Understanding the AI Bubble: Market Concentration, Valuations, and Diversification
Ben Felix
Summary:
This video discusses investor concerns about a potential "AI bubble" driven by high US market concentration and valuations, offering historical context and protective strategies.
- The S&P 500's top seven stocks now account for 36% of its value, an unprecedented level of concentration since 1927.
- US stock market valuations are nearing 1999 dot-com peaks, which preceded a decade of flat returns.
- Historically, market bubbles driven by new technologies like railroads and the internet involved massive investment, high asset prices, and subsequent crashes.
- AI-related stocks have significantly contributed to recent S&P 500 returns, earnings growth, and capital expenditure since ChatGPT's launch.
- While market concentration has a weak link to future returns, high market valuations generally predict lower future returns.
- Diversification across different markets (e.g., global stocks) and within markets (e.g., value and small-cap stocks) has historically mitigated the impact of market crashes, as seen in the Canadian Nortel bust and Japan's lost decade.
- Investors should maintain diversification and discipline, accepting that portfolios will always contain both winners and underperformers.
Concerns About the "AI Bubble" [0:00]
The video addresses growing investor worries about a potential "AI bubble," characterized by extreme market concentration and high valuations in the U.S. stock market.
Extreme Market Concentration [0:00]
- The S&P 500 index currently has 36% of its value concentrated in just seven stocks.
- The total U.S. market shows a 32% concentration in these top stocks.
- This represents the most extreme level of index concentration in U.S. market history, dating back to 1927.
High Market Valuations [0:21]
- U.S. stock market valuations are approaching their 1999 dot-com bubble peaks.
- The 1999 peak was followed by a decade of flat U.S. stock returns.
- A significant decline in these highly valued, concentrated stocks could substantially impact the overall market.
Historical Context of Bubbles [1:11]
The speaker clarifies that whether the current situation is an "AI bubble" is only discernible in hindsight but points to historical parallels.
The Nature of Stock Price Bubbles [1:49]
- Bubbles are periods of unusually high stock prices followed by much lower prices.
- They are often, but not always, triggered by new technologies promising huge profits.
- The history of technology bubbles dates back to at least the 1700s, following a similar pattern with each major technological innovation.
Productive Aspects of Bubbles [3:12]
- High stock prices fueled by speculation can facilitate the development and deployment of revolutionary technologies.
- Examples include the massive spending on fiber optic cables in the late 1990s and railway tracks in the 1840s.
- Despite the waste associated with bubbles (e.g., unused fiber optic cable, redundant railway track), essential infrastructure is created, potentially paving the way for economic growth.
- These "productive bubbles" are generally beneficial for the economy, even if they can be painful for individual investors.
The Cycle of Technological Bubbles [4:07]
- Investor excitement and high stock prices around technological revolutions follow a consistent pattern.
- Stock prices are driven by high profit potential and speculation.
- Eventually, prices revert, leading to low returns for those who bought near the peak.
- The rapid rise in prices of top U.S. stocks has been accompanied by substantial earnings growth, indicating some economic substance beyond pure hype.
- Since the launch of Chat GPT in November 2022, AI-related stocks have accounted for:
- 75% of S&P 500 returns.
- 80% of earnings growth.
- 90% of capital spending growth.
Lessons from Past Market Events [5:32]
The Nortel Example (Canada, 2000) [5:49]
- Nortel Networks, a telecommunications company, peaked at over 36% of the Canadian stock market index (TSSE 300) in July 2000.
- Canadian market valuations, influenced by Nortel, reached a Schiller CAPE ratio of 60.62, exceeding the U.S. dot-com peak.
- Nortel's downfall began with unprofitable acquisitions and was exacerbated by the dot-com bubble burst.
- The TSSE 300 index dropped by 43% between September 2000 and September 2002.
- Recovery and Resilience:
- The market recovered by July 2005 and delivered strong returns despite the initial crash.
- Canadian value stocks did not crash and delivered stronger returns post-recovery.
The Dot-Com Bust (U.S., 2000s) [8:17]
- The U.S. market in 1999 had high stock prices, largely unjustified by fundamentals, though less extreme concentration than Canada's Nortel period.
- The dot-com bubble burst led to a "lost decade" for U.S. stocks.
- The S&P 500 remained flat or below flat until July 2013, a brutal period for investors who bought at the peak.
- Mitigating Factors:
- Investors in U.S. value stocks and small-cap value stocks fared much better, earning positive returns while the broader market was flat.
The Japanese "Lost Decades" (1990s onward) [13:45]
- The Japanese stock market experienced a massive boom leading up to 1990, becoming the world's largest by market capitalization with exceptionally high valuations.
- The market crashed at the end of 1989 and has not recovered in real (inflation-adjusted) terms to this day.
- Survival Strategies:
- Diversification across markets: A globally diversified investor would have done well as the U.S. market subsequently outperformed.
- Diversification within the Japanese market: Japanese value and small-cap value stocks performed adequately despite the overall market's struggles.
Market Concentration vs. Valuations [10:01]
Market Concentration [10:01]
- Market concentration measures how much of the market's total value is held by a small number of stocks.
- The relationship between market concentration and future returns is statistically weak and economically noisy.
- Looking at U.S. market data from 1926, there's a slight negative correlation, but it's not statistically significant.
- Many other global markets are more concentrated than the U.S. market, yet still deliver positive returns.
- Example: Taiwan, one of the most concentrated markets in November 2015, outperformed the U.S. market in the subsequent decade.
- Periods of falling concentration in the S&P 500 (post-1950) have seen less positive but not disastrous returns.
Market Valuations (CAPE Ratio) [12:38]
- Market valuations (e.g., Schiller CAPE ratio) measure how expensive it is to buy expected future earnings.
- There is a clear monotonic relationship between starting CAPE ratio and future 10-year returns: higher starting valuations are associated with lower future returns.
- This relationship is economically strong, though statistically challenging to prove due to limited independent samples.
- High valuations do not guarantee an immediate crash, but they suggest moderating expectations for future U.S. market returns.
- The U.S. market has historically shown that high valuations can be followed by continued high returns, even if this is an unlikely outcome according to historical patterns.
Conclusion: Diversification and Discipline [17:15]
- The main lessons are diversification and discipline.
- A properly diversified investor should be comfortable with their portfolio through varying market conditions, accepting that it will always hold both winning and losing assets.
- The understanding that winners will outperform losers in the long run, and maintaining discipline, are crucial for long-term investment success.