Big Tech keeps $300bn of AI exposure off its balance sheets, FT reports
Guarantees issued in under a year back debt for data centres and chips while recording little of the exposure, and Nvidia's backing of an OpenAI site is the largest cited.
Big Tech companies are rapidly expanding their use of guarantees to back debt for AI data centres and chips, the Financial Times reported on Sunday. They have issued up to $300bn in commitments in less than a year while recording little of that exposure on their balance sheets. The paper called it a new way for Wall Street to turn the companies' credit strength into cheaper funding.
Posts summarising the report give the components. Nvidia is guaranteeing up to $105bn behind OpenAI's Ohio data centre project, according to one such summary. Broadcom has up to $29bn of exposure tied to AI racks for Anthropic. Meta carries a residual-value guarantee of about $28bn tied to its Hyperion site. We have not seen the article itself.
The figures sit inside a larger estimate. One post reproducing an FT chart cites Morgan Stanley's estimate of $3.1tn in total off-balance-sheet commitments across the hyperscalers. Bloomberg published its own piece on Sunday describing AI's influence on markets and the US economy as a house of cards. Any spending slowdown may be devastating, it said.
The report follows the FT's account on Thursday of an OpenAI presentation projecting $856bn of compute and infrastructure spending and $278bn of negative free cash flow through 2030. A guarantee is not a loss, and none of the reports says any has been called. Replies to the report ask who takes the first loss if utilisation or pricing slips before the next refinancing.