Fitch models 35 percent Wall Street drop in AI crash scenario

Fitch and the BIS are not saying an AI crash is inevitable, but they are quantifying what happens if investors decide frontier labs will not deliver the profits baked into current valuations. A 35 percent drawdown in U.S. equities tied to AI and a global growth slump below 1 percent would not just hurt shareholders, it would cool capital flows into compute, model training and the vast supply chains that feed this ecosystem. The piece makes clear that AI related capex already accounts for roughly a third of U.S. growth this year, and that European households are heavily exposed through holdings of U.S. tech stocks.

In AGI terms, this scenario is a reminder that financial cycles can be as constraining as physics. A sharp correction would likely slow the most capital intensive bets, such as trillion parameter frontier models and hyperscale data centers, while leaving more room for leaner, open source and domain specific systems. It could also push labs to demonstrate clearer unit economics rather than treating ever larger models as their own justification. At the same time, history suggests that foundational R&D often continues through downturns; the question is whether shareholders will tolerate multi year payback periods once the easy growth narrative breaks. If they do not, government funding and strategic investors may become even more important in setting the pace to AGI.