Banks Move to Offload Billions in AI Data Center Debt as Exposure Limits Loom

Major global financial institutions have begun seeking buyers for portions of the vast debt portfolios accumulated through underwriting artificial intelligence infrastructure, signaling growing caution among the primary lenders funding the AI expansion. Financial entities including JPMorgan Chase, Morgan Stanley, Sumitomo Mitsui Banking Corporation (SMBC), and Mitsubishi UFJ Financial Group (MUFG) are exploring avenues to offload or hedge their credit exposure, as detailed in reporting highlighted by Hacker News (https://startupfortune.com/banks-are-now-trying-to-offload-the-ai-data-center-debt-they-just-underwrote/).

The shift follows a competitive period earlier this year when investment banks rushed to finance multi-billion-dollar data center developments tied to technology giants such as Oracle and Meta. Lenders are now attempting to execute significant risk transfers—transactions that allow banks to retain loans on their balance sheets while shifting the highest-risk tranches to third-party investors—due to internal credit limits that cap their exposure to AI-related liabilities.

Distribution challenges have already emerged in large syndicate deals. A banking group led by JPMorgan and MUFG has spent over six months attempting to syndicate $38 billion in construction loans for data center facilities in Texas and Wisconsin leased to Oracle, according to Bloomberg reporting. To clear the balance sheet pressure, some underwriting institutions are offering portions of the debt package to non-bank institutional buyers at discounted rates. Describing the magnitude of current market offerings, Matthew Moniot, co-head of credit risk sharing at Man Group, told the Financial Times that transaction volumes are "out of scale to anything we've thought about, ever."

Oracle's financial commitments have become a focal point of market anxiety. Credit default swap spreads for Oracle's five-year debt expanded to approximately 151 basis points in early October before reaching 203 basis points, an 18-year high. The widening spreads follow a July decision by S&P Global Ratings to downgrade Oracle's credit rating to BBB-, one level above non-investment grade. The rating action reflected cash flow pressures stemming from aggressive AI capital expenditures, compounded by a legal dispute in which Oracle sued the Public Service Commission of Wisconsin over a directive requiring more than $7 billion in collateral for its Port Washington campus. The enterprise software provider is currently seeking to raise more than $40 billion in combined debt and equity this fiscal year to support facilities across Texas, Wisconsin, and New Mexico, following an $18 billion single-day bond issue last September.

Investor appetite for high-volume tech debt shows broad-based signs of cooling. Corporate bonds issued by major industry players—including Alphabet, Amazon, Meta, Microsoft, and Oracle—are currently trading at yield premiums compared to similarly rated debt from non-technology corporations. According to Bloomberg data, average order book oversubscriptions for bond sales from these firms declined from approximately five times offered amounts in February to below two times by July. A recent bond issuance by Meta generated one-third less investor demand than a comparable sale conducted by the company a year earlier.

The cooling demand coincides with broader macroeconomic headwinds in fixed-income markets. Benchmark 10-year U.S. Treasury yields rose to 5.304% this month, reaching their highest point since 2002, following a $39 billion Treasury auction that cleared at 5.30%, the highest auction yield recorded since 2000. Minutes from the Federal Reserve's September policy meeting highlighted corporate borrowing for AI infrastructure as a contributing factor to rising yields, though research firm Capital Economics attributed the upward move primarily to broader fiscal deficits and persistent inflation. Regardless of the driver, elevated benchmark yields have raised overall borrowing costs for technology infrastructure financings.

To keep debt off primary balance sheets, technology firms including Oracle, Meta, xAI, and CoreWeave have increasingly utilized special purpose vehicles (SPVs) to fund facility development. However, this structure has concentrated risk among the commercial banks underwriting the SPVs. In secondary credit markets, institutional capital is prioritizing investment-grade hyperscalers such as Microsoft and Google over lesser-capitalized borrowers. Yields on debt tied to CoreWeave facilities currently range between 7.00% and 9.875%, compared to yield spreads of 5.70% to 7.875% for facilities leased to top-tier hyperscalers, based on market data compiled by CCIR.

The developing bottleneck in debt syndication indicates that the rapid capital deployment driving AI expansion is encountering stricter financial terms. As underwriting banks enforce exposure caps and institutional buyers demand higher yields or steeper discounts, financing conditions for upcoming data center projects are tightening across both major hyperscalers and specialized infrastructure operators.