Why You Might Not See An AI Stock Market Bubble Once It's Formed

In purely economic terms, disruptive technologies are above all catalysts for growth, enhancing the productivity of companies, industries, economies, societies, or even humanity as a whole. People are quick to grasp the implications of a new technology, but tend to be overly optimistic about the time frame in which a new technology will have a real impact on economic productivity. Historical bubbles have formed and burst primarily because of misjudgement about the time horizon over which a disruptive technology will have a truly relevant impact. Were investors in the late 1920s wrong to assume that the mass production of automobiles would significantly change our civilization? Did investors in the late 1990s misjudge that the commercialization of the Internet would be the beginning of one of the greatest technological transformations of all time? Years after both bubbles had long since burst, the impact of the initially hyped technology was undeniable: in the 1950s, large segments of American society could afford cars, with enormous social and economic consequences; and today, the Internet is such an integral part of everyday life that network operators (initially treated as disruptors) are traded on the stock market at low price-earnings ratios like providers of basic utilities such as running water or electricity.

The emergence of generative pre-trained transformer-based neural networks (GPTs) and ChatGPT in particular in late 2022 has created a euphoria that is clearly reflected in the stock market valuations of AI-related companies in 2023, raising concerns about an impending AI bubble reminiscent of the dotcom bubble. But is it just hype to expect generative AI to transform almost every part of our lives, economy, and society, and boost our productivity like no other technology in decades? In my view, AI will lead to a transformation of today’s economy on the same scale as the steam engine transformed the artisan into the industrial economy in the 19th century. We don’t even have a word for an economy in which AI is a key factor of production. And yet I am certain that an AI bubble will form and burst. In my opinion, the only question is when, not if.

To understand this, we need to dissect the anatomy of a speculative bubble. “Bubble” usually refers to the hype-driven decoupling of the stock market from its fundamentals (usually earnings). But it’s not that simple. As George Soros explained in his 1987 book “The Alchemy of Finance”, stock prices are not just a passive reflection of fundamentals: The current AI buzz is likely to strengthen the fundamentals of the companies involved.
So anyone who expects AI stock prices to become bluntly disconnected from fundamentals in the coming months or years, until valuations become so stretched that the bubble has to burst, is likely to be wrong, in my view. The current hype around the wonders of AI will have an impact on the fundamentals of the companies involved. In almost every industry, AI is being touted as the solution to many, if not all, problems. Significant investments are being made that are driving the revenues and ultimately earnings of the companies that provide AI services, operate servers, build servers, and produce computer chips. This growth will continue to drive the valuations of these AI-related companies (i.e. investors will be willing to buy the stocks at higher price-earnings ratios). Obviously higher stock market valuations allow companies to raise more money for fewer shares in seasoned equity offerings (SEOs). Much more subtle, however, a positive feedback loop is emerging between higher stock market valuations and the fundamentals of these companies; George Soros called this phenomenon reflexivity. Higher stock market valuations allow companies to offer attractive stock-based compensation to employees, reducing labor costs and boosting profits, which in turn translates into lower price-earnings ratios that justifies higher stock market valuations. The companies at the forefront of a technological revolution will also become more attractive as employers and will be able to recruit the brightest minds, possibly at lower salaries. Although it should play a minor role, loans may also become cheaper for the fast-growing participants in the AI market. Lower interest rates ultimately lead to higher profits. Because all of these effects affect fundamentals, a bubble may not be visible to investors at some point.

At the same time the AI sector will expand horizontally and proliferate into all other industries, making it difficult to isolate a single sector bubble. New technology will create new problems, which in turn will be solved by new technology: Server operators, for example, are already groaning at the enormous power costs of running the new language models. One solution may be to use analog chips that can perform the matrix computations that underlie AI in a much more power-efficient way. As you can see, the focus may shift to sub-developments within the larger trend. To stay with this example: maybe two years from now, investors will be talking about an analog computing bubble, and the possibility of a bubble in the entire AI-related industry will not even be considered because this technology has become so ubiquitous.

So while I am certain that a bubble will form and burst, I have no idea what the timeline will be. I think we are witnessing just the dawn of this development. Today, AI can write impressive texts, code concisely, and generate beautiful images. In reality, however, we are still far from the day when AI can truly replace writers, programmers or designers. The world in 20 years will be completely different, but the world in 2 years won’t be. However, it is in the nature of the stock market to euphorically project new developments. AI-related companies will rush from one stunning earnings report and outlook to the next over the next few quarters or even years. At some point, most voices warning of an AI bubble will fall silent and the most dangerous phrase will be heard: “This time it’s different.” Until some event reveals the decoupling of market valuations from real industrial productivity gains and the bubble bursts. We don’t even know what caused the dot-com bubble to burst in 2000, maybe it was overpriced 3G rights in Europe.

This is the author's opinion, not financial advice.