Banks are now trying to offload the AI data center debt they just underwrote

The cost of financing new AI data center projects is about to rise significantly for less-established companies.

Major banks including JPMorgan, Morgan Stanley, SMBC, and MUFG are attempting to offload significant portions of AI-related data center debt, totaling billions of dollars. These institutions initially underwrote construction loans for projects leased to Oracle, primarily in Texas and Wisconsin. After six months of struggling to distribute this $38 billion in debt, some lenders are now willing to sell pieces at a discount to non-bank investors. This move indicates that banks are hitting internal exposure limits for AI-linked assets and are seeking to transfer risk, even at a lower price. Oracle's credit rating has been downgraded to BBB-, and its credit default swaps have reached levels not seen since 2009, reflecting increased concerns about the company's ability to service its substantial debt obligations.

The banks' eagerness to offload this debt, particularly when it involves selling at a discount, signals a growing caution in the market regarding the long-term viability and creditworthiness of AI infrastructure projects not directly tied to top-tier hyperscalers. This shift suggests that the perceived risk in AI data center development has increased, pushing investors to demand higher yields or to focus solely on the most secure tenants like Microsoft and Google.

This development indicates that the easy money for AI infrastructure buildout may be drying up, especially for companies without the balance sheets of giants like Oracle, Meta, or Amazon. As banks become more risk-averse, the cost of capital for emerging AI infrastructure players will likely escalate, potentially slowing down expansion or forcing them to seek alternative, possibly more expensive, funding avenues.