Is AI the Ultimate Tech Bubble?

AI might not merely be “a bubble,” or even an extraordinarily large one. It could potentially be the quintessential bubble – the kind that one might conceive in a lab if the objective were to engineer the Platonic ideal of a tech bubble, a bubble that could burst all others. This article aims to elaborate on this proposition.

The AI Bubble Perception Post – ChatGPT

Since ChatGPT achieved viral success in late 2022, compelling nearly every company in close proximity to Silicon Valley (and numerous others beyond) to pivot towards AI, the perception of a burgeoning bubble has been palpable. Headlines alluding to this phenomenon emerged as early as May 2023, and by this fall, it had become a rather widespread view. Financial analysts, independent research firms, tech skeptics, and even AI executives concur: we are indeed witnessing some form of an AI bubble.

However, as the discourse around the AI bubble intensified, it became evident that few were delving into a precise analysis of how AI constitutes a bubble, what this truly implies, and what the potential implications might be. Merely stating that speculation is rife, which is already quite apparent, or highlighting that there is now 17 times more investment in AI compared to internet companies prior to the dot – com bust is insufficient. While it is true that we are experiencing unprecedented levels of market concentration, and Nvidia has, at times, been valued almost on par with the entire economy of Canada on paper, it remains theoretically possible that the world deems AI worthy of such substantial investment.

Seeking a Reliable Framework

To gain a reliable and well – tested means of evaluating and comprehending the AI mania, it was necessary to turn to the scholars who have written extensively on tech bubbles. In 2019, economists Brent Goldfarb and David A. Kirsch of the University of Maryland published Bubbles and Crashes: The Boom and Bust of Technological Innovation. Through an examination of approximately 58 historical instances, spanning from electric lighting to aviation and the dot – com boom, Goldfarb and Kirsch developed a framework for determining whether a specific innovation led to a bubble. Some technologies, such as lasers, freon, and FM radio, which went on to become major commercial successes, did not create bubbles. In contrast, others like airplanes, transistors, and broadcast radio clearly did.

In contrast to many economists who view markets as the outcome of rational decisions made by purely rational actors – to the extent that some even posit that bubbles do not exist – Goldfarb and Kirsch argue that the narrative surrounding an innovation, including what it can achieve, its utility, and its potential for profitability, creates the conditions for a market bubble. “Our work places the role of narrative at the center,” they write. “We cannot fully understand real economic outcomes without also understanding the emergence of the stories that influence decision – making.”

Goldfarb and Kirsch’s Framework

Goldfarb and Kirsch’s framework for assessing tech bubbles takes into account four primary factors: the presence of uncertainty, pure plays, novice investors, and narratives surrounding commercial innovations. The authors identify and evaluate these factors, ranking their historical examples on a scale of 0 to 8, with 8 indicating the highest likelihood of predicting a bubble.

As the author began to apply this framework to generative AI, Goldfarb was contacted to obtain his perspective on the “bubbledom” of Silicon Valley’s latest craze. It should be noted that, unless otherwise stated, the conclusions presented here are those of the author, not Goldfarb’s.

Uncertainty in AI

In 1895, the city of Austin, Texas, acquired 165 – foot – tall “moonlight towers” equipped with arc lighting that burned carbon filaments and installed them in public areas. Spectators gathered to marvel as ash rained down. Goldfarb notes that with some technologies, the value is immediately evident. Electric lighting, for instance, “was clearly useful, and one could easily envision having it in their home.” However, as he and Kirsch write in the book, “despite the wonder of electric light, the American economy spent the subsequent five decades figuring out how to fully exploit electricity.”

“Most major technological innovations enter the world like electric arc lighting – remarkable, challenging, sometimes dangerous, and always raw and imperfect,” Goldfarb and Kirsch write in Bubbles. “Inventors, entrepreneurs, investors, regulators, and customers grapple with determining what the technology can do, how to organize its production and distribution, and what people are willing to pay for it.”

Uncertainty, thus, serves as the cornerstone of a tech bubble. Uncertainty regarding how the stories entrepreneurs tell about an innovation will translate into actual business, which parts of the value chain it might replace, how many competitors will enter the field, and how long it will take to reach fruition. In the case of AI, alarm bells are already sounding. From the outset, OpenAI’s Sam Altman has staked a great deal on building AGI (artificial general intelligence). He once addressed a group of industry observers who inquired about OpenAI’s business model and straightforwardly stated that his plan was to build a general intelligence system and then simply ask it how to make money (he has since distanced himself from this approach, stating that AGI is not “a super useful term”). Meta is aiming for “superintelligence,” though the exact meaning remains unclear. The goals keep shifting.

