The AI infrastructure construction boom is pushing the world's largest banks to their financing limits. On May 3rd, according to the Financial Times, as data center debt balloons dramatically, major lenders including JPMorgan Chase, Morgan Stanley, and Sumitomo Mitsui Banking Corporation (SMBC) are actively seeking to distribute the associated risks to a broader investor base. This move aims to free up balance sheet capacity and maintain their ability to continue lending.

Reports indicate that, according to informed sources, the banks are exploring private sales of debt stakes and using so-called "significant risk transfer" (SRT) instruments to reduce exposure to single, large borrowers. Matthew Moniot, Co-Head of Credit Risk Sharing at Man Group, stated plainly, "The scale we're talking about... far exceeds anything we've ever contemplated before, and the banks are going to be overwhelmed very quickly."

Analysis suggests this pressure reflects the unprecedented scale of financing demands from the AI industry. Data center operators like Oracle and CoreWeave have borrowed hundreds of billions of dollars to build facilities across the United States for AI labs. This wave is not only testing banks' risk tolerance but also reshaping the entire structure of the data center financing market.

Banks Struggle to Digest $38 Billion Debt, Seek Discounted Sales Informed sources reveal that lenders, including JPMorgan Chase and Mitsubishi UFJ Financial Group (MUFG), have spent over six months trying to syndicate a $38 billion construction loan related to Oracle's data center projects in Texas and Wisconsin.

Some banks have even sought to sell these Oracle-linked loans to non-bank lenders at a discount. This case clearly illustrates the current dilemma facing the banking industry: the sheer size of individual project financings is exceeding the capacity of the traditional syndicated loan model.

Faced with excessively concentrated risk exposures, banks have recently begun tentatively pitching various structured solutions to investors, including variants of SRTs.

SRTs (Significant Risk Transfers) have been widely used by European banks to reduce capital requirements – banks transfer the loss risk of part of their loan portfolio to investors like private credit funds and insurance companies in exchange for corresponding returns. In recent years, North American banks have also begun adopting this tool more frequently.

However, data center-related SRTs differ significantly from the traditional model. Traditional SRTs typically cover dozens of loans, whereas the schemes banks are now exploring involve slicing up a single, massive, and highly concentrated data center loan, transferring the riskiest portions off their balance sheets.

David Lucking of law firm Linklaters mentioned he has seen deals worth $500 million backed by a single borrower, noting that "SRT investors want confirmation that the bank retains a portion of the exposure to ensure alignment of interests."

Frank Benhamou, a Portfolio Manager at Cheyne Capital, pointed out that compared to traditional SRTs, data center-related transactions involve "a limited number of operators, extreme concentration, and significant construction risk," leading investors to demand higher return premiums.

Squeezed by Risk Limits and Regulatory Pressure The fundamental reason banks are eager to disperse risk lies in internal risk limits – these limits cap the maximum exposure a bank can have to a single borrower or a specific industry. Once these caps are hit, banks cannot provide financing for new projects, causing them to miss out on business opportunities presented by the AI infrastructure build-out.

Matthew Moniot stated, "If I were the chief risk officer of a bank, and my bankers were coming to me asking for multi-billion dollar credit lines for individual projects, I would ask them how they are going to distribute that risk."

Simultaneously, growing public opposition to data centers in some parts of the US is further increasing project risks. In April of this year, the state of Maine in the US Northeast passed a statewide ban on data centers, adding new uncertainty to related financing.