AI Credit Cycle Risks Compare to 2008 Subprime Crisis

A financial analysis warns that the current artificial intelligence boom is structured as a credit-driven cycle mirroring the 2008 subprime mortgage crisis. The industry is currently in a teaser period of deferred costs, where trillions of dollars in take-or-pay compute contracts signed by frontier labs remain in a construction phase with payments not yet due.

OpenAI and Anthropic have signed massive commitments, but a reset wall is expected in 2027–2028 as these contracts transition from booked backlogs to actual billing. The analysis suggests OpenAI's commitments may exceed its plausible revenue paths, potentially leaving its equity underwater and dependent on continuous refinancing through capital markets. In contrast, Anthropic's commitments are projected to be covered by revenue by 2027.

Cloud providers, including Microsoft, Alphabet Inc., and Amazon Web Services, face a secondary reset as massive capital expenditures shift into recurring depreciation charges. Oracle has already seen its remaining performance obligations grow 363% in a single fiscal year. While proponents argue the transformative potential of AI justifies the spending, critics maintain the current investment economics are unsustainable and risk a systemic re-rating of AI assets.