Nearly three years since AI took center stage in Silicon Valley, with the exception of Nvidia (whose chips would likely still be in use even if there were a bust), the major players have yet to demonstrate their long – term AI business models. OpenAI, Anthropic, and the AI – adopting tech giants are burning through billions. Inference costs remain high (these companies still incur losses on nearly every user query), and the long – term viability of their enterprise programs is at best a significant question mark. Will the product that justifies hundreds of billions in investment be a search engine replacement, a social media substitute, or workplace automation? How will AI companies factor in the still – sky – high costs of energy and computing? If copyright lawsuits do not go their way, will they have to license their training data and pass on the additional cost to consumers? A recent MIT study, which found that 95 percent of firms that adopted generative AI did not profit from the technology, has contributed to the latest wave of bubble fears.

“Typically, over time, uncertainty decreases,” Goldfarb says. People learn what works and what doesn’t. But this has not been the case with AI. “In the last few months,” he says, “we’ve realized that there is a jagged frontier, and some of the initial claims about AI’s effectiveness have been inconsistent or not as impressive as initially thought.” Goldfarb believes the market is still underestimating the difficulty of integrating AI into organizations, and he is not alone. “If we as a whole underestimate this difficulty,” he says, “then a bubble is more likely to form.”

AI may be more analogous to radio in history rather than electric lighting. When RCA started broadcasting in 1919, it was clear that it had a powerful information technology. However, how this would translate into business was less certain. “Would radio be a loss – leading marketing tool for department stores, a public service for broadcasting Sunday sermons, or an ad – supported entertainment medium?” the authors write. “All were possibilities, all part of technological narratives.” As a result, radio became one of the most significant bubbles in history, peaking in 1929 before losing 97 percent of its value in the crash. RCA was a major player, on par with Ford Motor Company in terms of high – traded stocks, and was, as The New Yorker recently put it, “the Nvidia of its day.”

Pure Play in the AI Landscape

Why is Toyota valued at $273 billion while Tesla is worth $1.5 trillion to investors, despite Toyota shipping more cars and generating three times as much revenue last year? The answer lies in Tesla’s status as a “pure – play” investment in electric (and to a lesser extent, autonomous) cars. In the 2010s, Elon Musk capitalized on the excitement and uncertainty surrounding EVs to paint a captivating picture of a future without internal combustion engines, enticing investors to bet heavily on a volatile startup over established companies. A pure – play company’s fate is tied to a particular innovation’s success, and entrepreneurs can tell more thrilling and fantastical stories about it, which is essential for a bubble to form. They are the vehicle through which narratives transform into tangible investments.

According to Silicon Valley Bank, 58 percent of all VC investment this year has gone to AI companies. While there aren’t an abundance of obvious pure – play investments accessible to retail investors (a factor in inflating a bubble), there are some significant ones. Nvidia tops the list, having staked its future on building chips for AI firms and becoming the first $4 trillion company in the process. When a sector has a large number of pure – play investments, as per Goldfarb and Kirsch’s framework, it is more likely to overheat and form a bubble. SoftBank plans to invest tens of billions of dollars into OpenAI, the purest AI play, though it is not yet open to retail investment (analysts speculate that if and when it is, OpenAI may become the first trillion – dollar IPO). Investors have also backed pure – play companies such as Perplexity (now valued at $20 billion) and CoreWeave ($61 billion market cap). In the case of AI, these pure – play investments are particularly concerning as the largest companies are becoming increasingly interconnected. Nvidia recently announced a $100 billion proposed investment in OpenAI, which in turn depends on Nvidia’s chips. OpenAI relies on Microsoft’s computing power through a $10 billion partnership, and Microsoft depends on OpenAI’s AI models.

“The key question is the proportion of investment in private versus public markets,” Goldfarb says. If most of the money is in private markets, private investors would bear the brunt of losses in a crash. If it’s mostly in public markets like stocks and mutual funds, then the crash would impact regular people’s pensions and 401(k)s. And indeed, AI investment is increasingly seeping into public markets (many market watchers have also pointed to the rise of private credit as a growing source of systemic risk, as more small investors have been able to invest in opaque deals over the past year). Either way, the sums are substantial. As of late summer 2025, Nvidia accounts for approximately 8 percent of the entire stock market’s value.

