AI Is the Ultimate Bubble—It Checks Every Box for a Historic Tech Crash

AI isn’t just another hype cycle, nor is it merely a large tech bubble. It could be the Platonic ideal of a tech bubble—the kind financial historians would engineer in a lab to study how extreme a market mania can get. The one bubble to burst them all. Here’s why.

Ever since ChatGPT went viral in late 2022, triggering nearly every company near (and far from) Silicon Valley to rebrand itself as AI-first, the fear of an inflating bubble has hung over global markets. Headlines were already warning of a bubble as early as May 2023, and by this fall, the idea became conventional wisdom. Financial analysts, independent research firms, tech skeptics, and even AI insiders all agree: we’re in some kind of AI bubble.

But as bubble talk has intensified, I’ve noticed almost no one has systematically broken down exactly how AI fits the definition of a bubble, what that classification actually means for the broader economy, and what fallout we can expect. It’s not enough to note that speculation is out of control (which it obviously is) or that AI investment today is 17 times what internet investment was right before the dot-com bust. Yes, market concentration is at unprecedented levels; yes, at points Nvidia’s market capitalization has nearly matched the size of Canada’s entire annual GDP. But in theory, the world could still decide AI is worth every dollar poured into it.

I wanted a proven, academic framework to analyze the AI mania, so I turned to the economists who literally wrote the book on tech bubbles.

In 2019, University of Maryland economists Brent Goldfarb and David A. Kirsch published Bubbles and Crashes: The Boom and Bust of Technological Innovation. Analyzing 58 historical examples of technological hype—from electric lighting to early aviation to the dot-com boom—they developed a framework to judge whether a new innovation will trigger a full-blown bubble. Interestingly, many game-changing technologies that went on to become foundational global industries (like lasers, freon, and FM radio) never created bubbles. Others, from commercial airplanes to transistors to early broadcast radio, absolutely did.

Unlike many traditional economists who view markets as the sum of rational, informed decisions (so much so that some even argue bubbles don’t exist at all), Goldfarb and Kirsch argue that the collective story people tell about an innovation—what it can do, how useful it will be, how much profit it will generate—creates the conditions for a bubble. “Our work puts the role of narrative at center stage,” they write. “We cannot understand real economic outcomes without also understanding when the stories that influence decisions emerge.”

Their framework evaluates tech bubbles across four core factors: uncertainty, pure-play investments, novice investors, and coordinated narratives around the innovation’s commercial potential. They rank historical bubbles on a scale of 0 to 8, with an 8 meaning a bubble is all but guaranteed.

I decided to apply this framework to generative AI, and reached out to Goldfarb directly to get his take on the current AI craze. For the record, the conclusions below are my own unless explicitly stated otherwise.

Uncertainty

Back in 1895, Austin, Texas installed 165-foot-tall “moonlight towers” outfitted with arc lighting across the city’s busiest public spaces. Onlookers gathered to stare up in awe as ash from the carbon filaments rained down on the crowd.

For some technologies, the value is obvious from day one. Goldfarb notes that electric lighting was so clearly useful that people could immediately imagine having the same tech in their own homes. Even so, Goldfarb and Kirsch write, “as marvelous as electric light was, the American economy would spend the following five decades figuring out how to fully exploit electricity.”

“Most major technological innovations come into the world like electric arc lighting—wondrous, challenging, sometimes dangerous, always raw and imperfect,” they write. “Inventors, entrepreneurs, investors, regulators, and customers struggle to figure out what the technology can do, how to organize its production and distribution and what people are willing to pay for it.”

In short, uncertainty is the foundation of any tech bubble. It’s uncertainty over whether the hype entrepreneurs sell will turn into actual sustainable revenue, which parts of existing supply chains it will displace, how many competitors will flood the space, and how long it will take to turn a profit. And if uncertainty is the base requirement for a bubble, AI is already blaring every alarm bell.

From the start, OpenAI’s Sam Altman has bet the entire company on building artificial general intelligence (AGI)—so much so that when a crowd of industry analysts once asked him about OpenAI’s business model, he openly replied that his plan was to build AGI, then just ask it how to make money. (He has since walked back that framing, saying AGI is no longer “a super useful term.”) Meta is chasing “superintelligence,” whatever that ends up meaning. The goalposts keep shifting.

