> Wouldn't the lender now also be in harms way and in turn have issues financing themselves?

Yes, they could be. There are potential multipliers here, and one failed project could cause other projects to be marked down, even if they're not failing.

But the mechanisms by which $1 trillion in data center debt could become multi-trillion dollar write downs just doesn't exist the same way it did with subprime mortgages.

In the mortgage market, credit default swaps let the same collateral be referenced without limit, CDO-squared structures re-tranched the losses into concentrated wipeouts, and the paper was on the balance sheets of firms funding 30 year assets with overnight repo.

A derivative market for individual SPVs where there's a mismatch between the debt term and the asset just doesn't exist in the data center market.

Pensions funds that are involved have 30 year liabilities and no redemption pressure. Insurers don't have run risk. Banks hold the debt against capital buffers and they have access to the discount window. The riskiest debt holders are levered private credit and semi-liquid retail vehicles, but even in these cases, you're looking at leverage around 2:1, not 30:1.

None of this is to say that an AI bust couldn't cause widespread pain. Just look at how much of US equities are dominated by AI-linked companies. But the issue is that Zitron is just wrong about the mechanisms and magnitudes. He's trying to paint a 2008-like scenario because it's much scarier.