What Financial Bubbles Reveal About Today’s Stock Market 

By Oliver Müller, Chief Investment Officer, Accresco Investment Management

Financial bubbles rarely begin with irrationality. They usually begin with genuine innovation, compelling economic narratives and the belief that this time the old rules no longer apply. Drawing on the lessons of history – from the dot-com boom to today’s AI-driven market – this article explores why innovation and speculation so often travel together, what current valuation signals suggest, and why disciplined investing matters most when optimism is at its peak. The challenge for investors is not recognising the future – it is determining what that future is worth today.

At the beginning of my career in equity markets, I arrived in London in early 2000 at a moment when the language of investing was changing rapidly. Traditional references to valuation, cash flows and competitive advantage were being displaced by concepts such as “new economy dynamics,” “eyeballs,” and “total addressable markets.” The underlying assumption was not simply that the world was changing, but that the scale of that change rendered conventional valuation frameworks increasingly obsolete.

A week after I joined Merrill Lynch, the Nasdaq peaked. At the time, of course, there was no way to know this. Market tops are never visible in real time. What is far more revealing is the intellectual environment that surrounds them: bubbles rarely emerge from ignorance, but from coherent narratives that gradually detach from the discipline of valuation.

“Passive investing allocates capital according to what the market has already decided. Active investing continually asks whether the market’s judgement remains correct.”

Financial bubbles have been studied for more than a century, yet the conclusions across the literature are strikingly consistent. From Charles Mackay’s Extraordinary Popular Delusions and the Madness of Crowds, through John Kenneth Galbraith’s The Great Crash 1929, Charles P. Kindleberger’s Manias, Panics and Crashes, and Robert J. Shiller’s work in Irrational Exuberance and Narrative Economics, the central pattern repeats with remarkable discipline across time and asset classes.

Bubbles rarely begin with speculation. They begin with something real: a technological breakthrough, a financial innovation, or a genuine shift in economic structure. What changes is not the validity of the initial insight, but the way in which it is translated into asset prices.

Kindleberger’s contribution was to emphasise the role of credit and liquidity in amplifying these cycles. Easy financing extends the lifespan of narratives by allowing speculation to persist beyond what fundamentals alone would support. Galbraith was equally explicit: financial sophistication often masks a simple expansion of leverage.

Shiller added a different but complementary dimension. Markets, he argued, are not merely inefficient in a technical sense; they are narrative systems. Once a compelling story takes hold, it becomes self-reinforcing. Rising prices validate the narrative, which attracts further capital, which in turn reinforces prices. In this phase, valuation does not disappear, but it gradually ceases to constrain behaviour.

Across these different perspectives, a common structure emerges. Bubbles are not defined by irrationality in the simple sense. They are defined by the progressive separation of price from discipline, enabled by narrative, liquidity and feedback loops between the two.

“Innovation creates opportunity. Valuation determines outcomes. One without the other is an incomplete investment thesis.”

The dot-com boom remains one of the clearest modern examples of how financial bubbles evolve. It is often portrayed as an episode of collective irrationality. That interpretation is too simplistic. The internet represented one of the most important technological breakthroughs of the twentieth century and ultimately transformed almost every aspect of the global economy. Investors were right about the technology. Their mistake was in assuming that an extraordinary future justified almost any price in the present.

Shortly after joining Merrill Lynch, I attended a major technology, media and telecoms conference in London. One presentation in particular remains vivid. It argued that traditional valuation methods – discounted cash flow analysis, earnings multiples, and comparable frameworks – could account for only a portion of the market’s value. The remainder, it suggested, lay in the vast “white space” of future opportunity created by technological change.

The argument was compelling and widely accepted. But it marked a subtle yet profound shift in investment thinking. Rather than using uncertainty about the future to refine valuation, investors increasingly used it to justify abandoning valuation altogether. In effect, the greater the perceived opportunity, the less constrained prices became. The possibility of extraordinary future outcomes was increasingly invoked to justify almost any price in the present.

As more and more capital flowed into the technology sector, this narrative strengthened rather than weakened. Rising valuations were interpreted as validation of the underlying thesis. Loss-making business models were financed on the assumption that scale would eventually resolve economics that had not yet been demonstrated. Valuation discipline was increasingly treated as backward-looking, while narrative alignment became forward-looking.

The adjustment, when it came, was severe rather than sudden. Investors had not simply overestimated future prospects; they had underestimated the uncertainty surrounding them. Valuations came to reflect a narrow set of highly optimistic outcomes. As markets began to price a broader range of possibilities, those valuations could no longer be sustained. The technology was ultimately vindicated. Many of the valuations were not.

What relevance, then, does the dot-com experience have for investors today? The obvious question is whether today’s market is different. In one sense, it clearly is. Artificial intelligence is a different technology, many of today’s leading companies are larger and more profitable, and the macroeconomic backdrop has changed considerably. But the purpose of studying bubbles is not to search for identical circumstances. It is to recognise recurring patterns in the relationship between innovation, narrative and valuation. Viewed through that lens, the parallels become difficult to ignore.

The most obvious place to look is the semiconductor sector. In both the dot-com era and the current AI cycle, semiconductors have provided the critical infrastructure for a technological revolution. The current AI cycle is now approximately 44 months old – almost exactly the point at which the semiconductor boom of the late 1990s reached its peak. When the two cycles are aligned from their respective lows, the resemblance is striking, not only in terms of elapsed time but also in the magnitude and acceleration of returns. While no two market cycles unfold identically, the similarities are sufficiently pronounced to warrant careful attention.

