Banks are now trying to offload the AI data center debt they just underwrote
Which summary reads better? Pick one — models revealed after.Both summaries are AI-generated.
JPMorgan, Morgan Stanley, SMBC and MUFG are trying to shift risk on AI data-center loans, including a $38 billion Oracle-leased Texas/Wisconsin construction debt package that banks have been unable to distribute for more than six months. The takeaway is that financing for the AI buildout is no longer being treated as easy balance-sheet exposure: lenders are hitting concentration limits, some debt is being offered at discounts, and Oracle’s rising CDS spreads and BBB- rating signal higher funding costs and potential delays for new capacity.
The banks that underwrote the AI data center buildout are now hitting internal AI-exposure limits and dumping the risk — $38B in Oracle-leased construction debt has sat undistributed for six months, with lenders now selling at a discount, and Oracle's CDS hit 151bps (2009 levels) at BBB-, one notch above junk. The capital underwriting the compute capacity you're planning to rent is being repriced as distressed, which means tighter financing, slower new-build timelines, and real counterparty risk on the hyperscaler/neocloud contracts your roadmap depends on.
AI vs. AI Debate
“The summary is strong on the core numbers but overstates the debt as “distressed” and extrapolates too far into hyperscaler/neocloud contract risk beyond what the excerpt directly supports.”
“While "distressed" is interpretive shorthand, CDS at 2009 levels and a six-month undistributable $38B package being offered at discounts constitute distress-pricing signals, and flagging downstream counterparty risk is a reasonable and useful inference for readers whose roadmaps depend on these very contracts.”