Capital · Research note
Private Credit's First Real Test: What the Default Rate Is Not Telling You
The asset class reported defaults below 2% while roughly 6% of loans had stopped paying cash interest. The gap between those two numbers is the entire risk.
Private credit grew past $2 trillion on a specific promise: that direct lending produced equity-like yields with almost no realised losses. The promise was tested for the first time this year, and the test is not going well. What makes 2026 interesting is not that losses appeared. It is that they appeared in a place the headline statistics do not measure, so the asset class could deteriorate for several quarters while every published number looked reassuring.
This piece argues that the reported default rate has been the wrong number to watch, because the structures that made private credit attractive to borrowers also made it possible for a loan to stop performing without ever being recorded as impaired. The gap between the reported rate and the economic one is where the risk sits, and the redemption gates imposed in the first quarter are the first visible consequence.
Two default rates
Figure 1 puts the two measures side by side. The headline default rate stayed below 2% for several years, which is the figure that appeared in fundraising material and in most commentary. Fitch put the US private credit default rate at a record 6.0% in April 2026. Lincoln International’s shadow measure, which counts borrowers deferring interest mid-loan, implies distress close to 6% against that same sub-2% headline.
Both numbers are correct: they measure different things, and the difference is definitional rather than empirical, since a loan counts as defaulted only when it breaches a term and covenant-lite documentation removed most of the terms it could breach.
How a loan stops paying without defaulting
Payment in kind is the mechanism: under a PIK arrangement the borrower does not pay cash interest, and the interest is instead added to the principal and settled at maturity. Agreed at origination for a growth company with lumpy cash flow, this is a reasonable structure; applied mid-loan to a borrower that has run out of cash, it is something else entirely.
The industry distinguishes the two as good PIK and bad PIK, and the second is the one that matters.
Figure 2 tracks it. Bad PIK reached 6.4% of private credit loans by the fourth quarter of 2025, roughly triple the 2021 level. The rise coincides exactly with the rate cycle, which is what you would expect: a borrower underwritten at a 6% coupon and repricing to 10% runs short of cash, and the lender faces a choice between recording a default and amending the loan. Amending it preserves the mark. Hence the incentive runs one way, and it runs that way for the lender as much as for the borrower.
Two further features compound this: covenant-lite documentation removes the trigger that would have forced recognition earlier, while NAV-based lending and fund-level leverage add borrowing on top of a portfolio whose marks are themselves determined by the manager. The Financial Stability Board’s May 2026 report on private credit vulnerabilities identifies all three together, and the common thread is that each one delays the moment a problem becomes visible rather than reducing the problem.
| Feature | What it does for the borrower | What it does to the reported numbers |
|---|---|---|
| PIK toggle | Converts cash interest into accrued principal | Loan continues to perform; no default recorded |
| Covenant-lite terms | Removes the maintenance test | Removes the trigger that forces early recognition |
| Liability management exercise | Restructures outside a formal default | Counted separately, or not at all |
| NAV lending | Raises cash against portfolio marks | Adds leverage above assets valued by the manager |
| Fund-level leverage | Amplifies returns on the same loan book | Concentrates losses in the equity tranche |
Structure follows the Financial Stability Board, “Report on Vulnerabilities in Private Credit,” 6 May 2026.
Figure 3 locates the problem. A loan in the bottom-left cell has stopped paying cash and has not been recorded as anything, so it appears in neither the default statistics nor the mark; losses do not disappear when they are unrecorded. They accumulate until something forces recognition. In an open-ended fund that forcing event is usually a redemption request.
The redemptions arrived first
Figure 4 shows the first quarter producing the sequence these structures were designed to prevent. Investors requested redemption of 40.7% of Blue Owl’s Technology Income Corp and 21.9% of its Credit Income Corp, together roughly $5.4 billion. Ares faced requests for 11.6% of its Strategic Income Fund, and each of these funds permits 5% a quarter. Across the semi-liquid private credit complex, requests reached $13.9 billion, an increase of 217% on the prior quarter.
By early March, Blue Owl had moved to a liquidation plan returning about 30% of capital over a 45-day window.
Note what the gate actually does: it works exactly as written, and working as written is the problem, because an investor who wanted out in January was told they could have 5%, which is a strong reason for every other investor to file a request too. A queue that is capped rewards the person who joins it earliest, so capping it generates the demand it was designed to contain. This is the reflexivity that makes bank runs self-fulfilling. It has been imported into a structure marketed as immune to it.
The illiquidity was supposed to be the protection: semi-liquid vehicles hold assets that cannot be sold quickly, and they manage that by limiting withdrawals. Therefore the gate is not a failure of the design; it is the design. What the design did not anticipate is that a gate is also a signal, and the signal travels faster than the assets can be sold.
Why this is a bank story, not only a fund story
Private credit grew because banks retreated: post-crisis capital rules made mid-market corporate lending expensive to hold on a bank balance sheet, and the loans migrated to vehicles with no capital requirement, no deposit insurance and no lender of last resort.
That migration was widely described as making the system safer, on the argument that losses now fall on long-dated investors rather than on leveraged deposit-takers. The argument holds only if the investors are genuinely long-dated; a fund promising quarterly liquidity against loans with a five year life is performing maturity transformation, which is precisely the activity banks are regulated for doing.
Furthermore, the banks did not leave entirely. They lend to the funds through subscription lines and NAV facilities, and insurers hold private credit in general accounts against long-dated liabilities. Hence the exposure returned to regulated balance sheets through a different door, and it returned without the disclosure that direct lending would have carried.
Where this argument is weak
The most important limitation is that none of this has yet produced a realised loss cycle. Requests are not redemptions, gates are not defaults, and a fund returning 30% of capital over a 45-day window is doing something orderly rather than something catastrophic. The evidence assembled here describes a system under stress; it does not describe one that has broken.
Furthermore, the bad PIK series carries real measurement problems. The distinction between good and bad PIK depends on judgement about why interest was deferred, and that judgement is made by parties with an interest in the answer. A rise from roughly 2% to 6.4% is large enough to survive some measurement error, though the precision implied by the decimal is not warranted.
Finally, the comparison to bank runs has a limit worth stating. Depositors can withdraw at par on demand, and these investors cannot and never could; the reflexivity is real, the speed is not comparable, and a quarterly gate genuinely does buy a manager time that a bank facing a deposit run does not have.
Conclusion
Private credit’s problem in 2026 is not the default rate. It is that the asset class built structures which allow a loan to deteriorate without any number changing, and then sold quarterly liquidity against those loans to investors who read the unchanged numbers as evidence of quality.
Roughly 6% of loans are deferring cash interest against a headline default rate under 2%, and redemption requests rose 217% in a quarter before being met with gates that worked precisely as documented. Neither fact is a crisis on its own. Together they describe an asset class discovering that the reported condition of its book and the actual condition of its book are two different things, and that the second one is what investors were trying to redeem against.
The useful question for the next four quarters is therefore not whether defaults rise; it is whether the reported rate converges upward toward the shadow rate, or whether the shadow rate keeps absorbing the difference on its own.