A second chart examines the semiconductor rally from a different perspective. The strongest monthly gain in the PHLX Semiconductor Index over the past thirty years occurred in February 2000, one month before the dot-com bubble reached its peak. The second strongest occurred in April 2026.

One data point, of course, does not establish a pattern. But periods of exceptionally strong monthly returns have historically tended to occur late in powerful market cycles, when investor conviction is greatest and expectations have become increasingly one-sided. They deserve attention not because they signal an imminent turning point, but because they suggest that markets have become increasingly dependent on the continued delivery of exceptionally favourable outcomes.

Valuation provides another important perspective. The cycle-adjusted price-to-earnings ratio (“CAPE”) of the global semiconductor sector now stands at more than four times its long-term median – the highest relative valuation of any major global equity sector. Software and hardware companies also trade at substantial premiums to their historical norms, while most other sectors remain much closer to long-term averages.

The implication is not simply that technology has become expensive. Rather, it suggests that investors have become increasingly willing to concentrate capital behind a relatively narrow set of expectations surrounding artificial intelligence. In other words, the current market is characterised not only by elevated valuations, but by an unusually high degree of valuation dispersion across sectors.

The same picture emerges at the level of the broader market. The cyclically adjusted price-to-earnings ratio of the US equity market currently stands at around 41 – its second-highest reading in almost 150 years of recorded history, exceeded only at the height of the dot-com bubble in 2000. It is higher than before the crashes of 1929, 1987 and the Global Financial Crisis.

On its own, this does not imply that a market reversal is imminent. Valuation has always been a poor timing tool. It is, however, one of the most reliable indicators of long-term return potential. History shows a clear inverse relationship between starting valuations and subsequent ten-year real returns. At CAPE levels comparable to today, investors have historically not experienced positive real returns over the following decade.

Elevated valuation regimes can persist for surprisingly long periods, particularly when supported by abundant liquidity and compelling narratives. Persistence, however, should not be confused with permanence. The longer valuations remain detached from underlying earnings, the more dependent market outcomes become on the continued delivery of exceptionally favourable assumptions.

Artificial intelligence undoubtedly represents one of the most important technological developments of recent decades. Yet this, in itself, is not historically unusual. The internet was an equally transformative general-purpose technology, reshaping communication, commerce and productivity in ways few could fully anticipate. The defining characteristic of the dot-com bubble was never the absence of innovation, but the willingness of investors to capitalise that innovation at almost any price.

As Robert Shiller has argued, narratives are most powerful when they are rooted in reality. The internet was real. Artificial intelligence is real. It is precisely this reality that makes extrapolation so persuasive, and why markets repeatedly struggle to distinguish between technological potential and investment value.

However, today’s market does differ from earlier bubbles in one important respect. During the dot-com era, the vast majority of capital was still allocated through active investment decisions. Prices were primarily set by investors making explicit judgements about valuation, fundamentals and expected returns.

Today, a much larger share of global equity flows enters markets through passive vehicles that allocate capital according to market capitalisation rather than fundamental value. This has brought enormous benefits in terms of lower costs and broader diversification. But it also means that as companies become larger, they automatically receive a larger share of incremental investment, reinforcing existing market leadership irrespective of valuation.

Price discovery has not disappeared. But a growing proportion of market participants are no longer engaged in it. Instead of asking what a company is worth, they buy it because it already represents a larger weight in an index. The responsibility for efficient pricing therefore falls on an increasingly smaller pool of active investors.

In periods of broad market participation, this distinction matters relatively little. In periods characterised by narrow leadership and powerful narratives, however, passive flows can reinforce existing trends, making prices more reflective of capital flows than of changing assessments of intrinsic value.

This is precisely the type of environment in which active management can demonstrate its greatest value. Not because active managers possess superior forecasting abilities, but because they remain directly engaged in the process of price discovery. They retain the flexibility to question consensus, reduce exposure to increasingly concentrated areas of the market and allocate capital where the relationship between price and value remains more favourable.

Passive investing allocates capital according to what the market has already decided. Active investing continually asks whether the market’s judgement remains correct.

None of this implies that investors should abandon equities. History overwhelmingly shows that shares in productive businesses remain the most effective long-term vehicle for wealth creation. Even when investments are made at elevated valuations, patience, diversification and a sufficiently long investment horizon have typically rewarded disciplined investors.

The final chart provides perhaps the most important perspective of all. It shows the annualised returns earned by investors entering the US equity market at different points in time and holding their investment for progressively longer periods. The message is striking. Short-term outcomes vary widely and can prove deeply disappointing. Over longer holding periods, however, the range of outcomes narrows markedly. History suggests that while valuation influences the journey, time remains the long-term investor’s greatest ally. In our dataset, every holding period of more than ten years produced a positive annualised return, while 97% of all rolling investment periods over the past 35 years generated positive returns.

The challenge, therefore, is not whether to invest, but how, and at what price.

Innovation has always been the engine of long-term wealth creation. But innovation alone has never guaranteed superior investment returns. Innovation creates opportunity. Valuation determines outcomes. One without the other is an incomplete investment thesis. History suggests that markets periodically forget this distinction. It also suggests they eventually rediscover it.