Hayes's core logic is not about whether "AI will make money," but rather "who is financing AI." He has consistently linked macro liquidity, dollar credit, and crypto prices, and this time he redefines AI data centers as quasi-real estate assets, shifting the discussion from tech valuation to leverage and collateral quality.

In terms of capital pathways, AI capital expenditure mainly flows to data centers, substations, cooling, networks, and land, which are more like long-term recovery assets that rely on financing for survival; once financial institutions, private credit, and governments participate in the expansion, the credit cycle will resemble a 2008-style balance sheet shock more closely than the internet bubble.

Comparable historical cases include railroads, power grids, and subprime mortgages: initially viewed as "new economies," they later triggered repricing due to rapid financing and slow returns. Hayes emphasizes not model failure, but rather that the credit chain breaks first, indicating that the current AI industry is more like building a capital-intensive system that requires continuous rolling financing.

Essentially, this is a liquidity repricing: if the AI bubble bursts, the first to be hurt will be heavy assets and credit markets, followed by central banks likely using looser policies to repair balance sheets; the reason this happens is that in modern finance, the biggest risk is not technological failure, but rather that the narrative around technology has pushed credit expansion to its limits.

ABAB News · Law of Cognition

The end of a new story often marks the end of old leverage.
Assets are first defined by financing, then corrected by cash flow.
Credit bursts first, liquidity follows.