantee, and it carries its own periods of underperformance. But as a structural complement to a market-cap-weighted index, it has a long record of reducing vulnerability during concentrated market selloffs.
3. Recalibrate return expectations The data on CAPE ratios and future returns is clear enough to warrant adjusting forward return assumptions for US equities. Financial plans built on the assumption that US equities will continue delivering 10–12% annual returns indefinitely may need stress-testing against scenarios of flat or modestly positive returns over the next decade.
4. Resist the behavioural trap Diversification's biggest enemy is not theory — it's psychology. Holding assets that are underperforming while your neighbour brags about Nvidia is genuinely difficult. The evidence is clear that diversification works, but it only works if you stay diversified through the periods when it feels pointless.
The Bottom Line on the AI Bubble
The AI bubble debate is, ultimately, unanswerable in real time. What is answerable is how to build a portfolio resilient enough to withstand a range of outcomes — including ones where the most expensive, most concentrated market in US history eventually reverts toward historical norms.
Concentration alone has not reliably predicted poor returns. Valuations have — modestly, imperfectly, but consistently enough to take seriously. And across every major historical episode of high valuations and concentration, from Japan to Canada to the US dot-com era, the investors who fared best were those who held diversified portfolios with meaningful exposure to value stocks and international markets.
The data does not say sell everything. It says: diversify deliberately, moderate your expectations, and hold the course — especially when it's uncomfortable.
This article is for informational purposes only and does not constitute financial advice. Always consult a qualified financial professional before making investment decisions.
Frequently Asked Questions
Is the US stock market currently in an AI bubble? No one can definitively answer this in real time — bubbles are only confirmed in hindsight. What is observable is that AI-related stocks have driven 75% of S&P 500 returns since November 2022, that the top seven stocks account for 36% of the index (the highest concentration since 1927), and that US CAPE valuations are near their 1999 peaks. Whether this constitutes a bubble depends on whether current valuations are ultimately justified by earnings. That question won't be answered until later.
Does high stock market concentration predict poor returns? The historical evidence suggests it does not — at least not reliably. Analysis of US data from 1926 to 2024 shows only a weak, statistically insignificant negative correlation between the concentration of the top seven stocks and subsequent 10-year returns. Globally, highly concentrated markets like Taiwan outperformed the US over a recent 10-year period. Concentration is a risk factor worth monitoring, but it is not a dependable predictor of crashes on its own.
What investment strategies have historically protected against bubble risk? Across three major historical episodes — Canada's Nortel crash, the US dot-com bust, and Japan's post-1989 collapse — two strategies consistently provided protection: geographic diversification and exposure to value stocks (companies with low prices relative to their fundamentals). In each case, investors in diversified global portfolios with value tilts significantly outperformed those holding concentrated market-cap-weighted positions. These are structural strategies, not timing calls.
Should I reduce my US stock market exposure right now? This article does not make investment recommendations, and neither should any single data point. What the evidence does suggest is that investors with very high US equity concentration relative to global markets may benefit from reviewing their geographic diversification. High CAPE valuations have historically been associated with lower subsequent 10-year returns — not crashes, but moderated expectations. Any changes to portfolio allocation should be made in consultation with a qualified financial adviser and in the context of your individual risk tolerance and time horizon.
About Zeebrain Editorial
Zeebrain publishes independent analysis of markets, investing, personal finance, and business. We disclose affiliate relationships, never accept payment for coverage, and fact-check all claims against primary sources. Read our editorial policy →
Disclaimer: Content on Zeebrain is for informational and educational purposes only and does not constitute financial advice or a recommendation to buy or sell any security. Always conduct your own research and consult a qualified financial adviser before making investment decisions. Past performance is not indicative of future results.
More from Business & Money
Related Guides
Keep exploring this topic
Should SpaceX, OpenAI and Anthropic Join the S&P 500?
Business & Money · S&P 500 · index investing
5 ETFs That Beat the S&P 500 and Build Wealth Faster
Business & Money · ETF investing · S&P 500
Market Breadth Warning: Why Index Highs May Hide Portfolio Risk
Business & Money · market breadth · NASDAQ
Federal Reserve interest rates: Impact on businesses and investments
Business & Money
Explore More Categories
Keep browsing by topic and build depth around the subjects you care about most.