Valuation comparison to 2000 dot-com bubble

Recently, debates over an AI bubble have intensified. From Bridgewater founder Ray Dalio to numerous Wall Street analysts, more people are beginning to question whether the AI-driven bull market is heading toward a bubble. In particular, rising capital expenditures and financing needs have led the market to worry about the sustainability of this rally. However, Howard Marks, Co-Chairman of Oaktree Capital, offers a different perspective. He believes that it is premature to conclude that the market has entered a true bubble phase. I personally love Howard's books "Cycle" and "The Most Important Thing," which can be considered my investment enlightenment bibles. Here, I quote Howard's views as corroboration for my own viewpoint from yesterday. In Marks' view, typical market bubbles usually have three characteristics: irrational exuberance, reckless investing, and a lack of investor alertness to risk. Although there is clear FOMO in the current market, investors are also continuously discussing AI valuations, capital expenditures, and financing risks. In other words, the market is not completely immersed in optimism. As for whether investment behavior has become universally irrational, he believes the existing signals are still complex and insufficient to draw a definitive conclusion. One key basis for Marks' judgment is the forward P/E ratio of tech stocks. Year-to-date, the forward P/E ratios of the S&P 500 and Nasdaq have actually declined. This does not mean stock prices fell; rather, the growth rate of corporate earnings expectations exceeded the pace of stock price increases, driving valuation multiples down. This difference is even more pronounced compared to the 2000 internet bubble. One month before the peak of that bubble, the average forward P/E ratio of the five largest tech companies by market cap in the S&P 500 was approximately 65x; today, that figure is around 23x. That is to say, despite significant gains in the stock prices of tech giants, their valuations remain far below those of the internet bubble era when viewed through the lens of expected earnings. Currently, Nvidia, Apple, and Microsoft combined account for about 21% of the weight in the S&P 500; whereas in 1999, nearly ten stocks were needed to achieve similar concentration. Meanwhile, narrowing market breadth and divergent performance within the tech sector have also exhibited some features previously seen during past bubble periods. However, investors find it difficult to easily abandon tech stocks. Over the past decade, the cumulative return of the Nasdaq has outperformed the S&P 500 by approximately 250 percentage points. Long-term excess returns have reinforced investor confidence in tech stocks, meaning that even as more people begin to worry about valuations, they may not be willing to actively reduce positions. This psychology itself could further concentrate funds into a few large tech companies. Determining whether an AI bubble exists cannot rely solely on how much stock prices have risen, nor can conclusions be drawn based on a single valuation metric. A relatively moderate forward P/E ratio does not mean financing risks can be ignored; ongoing market controversy does not imply that all investment behavior is sufficiently rational. My viewpoint in yesterday's article was clear: what investors truly need to focus on next are three questions: whether corporate earnings can meet current expectations, whether AI infrastructure investments will generate sufficient cash flow returns, and how much of this demand is built on sustainable commercial needs versus relying on financing-driven expansion. Judging a bubble requires combining valuation, capital flows, and market psychology, rather than simply equating rising stock prices with the formation of a bubble. For investors, identifying where risks are accumulating and whether current prices have left enough margin of safety for these risks is more important than debating whether AI constitutes a bubble.