Glonce

Banks try to offload AI data center debt, including $38 billion tied to Oracle

businessThe Glonce editorsHacker News🔥77Glonce rating77Impact65Novelty60Relevance70Surprise65Players80Credibility80
Banks try to offload AI data center debt, including $38 billion tied to Oracle

According to a startupfortune.com article drawing on reports from the Financial Times and Bloomberg, some of the banks that rushed to underwrite AI data center loans are now trying to move that debt off their books. JPMorgan, Morgan Stanley, SMBC and MUFG have been exploring significant risk transfers, per the FT. These are deals that let a bank keep a loan on paper while shifting its riskiest slice to outside investors. The article says the banks are doing this because they are bumping up against internal limits on how much AI-linked exposure any one lender can hold.

The clearest case is a $38 billion package of construction debt tied to a Texas and Wisconsin data center project leased to Oracle. Bloomberg has reported that a bank group including JPMorgan and MUFG has spent more than six months trying to distribute it. Some lenders are now willing to sell pieces to non-bank buyers at a discount just to move it. Matthew Moniot, co-head of credit risk sharing at Man Group, told the FT the deal sizes are "out of scale to anything we've thought about, ever."

Oracle's own credit signals have worsened. Its five-year credit default swaps, the cost of insuring against a default, climbed to around 151 basis points in early October, levels last seen in 2009, according to market data cited by Bloomberg. A caption at the end of the article gives a different figure, 203 basis points, called an 18-year high, and the two numbers are not reconciled. S&P cut Oracle's rating to BBB-, a single notch above junk, in July. Oracle is reportedly targeting more than $40 billion in additional debt and equity this fiscal year, on top of the $18 billion in bonds it sold in a single day last September, to keep funding its Texas, Wisconsin and New Mexico sites. Separately, Oracle sued Wisconsin's Public Service Commission over a rule that could force it to post more than $7 billion in collateral for its Port Washington campus.

Demand for hyperscaler paper has thinned. Bonds from Alphabet, Amazon, Meta, Microsoft and Oracle now trade at a premium to similarly rated non-AI corporate debt. Order books for their offerings shrank from roughly five times the amount on sale in February to under two times by July, per Bloomberg. Meta's latest bond drew a third less investor demand than its sale a year earlier. Capital is rotating toward data centers leased to investment-grade hyperscalers such as Microsoft and Google and away from weaker credits, which is why CoreWeave-leased paper prices at 7.00% to 9.875% against 5.70% to 7.875% for hyperscaler-leased bonds, per CCIR bond market data.

The wider rate backdrop is tight. The 10-year Treasury yield hit 5.304% this month, its highest since 2002, after a $39 billion auction cleared at 5.30%, the highest auction yield since 2000. The Fed's September meeting minutes named AI-related corporate borrowing as a contributor to that move, while Capital Economics argues inflation and government deficits are the bigger driver.

The article's own argument is that banks do not sell down debt they are confident in, and that spreading $120 billion of AI infrastructure spending across special purpose vehicles, the structure Oracle, Meta, xAI and CoreWeave have used to keep debt off their balance sheets, has moved the risk back to the banks that financed those vehicles. It calls six months of struggling to sell $38 billion the clearest evidence yet that the assumption of open credit is being tested. That reading is the author's opinion, not a statement from any bank.

Key facts

  • JPMorgan, Morgan Stanley, SMBC and MUFG have been exploring significant risk transfers on AI-linked exposure, per the Financial Times; the article says they are nearing internal exposure limits.
  • A bank group including JPMorgan and MUFG has spent more than six months trying to distribute $38 billion of construction debt tied to an Oracle-leased Texas and Wisconsin project, and some lenders will sell pieces to non-bank buyers at a discount.
  • Oracle's five-year CDS reached around 151 basis points in early October per Bloomberg data, levels last seen in 2009; S&P cut its rating to BBB- in July.
  • Order books for Alphabet, Amazon, Meta, Microsoft and Oracle offerings fell from roughly five times the amount on sale in February to under two times by July.
  • CoreWeave-leased paper prices at 7.00% to 9.875% against 5.70% to 7.875% for hyperscaler-leased bonds, while the 10-year Treasury yield hit 5.304%, its highest since 2002.

Why it matters

The AI buildout has leaned on bank-financed construction debt and special purpose vehicles. If the lenders closest to that money are seeking to shed exposure, the financing that data center projects depend on becomes harder and costlier. The article frames the stalled $38 billion Oracle-linked deal as evidence that credit for AI infrastructure is being tested now, not in some distant downturn. That framing is the author's view, but it rests on reported facts: six months without full distribution, discounted sales to non-bank buyers, and shrinking order books.

Who it affects

Oracle is the most exposed name in the story, with a BBB- rating, a reported $40 billion-plus funding target and a collateral dispute in Wisconsin. Banks such as JPMorgan, Morgan Stanley, SMBC and MUFG carry the underwriting exposure. Weaker or smaller AI infrastructure borrowers, CoreWeave among them, face wider pricing and compete for the same cautious pool of lenders. Hyperscalers with investment-grade tenants like Microsoft and Google are where investors are reportedly rotating.

How to use it

This is a market story, not a tool. The article points to indicators worth following: Oracle's credit default swap levels, order-book coverage on hyperscaler bond sales, the pricing gap between hyperscaler-leased and CoreWeave-leased paper, and the 10-year Treasury yield. Anyone planning to finance data center capacity should expect the article's picture of pickier lenders and higher benchmark rates.

How solid is it

The article is a secondary summary of Financial Times and Bloomberg reporting, plus CCIR bond data and Fed minutes. It gives no exact dates for those reports. It contains an unreconciled inconsistency: around 151 basis points for Oracle's five-year CDS in the body and 203 basis points in a closing caption. No bank has confirmed a specific risk-transfer deal; the source says only that they have been exploring them, per the FT. The bank group behind the $38 billion debt is named only as including JPMorgan and MUFG. Statements such as banks not selling down debt they are confident in are the author's commentary.

Risks and caveats

The source does not say how much of the $38 billion has been sold, what discount is being offered, or what the banks' internal exposure limits are. It does not say that any bank has taken a loss or that any borrower is in default. Capital Economics argues inflation and deficits, not tech bond sales, drive the Treasury yield move, so the rate backdrop may not reflect AI borrowing. The Oracle collateral rule is something that could force a $7 billion-plus posting, not one that has.

“out of scale to anything we've thought about, ever.”

— Matthew Moniot, co-head of credit risk sharing at Man Group, to the Financial Times, on the deal sizes involved