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
- Foundations What Is a Bubble? The five-phase anatomy, and a table of 15 historical bubbles: how far they fell and how long recovery took.
- Foundations Historical Case Studies Six deep dives, from the Railway Mania to the housing crash, each ending with its echoes in the AI boom.
- The Mechanism Circular Financing, Explained How booms fund their own demand: vendor loans, share calls, cross-shareholding, trusts holding trusts.
- The Mechanism The AI Money Loop The current map: Nvidia funding its customers, the Microsoft-OpenAI circle, GPU-backed debt.
- The Evidence Building Ahead of Demand Railways, grids, fiber, and now data centers: what happens when capacity outruns demand.
- The Evidence The Warning Lights Nine historical warning indicators and their current readings: flashing, amber, quiet, or not scoreable.
- The Evidence The Demand Question Is real revenue showing up? Company by company, plus the GPU depreciation controversy.
- The Evidence Outside Shocks Oil, Taiwan, the power grid, and the US fiscal position: what could pull the trigger.
- Weighing It The Case Against the Bubble The strongest version of the other side, including how often bubble calls have simply been wrong.
- Weighing It The Critics Ed Zitron's claims, his sourcing, and his track record: what held up and what didn't.
- Weighing It Conclusions: Where This Probably Goes The scorecard, the closest historical analog, and three scenarios you can watch for yourself.
- Reference About & Methodology How this site was made, what it can and can't tell you, and the July 2026 data snapshot.
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.