The AI Bubble Watch carries a number: the chance, as this page judges it, that the AI bubble bursts within the next twelve months. This page explains how that number is built.
First, what a burst means here. A fall of 25% or more in the AI-heavy end of the market, led by the AI story breaking. Not a bad week. Not a correction. The story itself giving way.
Where the number starts
It starts from the base rate. In an average year, the chance of a drop that size is roughly one in ten. So the number starts at 10%, and five gauges push it up or down from there. The gauges are scored the same way every time. That discipline is the whole method.
The five gauges
The gap. The big four technology companies plan to spend about $730 billion on AI infrastructure in 2026. Against that, the AI revenue they have actually disclosed adds up to about $62 billion. Roughly twelve dollars out for every dollar shown coming in. The wider the gap, the higher this gauge pushes.
Valuation. The Shiller CAPE ratio — the market’s price against ten years of earnings — stands at 40.7. Its long-run average is 17.4. Its all-time peak, set in December 1999, is 44.2. Expensive markets do not have to fall. They just have further to fall.
Concentration. The ten largest stocks now make up 39.2% of the S&P 500. When a handful of companies carry that much of the index, their story is the market’s story. If it breaks, there is nowhere to hide.
Earnings quality. All four companies are still passing their earnings tests — holding or raising their spending plans, reporting strong results. But investor Michael Burry has published a detailed critique: by stretching the assumed useful life of their chips and servers, the companies understate depreciation, by his count $176 billion across 2026 to 2028. Passing grades, partly on accounting. This gauge sits at mixed.
Funding. The borrowing behind the build-out still looks calm on the surface. Underneath, Burry counts roughly $3 trillion in purchase commitments, leases, and guarantees across the five biggest builders, much of it off the balance sheet, with data-center debt resting on equipment that goes obsolete in a few years. This gauge sits at moderate, and rising.
| Component | Points |
|---|---|
| Base | 10 |
| Gap | 12 |
| Valuation | 10 |
| Concentration | 8 |
| Earnings | 5 |
| Funding | 5 |
| Total | 50% |
What moves the number
The number is reviewed daily. It moves only on verified news, never on mood. The moves are set in advance, so the number cannot drift with the headlines:
- A spending guidance cut by a major builder: up 8 points
- A credit event in AI funding: up 8
- Write-offs or restated depreciation begin: up 10
- The accounting challenged by auditors or regulators: up 6
- An AI revenue miss at earnings: up 5
- The gap widening on new figures: up 3 to 6
- The CAPE passing its 1999 peak: up 3
- A clean earnings sweep, guidance held and revenue growing: down 4
- A material jump in disclosed AI revenue: down 3 to 6
- The gap narrowing two quarters in a row: down 3 more
- The CAPE cooling below 35 without a break: down 3
The number never goes below 5% or above 95%. Certainty is not on offer, in either direction.
What the number is not
It is not a forecast. It is not a measurement the way a thermometer measures. It is this page’s judgment, built from the five gauges above, scored the same way each time, with every move dated and explained in the Changes log on The AI Bubble Watch.
Two of the gauges lean on published work by Michael Burry of Scion Asset Management, credited where they appear on The AI Bubble Watch. Burry has published no probability of his own. The number is ours.
The Burst Risk number is reviewed daily by the Financial Spark staff. Figures as of October 30, 2026. Sources: company earnings reports, multpl (Shiller data), S&P Dow Jones Indices, published analysis by Michael Burry.