Two accounts of the same quarter appeared within days of each other. One reported that business development company results point to stabilisation, with managers cutting leverage and working through troubled positions. The other, based on regulatory filings from 44 BDCs, reported that portfolio values had moved further below cost across the first half of the year.
Both are accurate. The disagreement is grammatical rather than factual, and it is the most common way credit stories are misread.
First derivative, second derivative
"Deteriorating" is a statement about a level: the portfolio is worth less relative to what was paid for it than it was before.
"Stabilising" is a statement about a rate: it is getting worse more slowly than it was.
A portfolio can do both simultaneously and usually does, because credit cycles do not turn at a point. Most of the broad markdowns in this cycle landed in the first quarter; second-quarter losses at several prominent BDCs were concentrated in a small number of borrowers rather than spread across the book. That is genuine information — a narrowing of stress is different from a widening one, and it is what a stabilising rate looks like.
It is also entirely compatible with the level of impairment continuing to rise, which it did.
What the level is doing
Ares Capital is the largest listed BDC and therefore the most-watched single data point. Its non-accrual loans rose fifteen percent quarter-on-quarter, to $708m.
Non-accrual is the point at which a lender stops booking interest it does not expect to receive. It is a relatively honest category, because putting a loan on non-accrual costs the manager reported income immediately. A rising non-accrual balance at the largest player, while the sector narrative is stabilisation, is the level and the rate pointing in opposite directions in a single line item.
The number that does the concealing
Here is the figure to actually watch, and it appears in neither headline.
PIK income — payment in kind, where a borrower is permitted to add interest to the principal balance rather than pay it in cash — is approaching ten percent of income, a level widely treated as the threshold where a portfolio stops being a lending book and starts being an accrual exercise. In the first quarter, 32 borrowers accounting for $1.3bn of principal switched some or all of their interest to PIK.
Consider what a PIK amendment does to the statistics. A borrower who cannot pay cash interest is, economically, a borrower in difficulty. Amend the loan to PIK and that borrower is now current. It does not appear in non-accruals, because interest is accruing. It does not appear in defaults, because there has been no default. The reported yield on the position may even rise, since PIK rates are typically higher.
The exposure has not improved in any respect. It has grown, because unpaid interest is now principal, and the eventual loss will be measured against a larger number. What has changed is which statistic the borrower shows up in — and it has moved from the one everybody watches to the one almost nobody aggregates.
This is the same mechanism this desk described when a mark is an opinion until somebody has to sell. A PIK amendment is a mark expressed as a contract change: the manager's view that this borrower will eventually pay, converted into documentation, and thereby removed from the parts of the report where views are questioned.
Why software
The stress is reported as concentrated among software borrowers, which is worth a sentence because it is not obvious.
Software was the ideal private-credit collateral of the last cycle: recurring revenue, high gross margins, low capital intensity, and a valuation multiple that made leverage look conservative against enterprise value. The loans were written against revenue multiples rather than assets, on the reasonable premise that subscription revenue is sticky.
It is sticky. It is also repriceable by the customer at renewal, and this desk has been documenting the repricing all year — buyers discovering they own too many tools, free tiers disappearing, seat counts revisited. A borrower whose revenue is flat rather than growing breaches a covenant written for growth, and a lender who does not want the business converts the loan to PIK.
What would resolve it
The cycle turns when the level stops rising, not when the rate slows, and the level is visible in two series.
Watch non-accruals as a percentage of cost across the sector, not at one manager, and watch PIK income as a share of total investment income. If non-accruals flatten while PIK falls, the book is genuinely healing. If non-accruals flatten while PIK keeps climbing toward and past ten percent, the improvement is documentary: the difficult borrowers have been moved into a category that does not count them.
The distinction becomes academic at the maturity date, which is the wall this desk has written is taller than it looks. A PIK loan does not forgive anything. It postpones the conversation to a year in which the balance is bigger.
The review of regulatory filings from 44 US business development companies showing portfolio values further below reported cost in the first half of 2026, the concentration of stress among software borrowers and the observation that most broad markdowns fell in the first quarter are as reported by Reuters. The characterisation of second-quarter BDC results as pointing to stabilisation, and the actions attributed to Ares Management, Blue Owl Capital and BlackRock, are as reported by Private Equity Wire. The 15 percent quarter-on-quarter rise in Ares Capital non-accruals to $708m, the approach of PIK income toward a 10 percent threshold and the identification of 32 borrowers accounting for $1.3bn of principal switching interest to PIK in the first quarter are as reported by Reuters and Octus. All figures date from coverage published between June and early September 2026. The analysis is our own.
Topics marketsprivate creditearnings
