Analysis from Dimos Andronoudis about whether we are in an AI bubble and if so what kind of bubble it is.

• Across six bubble metrics, the current AI period has most in common with the GFC era: liquidity has become abundant and investor borrowing has built up, meaning that a sharp downturn would trigger margin calls, amplifying the fall
• On investor over-optimism, dot-com is a closer parallel, pointing to the risk of sharp equity falls if the bubble bursts
• Some reassurance comes from improving US corporate health: both average default probability and the share of firms with a default probability above 10% have declined, unlike during the GFC
• Even if our assessment says that AI is a bubble, it doesn’t tell us when it will burst; Fathom’s central scenario still has equity prices rising, but a 30% chance of a bubble burst in 2027 and a possible oil shock mean that economic and market risks are skewed to the downside
•