Find this research useful? Add aibubblequestion.com to your Google Preferred Sources (opens in a new window), one click, affects only your own results.

The AI Bubble Question
All pages

A public research project · Data snapshot: July 2026

Is the AI boom a bubble?

Trillions of dollars are being poured into AI chips and data centers. Either this is the most valuable infrastructure build-out in history, or it is the latest entry in a 200-year catalog of manias. This site puts the question against the historical record and lets you judge.

MOST PAST PEAKS: 1.5-4.5 YRS IN peak 0 0 2 4 6 8 10 12 14 years into the mania British Railway Mania: peaked about 1.5 years into the mania, fell 50-70% Railways 1846 US stocks 1929: peaked about 2.5 years in, fell 89% 1929 Japan 1989: peaked about 4 years in, fell about 80% over 13 years Japan 1989 Dot-com: Nasdaq peaked about 2 years into the mania, fell 78% Dot-com 2000 Telecom and fiber: peaked about 2.5 years in, equipment stocks fell 90-99% Telecom 2001 US housing: peaked about 4.5 years in, fell about a third nationally Housing 2006 or this? this? The AI boom, by the site's own editorial framing rather than a measured start date, roughly 3.5 years into its mania as of August 2026. Its peak, if there is one, is unknowable in advance. AI BOOM · AUG 2026
AI boom to date six documented bubbles Durations and declines drawn to scale from the case data; curve shapes and the AI line's height are stylized. The AI boom's "3.5 years into its mania" mark is the site's own editorial framing, not a measured figure; no primary source exists for when a mania starts. Peaks are only knowable in hindsight, which is the point.

New here? Start with What Is a Bubble? for the vocabulary and the 15-bubble dataset, or skip straight to Conclusions for where this probably goes. Every card below is one piece of the argument and stands on its own.

Where to start


What people mean by "the AI bubble"

When someone says the AI boom is a bubble, they usually mean some mix of three claims. First, that prices are too high: AI-linked companies are valued as if enormous future profits were certain. Second, that spending is ahead of demand: the industry's capital spendingMoney a company spends on long-lived physical things: buildings, machines, chips, data centers. Spent now, paid back (hopefully) over years.Full definition in the glossary on chips and data centers, on track for roughly $600 to 750 billion globally in 2026 by bank estimates (Goldman Sachs (opens in a new window)), dwarfs the tens of billions of yearly AI revenue anyone can actually count. Third, that the money is moving in circles: suppliers investing in their own customers, who use the money to buy from those same suppliers, a pattern this site calls the AI money loop.

None of those claims is obviously right, and none is obviously wrong. AI revenue is real and growing at triple-digit rates. The companies doing most of the spending are the most profitable businesses on Earth. And people who called bubbles too early have historically paid dearly for it. That is a serious case, and this site presents it at full strength.

Why the question matters

This is not just an argument for investors. AI spending was large enough by 2026 to account for a meaningful share of all US economic growth, with one macro estimate putting AI-related investment near 5% of US GDP (Betafinch (opens in a new window), as of Q1 2026). The ten biggest S&P 500 companies, dominated by AI-linked names, held 37.6% of the index by ticker, or 39.1% counting Alphabet's two share classes as one company, in fund-holdings data as of late August 2026, below the 40.7% "record" this site previously cited, a figure no located methodology reproduces. Anyone with an index fund, a pension, or a job in tech is exposed to how this resolves, whether they follow markets or not.

How this site approaches it

The method is simple: take 200 years of documented bubbles, from the British Railway Mania of the 1840s to the dot-com crash, and measure the AI boom against them on the same dimensions. How fast did prices rise? Who was borrowing, and how? What warning signs appeared before each peak, and which of those signs are visible today? What happened when spending ran ahead of demand? And what finally triggered each collapse?

One recurring finding shapes everything here: in most historical bubbles the technology was real. Railways, electricity, and the internet all changed the world. The bubbles were about prices and timing, not about whether the thing worked. "AI is transformative" and "AI is a bubble" can both be true, and history says they often have been.

Every number on this site links to its original source, uncertainty is kept as ranges rather than flattened into fake precision, and contested figures are flagged as contested. The data reflects a July 2026 snapshot.