Banks are scrambling to offload $38 billion in high-risk data center debt as AI infrastructure borrowing pushes lending limits. JPMorgan Chase, Morgan Stanley, and SMBC are using private deals and risk transfers to unload stakes, with Oracle-linked projects in Texas and Wisconsin at the epicenter. Matthew Moniot of Man Group called the scale “out of scale to anything we’ve thought about,” highlighting how lenders are “choking” under pressure. CoreWeave and Oracle have borrowed hundreds of billions for U.S. AI labs, forcing banks to innovate risk-sharing models. Frank Benhamou of Cheyne Capital noted that SRTs (Significant Risk Transfers) are now tailored to slice concentrated data center loans, with investors demanding higher returns for assuming construction risks. Carlos Mendez of Crayhill Capital warned banks are “having to find more counterparties” as public opposition, like Maine’s data center ban, adds volatility. Lenders are also turning to private credit and asset-backed securities to bypass traditional financing constraints.