Novice Investors in the AI Frenzy

Twenty – five years ago, on March 10, 2000, the stock market reached a milestone: the tech – heavy Nasdaq hit a then – high of 5,132 units. At the time, it seemed to be continuing its rapid ascent, having risen a staggering 86 percent in the previous year alone, fueled by an investor rush for internet companies like eToys, CDNow, Amazon, and Pets.com.

Today, hordes of novice retail investors are pouring money into AI via E – Trade and the Robinhood app. In 2024, Nvidia was the most – bought equity by retail traders, who invested nearly $30 billion in the chipmaker that year. AI – interested retail investors are also flocking to other major tech stocks like Microsoft, Meta, and Google.

While most of the investment to date has been driven by institutional investors, more pure – play and riskier AI startups like CoreWeave are going public or preparing to do so, alongside Nvidia and the tech giants. CoreWeave’s March IPO was initially lackluster but has been on the rise since, providing another avenue for retail investors to inject money into AI.

As Goldfarb points out, everyone is somewhat of a novice investor when it comes to AI due to its novelty, the high level of uncertainty, and the unknowns regarding its future. What differentiates today from 100 years ago, as Goldfarb and Kirsch note in the book, is that anyone can now participate. A century ago, stocks were too expensive for most working people to purchase, which significantly limited the potential to inflate bubbles (although this did not prevent the Great Depression). Now, with a simple tap on the Robinhood app, stocks of all sizes and types are accessible. Coupled with the “casino – ification” of the economy and the breakdown of a robust regulatory apparatus, novice investors now have a means to invest their savings in the nebulous promise of superintelligence.

Coordination through AI Narratives

In 1927, Charles Lindbergh completed the first solo non – stop transatlantic flight from New York to Paris. By then, the aviation industry had been receiving government subsidies for a quarter of a century, but this flight made global news. It was the most significant tech demonstration of the time and became a major coordinating event, similar to the launch of ChatGPT, signaling to investors to invest heavily in the industry.

“Expert investors correctly recognized the importance of airplanes and air travel,” Goldfarb and Kirsch write, “but the narrative of inevitability largely overshadowed their caution. Technological uncertainty was framed as an opportunity rather than a risk. The market overestimated the speed at which the industry would achieve technological viability and profitability.”

Consequently, the bubble burst in 1929. From its peak in May, aviation stocks dropped 96 percent by May 1932.

Regarding AI, the narrative of inevitability is perhaps the most obvious and affirmative indication of a bubble. The most prominent narrative pushed by AI industry leaders even before the boom is that AGI will soon be capable of doing almost anything a human can, ushering in an era of super – powerful technology beyond our imagination. Jobs will be automated, industries transformed, cancer cured, and climate change solved – AI will seemingly do it all. Adding to this is the industry narrative that we must “beat” China to AGI and thus refrain from regulating AI at all costs, further fueling the frenzy.

“Is this a compelling story?” Goldfarb asks. “The answer is resoundingly yes.”

It was clear early on what aviation would be good at – moving people more quickly than cars, trains, or horses. What elevates AI’s “bubbledom” to another level is that the promise of AI to investors seems nearly infinite. It is not just uncertain; it is unknowable. It should be noted that AI emerged after nearly a decade of near – zero interest rate policy, which led Silicon Valley investors to bet on companies with weak business models but strong narratives. Uber, the poster – child startup of that era founded in 2009, did not post a profitable year until 2023. And the AI narrative is like “Uber for X” on steroids. Different aspects of the AI story, such as “AI will cure cancer” or “AI will automate all jobs,” appeal to a wide range of investors and partners, making it uniquely potent in its ability to inflate a bubble and thus dangerous to the economy.

It is worth repeating that two of the closest historical analogs to AI in terms of tech bubbles are aviation and broadcast radio. Both were characterized by high levels of uncertainty, powerful coordinating narratives, were seized upon by pure – play companies, and were accessible to retail investors of the time. Both contributed to the inflation of a bubble so large that its burst in 1929 led to the Great Depression.

So, yes, Goldfarb says, AI exhibits all the hallmarks of a bubble. “There’s no doubt,” he says. “It meets all the criteria.” Uncertainty? Check. Pure plays? Check. Novice investors? Check. A compelling narrative? Check. On the 0 – to – 8 scale, Goldfarb says, it’s an 8. Buyer beware.

Update 10/27/25 3:45pm ET: Due to an editing error, an earlier version of this story was initially published.