TL;DR
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
JPMorgan, Morgan Stanley, SMBC and MUFG are exploring significant risk transfers for some AI-related lending, according to the Financial Times, as lenders face internal exposure limits. A separate $38 billion loan package tied to Oracle-leased data centers in Texas and Wisconsin has reportedly taken more than six months to distribute. The moves point to tougher financing conditions, but do not establish that banks are broadly withdrawing from AI infrastructure.
JPMorgan, Morgan Stanley, SMBC and MUFG are exploring ways to transfer risk from some of their AI-related lending, according to the Financial Times, as banks approach internal limits on exposure to the sector. The development comes amid a prolonged effort to distribute a $38 billion construction loan package for data centers in Texas and Wisconsin leased to Oracle, Bloomberg has reported, raising questions about how easily lenders can finance the next wave of AI infrastructure.
The banks are considering significant risk transfers, arrangements in which a lender can keep a loan on its books while transferring some of its potential losses to outside investors. The Financial Times reported that the lenders were exploring such deals. The reported motivation is exposure management: banks are nearing internal limits on how much AI-linked risk they can carry, rather than necessarily abandoning the loans or the sector.
Bloomberg has reported that a group of lenders including JPMorgan and MUFG has spent more than six months trying to distribute $38 billion in construction debt tied to an Oracle-leased project in Texas and Wisconsin. Some lenders are reportedly willing to sell portions to non-bank buyers at a discount. The available reporting does not specify how much has been placed, the size of any discount, or whether a completed risk-transfer transaction has occurred.
Credit-market indicators cited in the source report show investors distinguishing among borrowers. Oracle’s five-year credit default swaps rose to about 151 basis points in early October, according to market data cited by Bloomberg; S&P had lowered Oracle to BBB- in July, one notch above speculative grade. Bond data compiled by CCIR put yields on CoreWeave-leased debt at 7.00% to 9.875%, compared with 5.70% to 7.875% for debt leased to larger hyperscalers. These figures describe the cited market observations, not a forecast of future borrowing costs.
How Loan Risk Reaches New Investors
Risk transfers can let banks preserve lending relationships and keep financing projects while reducing the amount of loss exposure they retain. But they also move some risk beyond the banks that originated the loans, making it relevant to investors and other institutions that may buy the transferred slices. The reported effort does not itself show that losses have occurred or that the loans are in default.
The difficulty placing a large construction loan package matters because data center projects require substantial funding before facilities are operating and generating income. If lenders become more selective, borrowers with weaker credit or less established customers may face higher rates, stricter terms, or delays. That is a plausible financing consequence, not a confirmed outcome for any specific project. The reported pricing gap between CoreWeave-leased and hyperscaler-leased debt illustrates how investors may differentiate among tenants.
As an affiliate, we earn on qualifying purchases.
A Large Loan Meets Tighter Markets
AI data center developers and operators have used special-purpose entities to borrow for construction, with debt linked to specific facilities or leases. Such structures can separate project financing from a parent company’s balance sheet, but they do not remove the underlying risk from lenders or investors providing the capital. The source report says AI infrastructure borrowers including Oracle, Meta, xAI and CoreWeave have used these structures.
Oracle has been raising capital to fund data center expansion. The report says the company sold $18 billion in bonds in one day last September and is targeting more than $40 billion in additional debt and equity in the fiscal year, citing reported plans. It also cites Bloomberg data showing order books for bond offerings by major technology companies falling from roughly five times the amount on sale in February to under two times by July. Order-book demand is one measure of investor appetite, but it does not by itself establish whether a company can raise funds or on what final terms.
Broader interest rates are another part of the financing picture. The source report cites a 10-year Treasury yield of 5.304%, its highest since 2002, and says Federal Reserve meeting minutes identified AI-related corporate borrowing as one contributor to a rise in yields. Capital Economics, by contrast, has argued that inflation and government deficits are bigger drivers. The causes of the Treasury move are disputed; higher benchmark yields would still raise the reference cost for many borrowers.
The Scale of Risk Transfers Is Unknown
The reporting does not establish whether any of the four named banks has completed a significant risk transfer, how much debt they intend to shift, or which investors would take on the exposure. It is also unclear how much of the reported $38 billion loan package has been distributed and at what price. The banks’ internal exposure limits and their assessments of individual projects have not been detailed in the source material.
Rising credit protection costs and softer bond order books are market indicators, not proof that Oracle or other borrowers cannot raise capital. Nor does the reported activity show that all AI data center lending is facing the same conditions. The precise effect of higher Treasury yields on particular projects, and the relative roles of AI borrowing, inflation and government borrowing in moving rates, remain unsettled.
Watch Loan Placement and Borrowing Costs
The next evidence will be whether the lenders complete risk transfers or sell loan portions, and whether the Oracle-linked construction debt finds buyers on the reported terms. Further bond sales and market pricing may also show whether investor demand continues to separate established hyperscalers from less-established infrastructure borrowers. The source material does not provide a timetable for any transaction or confirm a forthcoming announcement.
For now, the reported moves indicate that banks are managing concentration as AI infrastructure financing expands. They do not establish a sector-wide lending retreat. Changes in loan placement, credit spreads and the terms available to data center developers will clarify whether this is limited portfolio management or a broader tightening in funding.
Key Questions
Which banks are reportedly seeking to transfer AI loan risk?
The Financial Times reported that JPMorgan, Morgan Stanley, SMBC and MUFG were exploring significant risk transfers related to AI-linked lending.
What is a significant risk transfer?
It is an arrangement that allows a bank to keep a loan on its books while shifting some of the potential losses to outside investors. The reported exploration does not confirm that a specific transaction has closed.
What is the $38 billion Oracle-linked loan?
Bloomberg has reported that a group of lenders has been trying for more than six months to distribute construction debt tied to data centers in Texas and Wisconsin leased to Oracle. The amount placed so far and any sale discounts have not been disclosed in the source material.
Does this mean banks are stopping AI data center lending?
No. The reports describe efforts to manage exposure and transfer risk, not a confirmed halt in lending. They do suggest that lenders and investors may be scrutinizing the size and credit quality of some projects more closely.
Source: hn
Fall Picks
fall essentials
As an affiliate, we earn on qualifying purchases.
