A Bigger Bubble?

03 November 2025

A key feature of our macro framework is that business cycles tend to be non-linear. This is antithetical to the efficient market hypothesis. The non-linearity exists because of a pro-cyclical or reflexive feedback loop. When a positive feedback loop develops between an underlying trend and a misconception relating to that trend, it sets a boom-bust process in motion. The boom-bust process is then amplified by credit and leverage. From our perch, most preconditions of a typical bubble are evident in the current episode. However, we are sympathetic to an even larger dislocation from equilibrium. The big picture question is whether investors should keep on dancing. A few thoughts.

First, according to UBS the typical preconditions for a bubble are in place, 1) extended period of strong equity market performance in absolute terms and relative to bonds, 2) prevailing bias that it is different this time, 3) extended gap from the previous bubble (25 years), 4) aggregate profits (economy wide) under pressure, 5) high market concentration, 6) retail buying and 7) loose monetary conditions.

A buy-on-dip mentality occurs when equities outperform bonds by more than 5% a year for more than a decade. Equities have outperformed bonds by 14% a year over the past decade. The prevailing bias that it is different this time is due to AI. There has been a 25-year gap since the TMT bubble in 1998 and AI in 2023. A new generation of investors believe that it is different this time. Overall corporate profits as measured by the national accounts are under pressure relative to stock market earnings. That was a reliable divergence signal in the last two major episodes and is evident today (chart 1). Retail buying in equities is evident in many regions (India, United States & Korea). Broad financial conditions are loose as a goose.

Chart 1

Source: Bloomberg

Second, the key mark of a bubble is very clear overvaluation. In the 2000 episode, the NASDAQ peaked at 60 times earnings. In the Japanese stock market bubble in 1989, the Topix peaked at 72 times earnings. Today the NASDAQ is on 34.6 times earnings. Historically, the gap between the earnings yield and the 10-year bond yield also becomes much greater. Of course, as we noted in September, the MSCI US Information Technology sector trades at 10.8 times sales which is more stretched than in 2000 (chart 2).

Chart 2

Source: Bloomberg

To be fair, earnings revisions and growth is much better than the previous episode. Optimists contend that we are still early in the capital expenditure cycle for AI and there are fewer corporate excesses compared to historical episodes. On the negative side, profit margins are very high and will probably compress as business become more capital intensive and competition grows. Put another way, mega-cap technology companies are no longer capital lite. The major companies can still fund the capital investment out of cash flow. However, there has been an increase in debt by some (notably META) and circular vendor financing that was a feature of the 2000 episode. In the near term, greater capital intensity has created a positive feedback loop between rising investment spending and profits. For now, the hyper-scalers could increase capital expenditure by a further 40% before they fund out of debt. Corporate leverage is much lower today.

Third, another near term catalyst might be extreme M&A. For example, Vodafone/Mannesmann in February 2000 and AOL/Time Warner announced in January 2000. Both of those deals were approximately $900 billion in today’s dollars. The final point to note is loose financial conditions might warrant an even bigger bubble. In the current episode, the Federal Reserve has cut the funds rate over the past year. In the TMT bubble, the Fed hiked the funds rate to 6.5% in June 2000 and close to nominal GDP before the market peaked. In the short-term lower rates and the end of quantitative tightening might warrant higher equity prices. However, the deterioration in national accounts (economy wide) profits might be a signal of weakness in final demand and a bearish (recession) signal for stock market earnings.

In conclusion, we are sympathetic to an even larger bubble. There is a reflexive boom driven by capital expenditure in AI. The hyper-scalers have room to increase this by a further 40% before they fund out of debt. Financial conditions are loose and might ease further. A key constraint relative to prior episodes is fiscal dominance (public debt) and a potential re-acceleration in consumer price inflation (core inflation remains above the Fed’s target). We are also cognisant of the weakness in economy wide profits, consumer sentiment, and potential final demand. While valuations might not be as extreme as the 2000 and 1989 episodes based on price to earnings, risk compensation is poor. If you are going to dance, do it near the exit or in equities that do offer attractive valuation and growth characteristics.

20251103_A_Bigger_Bubble.pdf

Download PDF • 401KB