The AI Bubble, Explained
When analysts warn the AI bubble is about to burst, what exactly would pop? The answer has everything to do with stock prices and very little to do with whether AI works.
You’ve seen the headlines. “This Is How the AI Bubble Bursts.” “Something Ominous is Happening in the AI Economy.” If you use AI tools regularly, you might be wondering whether this thing you’ve come to rely on is about to disappear.
What those headlines aren’t telling you: when analysts at Yale, Goldman Sachs, or the OECD warn about an “AI bubble,” they’re talking about money. Stock prices, company valuations, investment returns. They’re saying that AI stocks may be priced higher than future profits can justify.
They’re not saying AI technology doesn’t work, that AI capabilities are fake, or that ChatGPT will stop answering questions.
This distinction matters because the headlines blur two separate things. One is whether AI will keep working. The other is whether AI stocks are overpriced. Both can be true at once. A house can be a perfectly good house while being overpriced in the real estate market. You can cook dinner in its kitchen regardless of whether the listing price makes sense. AI works the same way. The technology can keep functioning just fine while the companies building it are overvalued by investors.
Once you see this distinction, the AI bubble conversation gets much easier to follow.
So What Is a Bubble, Anyway?
A bubble refers to an an economic bubble, which happens when prices climb far beyond what underlying value can justify. People buy not because of what something produces, but because they expect someone else will pay more for it later.
Bubbles are difficult to confirm while you’re inside one. Rapid price increases, widespread enthusiasm, and abandonment of traditional valuation methods describe both dangerous bubbles and the early stages of genuinely transformative investments. The internet did transform daily life, even though internet stocks were overvalued in the late 1990s.
This explains why informed people currently disagree about AI. Ray Dalio of Bridgewater Associates said in early 2025 that current AI investment levels are “very similar” to the dot-com bubble. Sam Altman, CEO of OpenAI, stated in 2025 that he believes an AI bubble is ongoing. Pat Gelsinger, who led Intel until late 2024, said the market is “of course” in a bubble. Others argue AI is different because the major players are established, profitable companies rather than speculative startups.
The Numbers Behind the Nervousness
The scale of AI investment is remarkable. Goldman Sachs projects companies will spend $527 billion on AI infrastructure in 2026. For comparison, the Apollo program for NASA to land humans on the moon cost about $300 billion in inflation-adjusted dollars over roughly a decade.
AI stocks have driven roughly 75 to 80 percent of S&P 500 returns since ChatGPT launched in November 2022. In late 2025, the five largest companies held 30 percent of the S&P 500 and 20 percent of the MSCI World index, which analysts noted was the greatest market concentration in half a century.
What concerns analysts is the gap between spending and returns. An August 2025 MIT report found that despite $30 to $40 billion in enterprise AI investment, 95 percent of organizations were seeing zero return. This means companies must demonstrate real payoff, and relatively soon, or investor confidence could erode.
The financial ties between major AI players have also grown unusually intertwined. Nvidia invests in OpenAI, and OpenAI buys chips from Nvidia. Microsoft holds a stake in OpenAI while being a major customer of CoreWeave, in which Nvidia also holds equity. This circulation of capital among a small group of companies can inflate headline figures while making actual outside demand harder to assess. Michael Burry, the investor known for betting against housing before 2008, has been betting against Nvidia, writing that “true end demand is ridiculously small” and that “almost all customers are funded by their dealers.”
When Bubbles Burst, What Happens to the Technology?
The dot-com crash offers a clear lesson. The Nasdaq fell 78 percent between March 2000 and October 2002. More than half of public dot-com companies failed by 2004. Trillions of dollars in market value vanished.
But websites kept working, email kept running, and Amazon kept selling books. What collapsed were stock prices and companies that had built unsustainable businesses, spending investor money without becoming profitable.
The fiber optic cables laid during the bubble stayed in the ground. The resulting oversupply of bandwidth helped make internet services cheaper to deliver, enabling a later generation of web companies to emerge. Infrastructure built during the frenzy became part of the foundation for the mature internet economy.
The British railway bubble of the 1840s shows a similar pattern. At its peak, railway investment reached 6.7 percent of British national income, and the eventual crash ruined many investors. But the railways themselves remained. Track authorized during the bubble years came to represent about 90 percent of Britain’s eventual rail system.
Jamie Dimon, CEO of J.P. Morgan, put it this way when discussing AI: the technology “will pay off, just like cars in total paid off, and TVs in total paid off, but most people involved in them didn’t do well.”
What This Means for How You Think About AI
If a bubble bursts, some AI products will disappear when their parent companies fail. Others will raise prices or cut free tiers as investor subsidies dry up. The tools you use today may not be the tools available in two years.
But the skills transfer. Knowing how to write an effective prompt, how to evaluate whether an output is trustworthy, when AI is the right tool and when it isn’t: these abilities work across platforms. If one chatbot vanishes, another will exist. The interface might change, but the underlying competency remains yours. This is why learning to use AI well matters more than learning to use any particular AI product.
The bubble conversation also reveals something important about how AI gets covered in the media. Much of what gets written about artificial intelligence is shaped by market enthusiasm or market fear. A company raising billions in funding is not proof that its technology works well. A stock price falling is not proof that the technology stopped working. Breathless headlines about “AI collapse” are about investor returns, not about whether your AI tools will keep functioning.
Once you can separate the money story from the technology story, the headlines get a lot less confusing.
What Comes Next
Current conditions share characteristics with previous bubbles: high market concentration, stretched valuations, widespread speculation. Whether this means a correction is imminent, years away, or avoidable is impossible to know. Bubbles can last longer than skeptics expect. The dot-com bubble continued growing for years after Alan Greenspan warned of “irrational exuberance” in December 1996.
The sensible approach is to understand what’s happening and make choices that would hold up across different outcomes. If you have financial exposure to AI stocks, the moment deserves attention. If you mainly use AI as a tool, a bubble bursting would bring real but manageable changes: shifting prices, disappearing products, industry consolidation, but continued access to what AI can do.
The technology and the market are two different things. The headlines blur them together, but you don’t have to.
Hello, I just found your site and love the content. I am former librarian too, but now I work for a corporation as an reference and info researcher. my company is pushing AI enormously, even in areas when it does not make any sense. my team is trying to point out the dangers and pitfalls (check your original sources, for god's sake), but it often feels like fighting against windmills.
thanks for posting this, very valuable!
Really important to keep this distinction between market and technology in mind when thinking and talking about this