Despite investors' fears, private debt is far from collapsing because not all risks are the same.
The cracks in the private credit market seem to be widening.
Private credit is an alternative to syndicated bank loans as a source of business capital provided mainly by private equity (PE) firms. The market is heavily involved in financing data center capacity, which is growing in strength and demand for artificial intelligence. Investors fear that rising costs of artificial intelligence threaten the software industry and could create a market bubble that leaves private debt funds overexposed.
Yet there are reasons to believe that the potential damage to the sovereign debt market is still manageable and contained.
This article appears in the May 2026 issue of Global Finance Magazine. .
To be sure, when auto parts retailer First Brands announced bankruptcy late last year, which was backed by a debt fund backed by investment bank Jefferies Group, it raised alarm bells elsewhere. Underscoring the bleakness of private credit, which is largely unregulated, were allegations that First Brands had borrowed against the same receivable more than once. Meanwhile, defaults in other areas of the credit sector reached a record high in 2025, according to Fitch Ratings, reaching a rate of 9.2%, more than double the 3.6% recorded in 2023. Default rates continued to rise this January, reaching 9.4% before falling slightly in February to 5.4%.
As the financing of First Brands reveals, banks and PE firms are involved in private debt, either by financing investment funds sponsored by Ares Capital, Antares, Apollo, Blackstone, Blue Owl, and the like, or with their own funds. With pension funds, insurance companies, and increasingly, individuals investing in private debt, law firm Quinn Emanuel warned in a March client memo that the trend could create systemic risk, even though private debt is still a small part of the overall loan market.
“The result is a chain of transmission from technology companies, through private credit originators, to regulated banks that borrow, to insurers and pension funds that invest alongside them, and potentially to the retirement accounts of ordinary Americans,” the memo's authors warn.
Only a minority of small-company borrowers are in trouble, and companies with EBITDA of $25 million or less experience significantly higher default rates—15.8%—than large companies by 2025. Healthcare and consumer companies have high rates of automation. Fitch also notes that losses incurred by original lenders are limited, and most cases result in full or high percentage recoveries.
Notably, private mortgage rates have historically tended to be higher than those for commercial mortgages, a trend some observers attribute to customized, and sometimes distressed, lending policies. January's uptick was largely driven by “distressed” swaps and payment-in-kind (PIK) interest, according to Fitch.
AI concerns
Concerns are growing about PE funds exposed to software. Investors are concerned that AI will disrupt the software industry, leading to defaults among the sector's private equity loan portfolios. But many such funds are diversified, and even those that aren't may not be as vulnerable to AI disruption as investors fear. That's because the major language models that support AI require user interface interfaces to work, so software may be needed to facilitate the use of the technology.
“Using AI still requires a lot of effort to make it work in a certain environment,” Allen Lin, senior director of North American companies, technology, at Fitch Ratings, told the audience at a recent webinar held by the firm.
Of course, a lot depends on the type of application involved. As Fitch notes, companies that produce software that is deeply embedded in business technology systems, uses proprietary data, or operates in highly regulated industries such as healthcare and financial services can benefit from AI development. Conversely, those that produce software for less embedded applications, such as digital content creation or certain types of analytics and visualization tools, are more exposed to AI disruption.
Even if this AI bubble bursts, that risk is unlikely to disappear, said Lyle Margolis, a senior director in the Fitch corporate group, where he runs the private equity business, in an interview. Global Finance. “AI is here to stay and it will disrupt certain parts of the software market,” he says.
However, the risks may be overestimated. Whether measured by ratio, interest coverage, or EBITDA, “trends in the software sector have been somewhat positive,” he noted. The risk of refinancing the sector is not without risk. And data center construction presents one of the few “significant challenges” to private credit in the software sector, added Dafina Dunmore, Fitch's senior director of non-bank financial institutions in North America.
Another mitigating factor: Redemption risk, which could see a large cash outflow. However, it is largely limited to business development companies (BDCs), a more liquid, commercial-oriented type of private debt investment vehicle. Blue Owl, for example, recently banned the use of one of its BDCs and closed others. And the $33 billion Cliffwater Corporate Lending Fund, the largest U.S. private credit facility at the time saw redemption requests up 14%.
Although defaults increase in these portfolios, redemption risk is not an issue in most debt funds, because investors are locked in until maturity. In addition, pressure is focused on direct lending: corporate debt that finances operations and growth.
Hidden Dangers
Certainly, many such risks may be hidden, given the fragility of private credit. Blue Owl's exposure to software debt, among the highest in the industry, stretches twice as far as its public filings show, according to a recent Wall Street Journal analysis. The paper also found other PE firms whose credit ratings show software exposures that exceed what is publicly stated include Blackstone, Ares, and Apollo.
Investor concerns may exacerbate Blue Owl's bailout problems as its financial data center deals involve accounting practices that hide the risks involved. The main source of concern is likely Blue Owl's $27.3 billion financing of Meta's Hyperion data center in Louisiana.
However, S&P rates the bond backing the deal, called a Beignet, as a Meta bond, indicating it has default risk. Indeed, investors seem to like that cash-rich Meta behind Beignet. A bond was recently issued to fund a CoreWeave data center, which can be supported by a hyperscaler.
Still, some wonder if the risks are worth the trouble.
Quinn Emanuel warns that the vagaries of Meta's accounting management could lead to lawsuits between the parties over who bears the losses if AI fails to meet expectations and Meta chooses not to renew the lease. Blue Owl finances Oracle's data center in a similar way, but that bond trades at a discount to Meta, partly because Oracle doesn't support it and partly because the primary tenant is not financially stable OpenAI.
“When we rate data centers, to some extent we look at the credit quality of the top tenant,” said Victor Leung, vice president of project finance at ratings firm DBRS Morningstar.
This type of difficulty led Quinn Emanuel to warn in his March 13 memo that, “the creation of an AI data center—projected to require $5.2 billion in infrastructure investment by the end of the decade—has presented complex financial structures that pose significant litigation risks.”
Mark Koziel, CEO of the International Association of International Certified Professional Accountants and president-CEO of the American Institute of CPAs, says he will raise the issue of current accounting rules for these financial systems at an upcoming meeting with the Financial Accounting Standards Board. And last month, the US Treasury Department said it would meet with industry representatives and investors to discuss the potential risk of sovereign debt in the financial system.
So far, warnings of a private credit crunch seem overblown.
Asset-backed credit funds (ABF), which are based on the value of the borrower's assets and are the fastest growing sector in the market, are not immune to stress, due to their self-liquidating nature. In contrast to direct loans, the principal on equity-financed loans is repaid over the term of the loan. As a result, ABF funds do not face the same risk of repayment as direct lenders.
Direct lenders “don't have the benefit of that cash flow going toward paying off the loan,” notes Fitch's Margolies.
Apart from the acquisitions of First Brands and Jefferies, the ABF segment has yet to be fully evaluated. But there may be a test soon: Beignet is also supported by the property. Or kind of.
The loan principal is always different for each renewal, so it is not completely self-paying. As a result, DBRS Morningstar's Leung notes, “you face the risk that your property will lose its source of income.” Therefore, Meta guarantees that it will cover any losses faced by investors if it fails to renew the lease and the residual value of the facility falls below a certain limit.
That situation is not far-fetched, Quinn Emanuel warns, noting that it is expensive to convert an AI data center into general-purpose cloud computing or other uses: “If the demand for AI computing contracts, these centers may serve as idle assets with limited alternative use and depressed resale value.
