Original post · Original
Rohan Paul@rohanpaul_ai
Big tech’s AI spending, measured against the size of the economy, is following the same path as the 1990s telecom boom, and it’s speeding up just as the companies start borrowing heavily to pay for it.
The bigger exposure doesn’t show up as debt. Amazon, Alphabet, Microsoft, Meta, Oracle, Nvidia and Broadcom have taken on over $3T in future leases and guarantees, mostly in the last 2 years, so if the AI payoff disappoints, they may owe far more than their balance sheets show.
Fed releases financial stability report
> AI is now the top financial risk for Wall Street. 87% of the market pros the New York Fed polled named it, up from 50% in spring.
> The Fed says “market enthusiasm for super intelligence” is part of why stock valuations are so high.
> Debt raised for AI infrastructure passed $680B by October. That’s about 5x all of 2023 and about 31% of investment-grade corporate borrowing.
> Big tech and chipmakers also carry over $3T in future leases and guarantees, so their real AI exposure may be far larger than their reported debt.
> GPU makers guarantee the value of the chips they sell to AI firms and often invest in those same buyers, which creates “circular financing.”
> The largest US banks have close to $1T of exposure to big tech, data center builders and the vehicles that finance them.
> Hyperscaler spending as a share of GDP may soon pass the 1990s telecom boom.
View this post on X ↗
· UTC+8