The bull market in US stocks enters its fourth year: Taking history as a lesson, will the "AI bubble burst" or "this time is really different"?

As artificial intelligence trading continues to drive US stocks to record highs, the market is increasingly questioning whether it is experiencing another financial bubble destined to burst. But the answer is not so simple, at least from a historical perspective. In 2025, the S&P 500 index rose by 16%, with the winners in the field of artificial intelligence including NVIDIA Corporation(NVDA.US), Alphabet(GOOGL.US), Broadcom Inc.(AVGO.US), and Microsoft Corporation(MSFT.US) contributing the most. However, at the same time, there is growing concern about whether the billions of dollars that large tech companies are committing to invest in artificial intelligence infrastructure will bring matching returns. Data shows that capital expenditures for companies like Microsoft Corporation, Alphabet, Amazon.com, Inc.(AMZN.US), and Meta Platforms(META.US) are expected to grow by 34% in the next year, totaling around $440 billion. Meanwhile, OpenAI has pledged to invest over $1 trillion in artificial intelligence infrastructure, a staggering figure for a non-profitable private company. But perhaps more concerning is that many of OpenAI's partnership arrangements have a circular nature, with investments and expenditures flowing back and forth between OpenAI and several public tech giants. Brian Levitt, Chief Global Market Strategist at Invesco Ltd., pointed out that throughout history, excessive investment is often seen when there is technological progress that can change society. He cited examples such as railways, electricity, and the internet. This time may not be an exception. He said, "Sometimes infrastructure development may exceed the short-term needs of economic development. But that doesn't mean the railways weren't built, or the internet didn't emerge, right?" Despite this, as stock valuations continue to climb, the S&P 500 index has just achieved double-digit percentage gains for the third consecutive year, investors are starting to worry about how much more room for growth there is in the field of artificial intelligence, and how much market value could be lost if artificial intelligence fails to meet expectations. NVIDIA Corporation, Microsoft Corporation, Alphabet, Amazon.com, Inc., Broadcom Inc., and Meta Platforms account for nearly 30% of the weight in the S&P 500 index, so a sell-off in artificial intelligence could have a significant impact on the index. Gene Goldman, Chief Investment Officer at Cetera Financial Group, who does not believe there is a bubble in artificial intelligence stocks, said, "Bubbles are likely to burst in a bear market. We just don't see a bear market coming soon." Taking a lesson from history, what are the similarities and differences between the current artificial intelligence boom and past market bubbles? Gains and Duration One simple way to measure whether the surge in technology stocks driven by artificial intelligence is excessive or too quick is to compare it to past bull markets. Research by Bank of America Corp strategist Michael Hartnett shows that looking back at the 10 global market bubbles since 1900, the average duration slightly exceeded two and a half years, with gains of up to 244% from trough to peak. In contrast, the current uptrend driven by artificial intelligence has entered its third year, with the S&P 500 index rising by 79% since the end of 2022, and the tech-heavy Nasdaq 100 index rising by 130%. Although it is difficult to draw conclusions from data, Hartnett warns investors that even if they believe there is a bubble in the stock market, they should not sell their stocks, as the final stages of a rally are usually the most significant, and the cost of missing out on an opportunity will be very high. He suggests that a way to hedge risk is to buy value stocks that are priced cheaply, such as UK stocks and energy companies. Concentration The top 10 largest stocks by market value in the S&P 500 index currently account for about 40% of the index, a level of concentration not seen since the 1960s. This has raised concerns among some investors, including Wall Street veteran research expert Ed Yardeni. He stated last December that it no longer makes sense to advise investors to overweight tech stocks. Market historians believe that although the current level of concentration seems high compared to recent times, there are precedents. Paul Marsh, a professor at the London Business School, studied global asset returns over the past 125 years and noted that the percentage of top stocks in the US market reached similar levels in the 1930s and 1960s. Marsh said that in 1900, 63% of the total market value in the US was related to railway stocks, whereas by the end of 2024, this ratio was only 37%. Fundamentals Dario Perkins, an economist at TS Lombard, said that asset bubbles are often harder to spot in real-time than in hindsight because fundamentals are usually the focus of debate, and the indicators that investors are paying attention to may change. He said, "Tech enthusiasts are quick to claim 'this time it's different' and that fundamentals and valuations will never be the same again." However, some fundamental factors remain crucial. For example, compared to the dot-com bubble era, the debt-to-earnings ratio for today's artificial intelligence giants is lower than companies like WorldCom Inc. And companies like NVIDIA Corporation and Meta Platforms have reported strong profit growth from their artificial intelligence businesses, which was not the case during the speculative era 25 years ago. Potential credit risks in artificial intelligence trading are causing concern among some investors. After Oracle Corporation(ORCL.US) issued $18 billion in bonds on September 24th, its stock price dropped by 5.6% the next day, and has since fallen by a total of 37%. According to estimates by Industrial Bank of France, Meta, Alphabet, and Oracle Corporation alone will need to raise $86 billion in funding in 2026. Valuation The valuation of the S&P 500 index has reached its highest levels since the early 2000s, at least according to its cyclically adjusted price-to-earnings ratio (Shiller P/E ratio). This ratio, invented by economist Robert Shiller, divides stock prices by the average inflation-adjusted earnings over the past 10 years. Bullish investors believe that while tech stocks have driven the market valuation higher, the pace is far lower than during the dot-com bubble era. In 2000, Cisco Systems, Inc.(CSCO.US) had a P/E ratio that exceeded 200 times the past 12 months' earnings, while NVIDIA Corporation's P/E ratio is currently below 50 times. Richard Clode, a fund manager at Janus Henderson, said that in an environment where valuations are not controversial, stock prices can decouple from earnings growth. He said, "We haven't seen that yet." Investor scrutiny Discussions about a potential stock market bubble have been ongoing throughout the year, but in November and December, with warnings from prominent investors like Michael Burry, a prototype from the movie "The Big Short," and the Bank of England, the discussions have significantly intensified. Data shows that in November, there were more than 12,000 news reports mentioning the term "artificial intelligence bubble," roughly equal to the total in the previous ten months. A survey by Bank of America Corp in December showed that investors see the artificial intelligence bubble as the biggest "tail risk" event. More than half of the respondents said that the "seven tech giants" are the most crowded trade on Wall Street. This is in stark contrast to the dot-com bubble era, as Venu Krishna, Head of US Stock Strategy at Barclays PLC Sponsored ADR, said people were "incredibly excited" about how the internet would fundamentally change everything at the time. With the increasing issuance of debt, doubts about the returns from investing in artificial intelligence are also growing. He said, "I wouldn't ignore this, but I generally think this kind of scrutiny is beneficial. In fact, it is this scrutiny that can prevent extreme behavior like a collapse."