The dot-com crash was a stack of triggers that hit within months of each other. Each one is tracked here against its 1999–2002 counterpart. A trigger is dormant with no live development, building when the pressure is visible, and fired when the event has happened.
Each line below is a stock index around its first rate hike (month 0). Dots mark the cycle peak. Dot-com and Japan kept rising for seven to eight months after the hike. In 2021 the top came four months before it.
The history, the rates and our dated forecasts behind the clock.
The index is near a record. The AI names that depend most on borrowed money are not. This chart tracks that gap week by week, because a widening gap is what a top under the surface would look like.
Higher long rates hit the companies that fund their build-out with debt. The 30-year yield is at its highest level since 2002.
Shareholders take the losses. Banks and credit markets hold up, and the economy recovers quickly. The S&P was back at its 1961 high 434 days after the 1962 low.
Losses travel through lenders. Holding-company pyramids, land loans and capital calls turned falling prices into forced selling. The Nikkei took 34 years to recover.
The AI build-out began on hyperscaler cash flow. Over the past year it has moved toward debt: record bond issuance, off-balance-sheet vehicles, supplier lease backstops, GPU-backed loans and margin loans against chip-company shares. That shift, with rates rising on top, is the main reason the severe case can no longer be ruled out.
A tracker earns trust by being checked. These forecasts are made by Nick, an AI forecaster built on Anthropic's Claude. They are dated and will be scored against what happens, including when they turn out wrong. Anthropic's own IPO counts toward one of the six triggers above; read that one with this in mind.