Robert Kuttner writes:

A new report to be presented today at the Brookings Institution shows that the total cost of the AI build-out through 2032 will be $10.3 trillion. That amounts to 3.63 percent of GDP per year, many times that of previous large-scale infrastructure investments, from the 19th-century canals, the railroads, telephone, and electric grids, to the interstate highway system.

And unlike AI, whose benefits are speculative, unproven, and with hard-to-fathom risks, these earlier investments in infrastructure more than paid for themselves in productivity gains to the broader economy. The Brookings report also raises the alarm that much of the funding for the AI build-out is financed by debt, some of it unknown because it is hidden in off-balance-sheet financing. That immense AI debt also crowds out more productive borrowing, and raises interest rates.

Some of this debt will never be repaid. And a great deal of both the stock market boom and the surprising continued GDP and job growth rests on the artificial and unsustainable stimulus of the AI bubble.

Combined with citizen pushback against giant data centers, the concern of some AI executives that they have overreached and need an excuse to pause, and the rising interest rates due to Fed policies and bond market panic, the stage is set for a financial collapse comparable in scale to 2008 or 1929.

On Thursday, right on cue, it was revealed that Oracle, which is behind on its plans to develop a massive AI data center in New Mexico, sent a notice to the center’s property developer, Blue Owl Capital, that it would be unable to meet its payments on time. Oracle is $18 billion in debt on the deal. Astonishingly, Oracle cited “force majeure,” a boilerplate provision in contracts excusing failure to perform usually reserved for wars and weather disasters. [Continue reading…]

Kim Forrest, founder & CIO at Bokeh Capital Partners, joins BNN Bloomberg to discuss how data centre uncertainty is clouding AI trade.: