As a forewarning to the reader, I am a physicist, not an economist. What I wrote below may very well be naive. In fact, I hope to stand corrected by readers in the comments thread - that has been an effective way to self-improvement over my 20+ years of blogging, and I continue to value it today, even though our AI friends can be just as quick at explaining my fallacies. Anyway, this is my blog, so it is entirely appropriate to be naive.
So, the writing is on the wall - AI has come to dominate the scene of technological advancement in the 2020ies, and with it has come an enormous interest of the financial markets around the globe. Undoubtedly, the rise of value of stocks connected with AI is polarizing the scene, much as it happened at the end of the XXth century with the rise of the "dot-com"s. Back then, the internet had recently become commercially accessible, and every new month seemed to bring in a new revolutionary company. Investors rapidly concluded that the internet would change everything, and therefore every internet company was worth an enormous amount. The preamble of the above sentence was absolutely true, but the conclusion was short-sighted.
The .com companies had no profits, no customers, and often barely a product to sell, but the general perception boosted their value and they managed to raise hundreds of millions of dollars. The old economical thinking, wherein stock prices were valued based on earnings, cash flow, and profitability, went dismissed in favor of page views and number of users. Of course this situation of continuous overevaluation could not sustain itself; companies could not make money, capital dried up, and many of the startups disappeared, producing an 80% drop in the NASDAQ index.
But the internet did not fail. And that bubble ended up having financed an enormous amount of infrastructure, including fiber optic highways, data centres, software, and expertise. The technology was real and it was indeed transformative; the investments were simply too early and too enthusiastic, and grew at a pace that could not be sustained. Lesson learned?
Today, AI again promises to change everything, and this concept may again be true. Yet, again we see the pattern - hundreds of billions of dollars are being spent on GPUs, data centres, power infrastructure and networking. For investors, the fear to miss out on new gains is again very strong, fueling overinvestments. But the situation is different from that of the 1990ies. While dot-coms had continual fundraising needs, the AI leaders of today include extremely rich companies (Micro$oft, Alphabet, Amazon, Meta). These generate tens of billions of dollars annually, so they have wider and stronger shoulders.
It seems to me as if the role of speculative investors is much smaller today than it was in the dot-com era. Furthermore, AI already generates revenue, selling cloud services, medical applications, scientific and enterprise software. We may be looking at a phase which is less like one of speculative websites and more like one when railroads are built. Much of that infrastructure will remain useful in the future. This is common with major new technologies: a phase of speculative overinvestment is the prelude to a sustainable growth.
One additional reason why the situation today is markedly different is that governments are involved, as AI has become part of national security and military competition among countries. Governments are encouraging AI investment, and this creates demand beyond the purely commercial market.
So will there be a crash? Is the AI bubble going to burst any time soon?
From what I can see, it is likely that many speculative AI startups will fail in the short term; chip makers will become overvalued, and weak companies will disappear. All the while, AI will continue to spread through the economy, and strong firms will not only survive but weather the storm to their advantage, getting less and less trouble from competition.
In other words, it is quite predictable that the technological revolution will continue. Who wants to live in a world without LLMs today? Or without the incredibly more powerful tools that AI developers are designing today, and that AI itself will be designing tomorrow? Of course, you could be among the minority of neo-luddites who hates these new developments, but realistically there is an evolutionary trend that cannot be stopped. It can be for better or for worse - that is to be seen - but it is futile to resist it.
As far as I am concerned, I consider myself incredibly lucky to be experiencing this, and to be living through these years of AI revolution. I was fascinated by the prospects of artificial intelligence as a teenager, when I read books such as "Godel, Escher, Bach - an Eternal Golden Braid" by Douglas Hofstadter and fantasized about a world where information and intelligence could be freely available in no time and with no effort, on everything. Those days are here now, and more is yet to come. True - while all of that is becoming real, we live in an increasingly disturbing society where ethical values have hit record lows, inequalities are abominable, and we insist in not solving world hunger and exploitation of human over human. That, too, might be among the growing pain of a better future. Or not - we shall see.
