“In the expectation that artificial intelligence will be transformative, large American tech firms poured over $400bn into data centres and other necessary infrastructure during 2025; by one estimate, a staggering $7trn will be spent by the end of the decade,” wrote The Economist in its November 10 issue on “The World Ahead 2026.” “Yet revenues from AI so far amount to a measly $50bn a year, about an eighth of Apple or Alphabet’s entire annual revenues.”
Not surprisingly, concerns keep rising over a potential AI bubble. “Will the bubble burst?,” The Economist asked. “As with railways, electricity and the internet, a crash would not mean that the technology does not have real value. But it could have wide economic impact.”
But, in a recent NY Times opinion article, “AI Is a Bubble. Maybe That’s OK,” economist Mohamed El-Erian noted that “Bubbles seem by definition irrational. They grow as investors — often hostage to exuberant, herd-like behavior — push valuations well beyond anything warranted by the fundamentals on the ground.”
“A bursting bubble could indeed be painful in the short term. But what if we’re in a rational bubble that, unlike other big speculative manias in history, takes our economy to a fundamentally better place?,” added El-Erian. “I’m borrowing the phrase rational bubble from conversations with a Nobel laureate in economics, my friend A. Michael Spence,” — a co-recipient of the 2001 Nobel Prize in Economic Sciences.
Michael Spence recently participated in a mid-November conference in China where he said that “The AI investment boom is a rational bubble” according to an article reporting on his remarks.
“Artificial intelligence is revolutionary in promoting the progress of various scientific fields, and its impact on the economy will come more slowly,” explained Spence. “Because you need to figure out how to use it, conduct experiments, change existing behaviors, learn new skills, change business models and corporate structures, all these things we are familiar with. But I think it is reasonable to speculate that if used properly, AI will have a very huge impact.”
“The artificial intelligence we talk about today may be completely different from what we will have tomorrow,” Spence added. “If we don’t know the development direction of AI agents and how reliable they will be in the future, it is difficult to accurately predict what impact they will bring.”
In his NY Times article, El-Erian further explained that the the reason the AI bubble will pop is not because investors are overestimating AI, but because of three key forces:
- The arms race among the largest tech companies, all of which are working on groundbreaking AI models. “Not all will flourish, especially as the required funds — for elite engineers in high demand, sprawling data centers and energy consumption — continue to soar.”
- AI washing, — the gold rush that attracts prospectors who plaster an AI label onto mundane services reminiscent of the dot-com bubble of the late 1990s when so many startups added .com their names.
- External developments will derail some companies, including regulatory changes, a lack of widespread AI adoption, geopolitical competition, and the presence of bad actors.
“Whichever way you look at it, the potential payoffs of A.I. adoption are staggering — for the economy, for social sectors and, of course, for investors. … From their perspective, what some may ultimately see as enormous overspending is, in fact, a calculated portfolio play that fuels competition and innovation. … The belief in A.I.’s transformative power is justified. The resulting flood of capital is a logical response. Some will lose. Overall, we will be much better off.”
A number of other economists have expressed similar views. For example, in a recent article, “Is AI a Bubble? Not So Fast?,” George Mason University professor Tyler Cowen noted that “it is premature to write off current AI valuations as a bubble. How do I know this? First, because a lot of so-called bubbles pay off in the long run. You could argue that the U.S. real-estate sector was a bubble in 2007. Indeed, home prices crashed soon after. Yet in most parts of the country, prices later came roaring back.”
“The same was true of Amazon shares during the dot-com bubble of the late 1990s,” added Cowen. “After the crash of internet stocks, it took Amazon shares years to return to their previous heights. But they did, and then far exceeded them. The lesson is clear: If you see an investment that looks steady, sometimes the best thing to do is to jump on it. No one can know when the market bottom is going to come, and either way, it’ll probably spike again fairly soon.”
For a reminder of what the internet dot-com bubble was like, let me share this article, “The Bubble – Reconsidering the Boom and the Bust,” written in 2004 by my friend and former IBM colleague John Patrick.
Last week I posted “Tracking the Evolution of AI,” a blog based on “The Longitudinal Expert AI Panel” (LEAP), a report which raised some very intriguing questions in my mind about the long term evolution of AI. What if there are two kinds of historically transformative technologies: those whose impact plays out over a relatively small number of decades, — say four to six, — and those whose impact play out over a much longer timeframe?
LEAP is a new project that aims to create the mosts reliable views of the long term evolution of AI by tracking the predictions of leading computer scientists, economists, industry professionals, and policy researchers, as well as those of highly accurate superforecasters and engaged members of the general public.
One of LEAP’s prediction particularly caught my attention. According the LEAP, the median experts predict that the impact of AI by 2040 will be comparable to the impacts of what it calls a ‘technology of the century’, — e.g., steam power, railroads, electricity or automobiles. According to this prediction, the impact of AI will likely be comparable to that of the technology revolutions we’ve had every four to six decades over the past two centuries.
But LEAP also uncovered a very different prediction. “Experts also give a 32% chance that AI will be at least as impactful as a ‘technology of the millennium,’ such as the printing press or the Industrial Revolution.” According to LEAP, a ‘technology of the millennium,’ means that its impact played out over a century or more rather than over a few decades.
For example, the printing press, invented by Johannes Gutenberg around 1440, dramatically accelerated the spread of knowledge and literacy across Renaissance Europe. Gutenberg’s printing revolution influenced almost every facet of life in the centuries that followed, as printed books significantly expanded the knowledge available to society.
Similarly, the Industrial Revolution transformed the economy by introducing machines, manufacturing processes, and related technological advances into activities that had previously relied on manual production. Starting in Great Britain, the Industrial Revolution spread to continental Europe and North America throughout the 19th century, and later to much of the world in the 20th century.
As Wikipedia explains: “The Industrial Revolution influenced almost every aspect of life. In particular, average income and population began to exhibit unprecedented sustained growth. Economists note the most important effect was that the standard of living for most in the Western world began to increase consistently for the first time, though others have said it did not begin to improve meaningfully until the 20th century. … Economic historians agree that the onset of the Industrial Revolution is the most important event in human history, comparable only to the adoption of agriculture with respect to material advancement.”
What if AI turns out to be a technology of the millennium akin to the Industrial Revolution? What if our newfound ability to analyze huge amounts of data with sophisticated algorithms and powerful computers is now leading to a new kind of cognitive revolution that will transform the economy and influence almost every aspect of life by introducing cognitive technological capabilities that were previously viewed as the exclusive domain of humans?
Looking at AI as a general purpose ‘technology of the millennium’ makes it easier to understand why Mohamed El-Erian, Michael Spence, Tyler Cowen and other economists are now arguing that AI is a “rational bubble.”
“It’s not just that many existing activities will be done better and more efficiently,” wrote El-Erian. “A.I. is poised to open the door to discoveries, particularly in health and education. Such gains would allow the economy to grow faster without kicking off inflation, something economists describe as raising the ‘speed limit’ for noninflationary growth. Increased productivity and a larger economy provide us with more opportunities to address the problems that my generation is leaving our kids and grandkids: high levels of debt, climate change and excessive income inequality.”