We build a novel property-level database of U.S. datacenter mortgages, documenting a source of AI-buildout financing that has received little attention. Unlike corporate debt, which is observable through public bond and loan disclosures, property-level mortgage lending is fragmented across property records, local lenders, and non-public loan documents. By linking these mortgages to lender balance sheets, we document a large, recent, and highly concentrated market. The market is dominated by banks with a high propensity to originate loans for distribution rather than retention, that is, originate-to-distribute (OTD) banks, and by Undiversified Datacenter borrowers, firms whose repayment capacity depends on continued AI demand for datacenter capacity rather than diversified revenue. We then ask whether OTD incentives relax lending standards in this market. We find that High-OTD lenders extend substantially higher leverage and finance more Undiversified Datacenter deals. Moreover, the leverage premium itself concentrates among these same borrowers.