Shadow credit: $70 billion quietly props up the AI buildout

The AI boom is increasingly financed through structures that never show up directly on a corporate balance sheet. According to calculations reported by Bloomberg this weekend, AI companies and their partners have built up roughly $70 billion in so-called shadow credit backstops — commitments that secure bonds and loans for data-center projects without the tech firms taking on the debt themselves.

The pattern is usually the same: a project vehicle builds the data center and raises debt to do so. A tech company guarantees lease payments or agrees to step in through backstop arrangements if the project cannot service its obligations. Formally, the corporate balance sheet stays clean — economically, the company still carries the risk.

That is why unease is growing in the bond market. Traders and analysts are asking how robust these guarantees would prove under stress, and how many further commitments exist that have not yet been quantified publicly. Rating agencies treat the structures inconsistently, which makes it harder to assess true leverage.

The figure fits a larger picture: AI infrastructure spending is estimated in the hundreds of billions of dollars for 2026, and a growing share runs through private credit funds, bonds and guarantees rather than classic corporate financing. Just this week, Nvidia was reported to have adjusted its backstop commitment for an OpenAI site in Ohio.

For investors, the core question is what happens if revenue from AI services grows more slowly than the cost of capital. Theoretical guarantees would then become real payment obligations — and invisible leverage a visible problem. The reports do not point to any acute distress; the warning is about opacity, not an actual default.