AI ‘exuberance’ risks ending in lengthy investment bust, BIS warns

The signal

• Terminology. "Exuberance" is a deliberate choice from a body that picks words carefully. The Financial Times headline is explicit: the downside is a long bust, not a sharp reset.

• Scope. Per Global Banking & Finance Review, the warning is bundled with debt levels and fragilities. This is a macro read from Basel, not a sell-side sector note.

• Source depth. What we have: headlines and framing. The full BIS report and the FT paywalled analysis hold the granular exposures, country splits, and bank-level data. Operate on the framing, not on invented specifics.

Operator math, starting now

• Runway. Model the next round at flat or down. If the board deck assumes flat-up, the deck is wrong.

• Burn multiple. Strip every line item that does not directly compound revenue. Default-alive on the next raise is the only benchmark that survives a credit tightening.

• Cap table. Liquidation preferences compound in a down round. Clean the structure now, before the market forces punitive terms.

• Revenue mix. Three-logo concentration is not a business. It is a single point of failure and a target for procurement renegotiation.

• Headcount. Tie every new seat to a measurable output inside 90 days. Speculative hires are the first line item to get cut when capital tightens.

• Vendor spend. Re-license aggressively. GPU and inference contracts priced on 2025 multiples are an obvious compression target.

The verdict

Offense: cost structure. Defense: cash. The base case is capital reprices, AI multiples compress toward software norms, and the spread between cash-generative operators and story-driven burners widens fast. Plan for that scenario now. The alternative is a bet that Basel is wrong. We do not recommend that bet.