Research paper · November 2026
Lending Against Machines Nobody Appraises
Abstract
Between July 2025 and October 2026, AI cloud operators borrowed tens of billions of dollars in facilities described as secured by graphics processors. This paper reads the public record for eleven of them, plus one corporate secured facility included as a comparator, and codes what each discloses. The record comprises issuer releases, SEC Form 8-K and 6-K filings, registration statements, rating-agency releases and, for one facility, the full credit agreement. Pricing is always disclosed. In this sample it sorts cleanly by the credit of the customer behind the loan. Facilities tied to investment-grade contracts priced at SOFR plus 2.25 to 3.00 percentage points, or 5.9 to 6.8 percent fixed, while facilities tied to sub-investment-grade or unrated customers priced at SOFR plus 4.00 to 5.50, or 9.0 to 9.95 percent fixed. What the record does not contain is any figure that would let an outside reader value the machines themselves. No facility in the sample discloses an advance rate against appraised value, a market residual value assumption, an obsolescence or refresh covenant, or any verification by a party other than the borrower that the collateral exists and performs. The one credit agreement read in full defines a six-year straight-line depreciated amount for the GPUs and requires the borrower to report rack serial numbers and installation progress, which is the closest the record comes. Three disclose a loan-to-cost figure. One discloses that renewal and re-lease criteria exist, with their key term redacted. Data-centre securitisations rated by the same agencies report appraised loan-to-value, capitalisation rates and haircuts as a matter of routine. The paper argues that these instruments are priced as contract credit and described as asset-backed, and proposes five disclosure fields that would let the collateral be priced in its own right.
Cite
Abhay Chakra Sadineni (2026). Lending Against Machines Nobody Appraises. Research paper. https://doi.org/10.5281/zenodo.24210925 https://www.abhaychakra.com/papers/lending-against-machines-nobody-appraises/