Matt Barrie warns $1.65T AI debt bubble risks subprime crash

The AI labor panic just collided with hard economic data.

Despite headline-grabbing corporate layoffs, artificial intelligence created nearly one million net new US jobs, according to economist Alex Domash on The Intelligence from The Economist. The surge split between 350,000 blue-collar infrastructure roles building data centers and 750,000 white-collar engineering positions. On The AI Daily Brief, Anthropic head of economics Peter McCrory confirmed that 18 months of user data shows no spike in aggregate unemployment, as workers use AI to expand output rather than replace colleagues.

"The narrative of instant AI mass unemployment just hit reality."

- Alex Domash, The Intelligence from The Economist

Yet beneath this hiring boom lies a precarious financial structure.

Freelancer.com CEO Matt Barrie warned on Macro Voices that Wall Street and tech firms built a $1.65 trillion debt bubble over five years to fund AI data centers. Using special purpose vehicles and private credit to hide balance sheet exposure, this debt pile eclipses the $1.3 trillion subprime mortgage crisis of 2007. Worse, the entire buildout relies on a hyper-concentrated customer base: OpenAI and Anthropic generate nearly three-quarters of all cloud AI revenues.

"Special purpose vehicles and private credit hide massive balance sheet exposure for firms like Meta."

- Matt Barrie, Macro Voices

To control soaring token costs, enterprise software strategy is pivoting sharply. Barrie noted burning four billion tokens in a single day across 40 AI agents, running up a $1,300 daily cloud bill. In response, companies are migrating to custom small models or localized hardware like Nvidia DGX Spark clusters. On The AI Daily Brief, Microsoft outlined switching enterprise tools to specialized internal models, cutting operational costs by 84 percent while extending chip lifespans.

A day later on What Bitcoin Did, fund manager Jeff Ross framed the infrastructure surge as an economic escape hatch rather than an impending collapse. Ross argued Washington will treat data center buildouts as vital national security assets, backing them through central bank liquidity injections by early 2027. In his view, rapid productivity gains will spark a domestic manufacturing expansion that generates price deflation across consumer goods, healthcare, and energy.

Whether AI spending fuels a historic productivity boom or triggers a debt default depends on who blinks first: lenders adjusting interest rates or hyperscalers desperately hunting for margin.