Nearly three years after AI became Silicon Valley’s top priority, most major players (with the exception of Nvidia, whose chips will likely remain in demand even after a bust) still have not proven what their long-term AI business model will actually look like. OpenAI, Anthropic, and big tech companies betting big on AI are burning through billions of dollars a year. Inference costs have not come down, and these companies still lose money on nearly every user query. The long-term viability of their enterprise AI products is, at best, a huge open question.

Will the product that justifies hundreds of billions in investment be a Google replacement? A new social media platform? Workplace automation? How will AI companies account for sky-high energy and computing costs? If they lose ongoing copyright lawsuits over training data, will they have to pay to license content—and pass that extra cost onto consumers? A recent MIT study sent shockwaves through the market when it found that 95% of companies that adopted generative AI have not seen any profit gain from the technology.

“Usually over time, uncertainty goes down,” Goldfarb says. Market participants learn what works and what doesn’t. With AI, that hasn’t happened. “What has happened in the last few months,” he says, “is that we've realized there is a jagged frontier, and some of the earliest claims about the effectiveness of AI have been mixed or not as great as initially claimed.” Goldfarb thinks the market is still vastly underestimating how hard it is to integrate AI into existing businesses, and he’s far from alone in that view. “If we are underestimating this difficulty as a whole,” Goldfarb says, “then we will be more likely to have a bubble.”

The closest historical parallel for AI here isn’t electric lighting—it’s early radio. When RCA launched its first broadcasting service in 1919, everyone immediately recognized it was a transformative new information technology. But no one could agree on how it would make money. “Would radio be a loss-leading marketing for department stores? A public service for broadcasting Sunday sermons? An ad-supported medium for entertainment?” the authors write. “All were possible. All were subjects of technological narratives.” As a result, radio became one of the biggest bubbles in history, peaking in 1929 before losing 97% of its value in the crash. RCA wasn’t a niche player—it was one of the most heavily traded stocks on the market, alongside Ford. As The New Yorker recently put it, it was “the Nvidia of its day.”

Pure-Play Investments

Why is Toyota valued at $273 billion while Tesla hit a $1.5 trillion market cap, even though Toyota sold more cars and generated three times the revenue of Tesla last year? The answer boils down to Tesla’s status as a “pure play” investment in electric (and eventually autonomous) vehicles. In the 2010s, Elon Musk leveraged the widespread uncertainty and excitement around EVs to sell a narrative of a future free of gas-powered cars that was so compelling that investors chose to bet big on a volatile startup over a proven, profitable incumbent.

A pure-play company is one whose entire future is tied to the success of a single new innovation, making it the perfect vehicle for entrepreneurs to sell exciting, exaggerated hype—and you need pure plays to inflate a major bubble. They turn stories into massive market bets.

So far this year, Silicon Valley Bank data shows 58% of all venture capital investment has gone to AI companies. While there aren’t a huge number of pure-play AI options available to retail investors (another key bubble criteria), there are several major ones. Nvidia is at the top of the list: it has staked its entire future on building chips for AI firms, and in the process became the first company in history to hit a $4 trillion market cap.

Per Goldfarb and Kirsch’s framework, the more pure plays a sector has, the more likely it is to overheat and form a bubble. SoftBank is already planning to pour tens of billions into OpenAI, the purest AI play of all, even though OpenAI isn’t yet public. Analysts speculate that when OpenAI does go public, it could be the first trillion-dollar IPO. Investors have also backed other pure-play AI companies like Perplexity (now valued at $20 billion) and CoreWeave (which has a $61 billion market cap).

For AI, these interconnected pure plays are extra worrying: the biggest companies are increasingly financially tied to one another. Nvidia just announced a proposed $100 billion investment in OpenAI, which relies entirely on Nvidia’s chips to operate. OpenAI relies on Microsoft’s cloud computing power from a $10 billion partnership, and Microsoft in turn relies on OpenAI’s AI models to compete in the space.

“The big question is how much of that is in the private markets, and how much of that is in the public markets?” Goldfarb says. If most of the investment is private, only wealthy private investors will lose heavily in a crash. If most of it is in public markets like stocks and mutual funds, a crash will eat into everyday people’s pensions and 401(k)s. And right now, AI investment is increasingly moving into public markets. (Many analysts also point to the rise of private credit as a growing systemic risk, as more small investors have poured money into unregulated, opaque AI deals over the past year.) Either way, the sums involved are staggering. As of late summer 2025, Nvidia alone makes up roughly 8% of the total value of the entire U.S. stock market.

Novice Investors

Twenty-five years ago, on March 10, 2000, the tech-heavy Nasdaq hit an all-time high of 5,132 points, capping an 86% surge in just one year, fueled by a gold rush for internet stocks like eToys, CDNow, Amazon, and the infamous Pets.com.

Today, hordes of novice retail investors are pouring money into AI via trading apps like Robinhood and E-Trade. In 2024, Nvidia was the single most bought stock by retail traders, who poured nearly $30 billion into the chipmaker that year alone. Retail investors chasing AI are also piling into other big tech stocks with AI exposure like Microsoft, Meta, and Google.

While most AI investment so far comes from institutional investors, more risky pure-play AI startups like CoreWeave are going public (or preparing to go public) alongside Nvidia and the big tech giants. CoreWeave’s IPO in March was initially seen as underwhelming, but it has surged in value since, giving retail investors another vehicle to pour money into AI.

As Goldfarb points out, literally everyone is a novice investor when it comes to AI. It’s such a new field, there’s so much inherent uncertainty, no one can reliably predict how it will play out. Goldfarb and Kirsch note that what makes today different from a century ago is that anyone can get in on the hype. A hundred years ago, most working people couldn’t afford to buy individual stocks, which limited how big bubbles could get (though that didn’t stop the Great Depression). Today, you can buy any stock, from any size company, with a single tap on a phone app. Combine that with the increasing casino-ification of the U.S. economy, and the collapse of meaningful regulation to rein in excessive hype, and you have the perfect setup for novice investors to sink their life savings into the vague promise of superintelligence.

Coordinated Narrative of Inevitability

In 1927, Charles Lindbergh completed the first solo nonstop transatlantic flight from New York to Paris. The aviation industry had already been propped up by government subsidies for 25 years at that point, but Lindbergh’s flight dominated headlines around the world. It was the biggest tech demo of its era, and it became a massive coordinating event—on the same scale as ChatGPT’s 2022 viral launch—that signaled to investors it was time to pour money into aviation.

“Expert investors appreciated correctly the importance of airplanes and air travel,” Goldfarb and Kirsch write, but “the narrative of inevitability largely drowned out their caution. Technological uncertainty was framed as opportunity, not risk. The market overestimated how quickly the industry would achieve technological viability and profitability.”

The aviation bubble popped in 1929: from its peak that May, aviation stocks lost 96% of their value by May 1932.

When it comes to AI, this narrative of inevitability is the clearest check mark on the bubble framework. There is no bigger, more unifying story than the one AI leaders have been pushing for years: AGI will soon be able to do almost any job a human can do, and will usher in an era of transformative technology we can barely imagine. AI will automate millions of jobs, rewrite entire industries, cure cancer, solve climate change—it will do literally everything. Add to that the industry’s narrative that the U.S. must “beat China” to AGI, so AI can’t be regulated at all, and you have even more fuel for the hype fire.

“Is this a good story?” Goldfarb says. “The answer is profoundly yes.”

Even for early aviation, it was always clear that aviation’s core purpose was moving people and goods far faster than cars, trains, or horses. What pushes AI into a league of its own as a bubble is that its promise to investors is almost infinite. It’s not just uncertain—it’s unknowable. We also have to remember that AI’s boom came after a decade of near-zero interest rates that taught Silicon Valley investors to bet on companies with no viable business model, just a big compelling narrative. Uber, the poster child of that era, was founded in 2009 and didn’t turn a profit until 2023. The AI narrative is that same “Uber for X” hype on steroids. Every part of the AI story—whether it’s “AI will cure cancer” or “AI will automate all white-collar work”—appeals to some segment of investors, making it uniquely powerful at inflating a bubble, and uniquely dangerous to the broader economy.

It’s worth repeating that AI’s closest historical parallels are 1920s aviation and early broadcast radio. Both had extreme levels of uncertainty, both were hyped with incredibly powerful unifying narratives, both were dominated by pure-play companies looking to cash in on the new transformative tech, and both were accessible to the retail investors of the era. Both helped inflate a bubble so big that when it popped in 1929, it triggered the Great Depression.

So, Goldfarb confirms, AI has every single hallmark of a full-blown bubble. “There’s no question,” he says. “It hits all the right notes.” Uncertainty? Check. Pure plays? Check. Novice investors? Check. A compelling unifying narrative? Check. On that 0 to 8 scale Goldfarb and Kirsch use to rank bubbles, Goldfarb says AI is an 8. Buyer beware.

---

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