Sponsorship: Private Debt News reaches institutional investors, credit professionals, and LP decision-makers. Early sponsor rates available. Contact us here or reply directly to this email.
Direct lending volume came in at $38.9bn for the third quarter, the lowest since the second quarter of 2023. September alone was $10.4bn, down 66% from a year earlier and less than half the recent monthly run rate of about $23bn. The money is still leaving faster than it arrives. Year to date, $22.9bn of private credit has been refinanced into bank loans against $9.2bn moving the other way.
That inversion is the unusual part. Direct lenders gained share in every prior volatility window because banks retreated and they didn’t. This time the banks are the ones taking share, because the private funds are busy nursing troubled books, managing $15bn of redemption queues, and finding they have less to deploy. Churchill’s Randy Schwimmer called it “the period of zero gravity.” Rates are up, exits are slow, deal flow is down, and the capital that remains has nowhere obvious to go.
Where it’s going is downmarket. Loans to companies under $20mm of EBITDA more than doubled in the second quarter and lower- and core-middle-market deals made up ~2/3rds of third-quarter volume. EOS Corrugated, a Sagard-owned maker of parts for cardboard manufacturers with about €25mm of earnings, is drawing a 5.5x package from Barings and CVC Credit at as tight as E+475. Borrowers under $20mm of EBITDA now pay roughly what borrowers over $100mm pay. “We’re definitely seeing larger managers looking at smaller deals to put more money to work,” said HarbourVest’s Greg Ciesielski.
The industry’s trade body answered with a report arguing sentiment has become “divorced from the market-wide reality,” citing 17% AUM growth to $4.1 trillion and a decline in loan amendments to 10.6% from 11.7%. The same report found ~1/6th of surveyed funds with non-accruals above 5%, and Man Group’s Kevin Marchetti put roughly 70% of lender takeovers in the 2021 and 2022 vintages.
Apollo described its $2.6bn Yankees financing to investors as “AI-resilient leisure” after cutting $750mm of software. PIMCO said software loans still trade 8 points behind non-software peers with $62bn maturing in 2028 and 2029. And Cox Capital, having failed twice to buy gated BDC shares at a discount, is now offering to lend against them at a fixed 10%.
Key Market Themes
1. The Quarter the Banks Won
Third-quarter direct lending volume of $38.9bn is the weakest print since mid-2023. September’s $10.4bn was a 66% year-over-year decline. The refinancing flow stands at $22.9bn out of private credit into syndicated loans versus $9.2bn in, a ratio of roughly 2.5 to 1 and the opposite of what direct lenders have delivered in every prior stress period. Large-cap LBOs remain scarce, and a number of managers are tending troubled funds with less capital to put out.
Beyond direct lending, Blackstone and Apollo are still writing multi-billion checks for investment-grade AI infrastructure borrowers, which is where the large-ticket activity has migrated.
Share loss during volatility is new
The pitch to LPs for a decade was that private credit was countercyclical. Banks pull back in a shock, direct lenders step in, and spreads widen in their favor. The 2022 rate shock proved it. This cycle is doing the reverse, because the shock hit the lenders’ own funding first. A manager with a 14% redemption queue, a stressed software book, and a revolver up for renewal doesn’t have the balance sheet to underwrite a $2bn unitranche, so the bank does it at S+275 and the borrower saves $29mm a year. The volume collapse isn’t a demand problem. It’s the supply side choosing liquidity preservation over deployment, which is rational and also the first time the countercyclical story has failed.
2. Small Ball at Large-Cap Pricing
Lending to sub-$20mm EBITDA companies more than doubled in the second quarter as larger managers moved down to find deployment. EOS Corrugated at 5.5x and around E+475 is the template, a leverage level and spread that would have been unusual for a €25mm earnings business in any prior year. KBRA DLD’s rolling spread data shows the sub-$20mm, $20-50mm, $50-100mm, and $100mm-plus cohorts converging into a band of roughly 490 to 530 bps. Recent examples include Ares backing Metalock, Golub on Atoz Services, BlackRock refinancing iLC, and MidCap lines of $60mm and $80mm to Ceribell and RoC Skincare.
The size premium disappeared
Smaller borrowers used to pay 100 to 150 bps more than large ones for thinner cushions, shorter track records, and less diversification. That premium is now zero, because the capital that can’t find $100mm EBITDA deals is bidding for €25mm ones at the same price. EOS at 5.5x with E+475 is a large-cap structure on a lower-middle-market company. Whether that turns into a problem depends on whether these businesses can absorb a setback at 5.5x, and the historical answer is that they can’t as easily. The lower middle market was the segment with the best covenant protection and the widest spreads two years ago. It’s now getting the same loose terms the upper market got in 2021, from managers who are there because they have to be somewhere.
3. The Trade Body Says Everything Is Fine, Then Lists the Problems
The Alternative Credit Council’s report argues public sentiment has detached from reality. AUM grew 17% to $4.1 trillion. Significant amendments fell to 10.6% of loans from 11.7%. Stress is “concentrated rather than indicative of broader market deterioration.” The report also notes lenders now require more covenants, wider spreads, and in some cases won’t extend maturities for AI-vulnerable borrowers.
The concentration argument comes with numbers. About ~1/6th of surveyed funds carry non-accruals above 5%. Marchetti at Man Group puts ~70% of lender takeovers in 2021 and 2022 deals “when purchase multiples were unsustainable and rates were low.” KKR’s Michael Small argued the AI threat is being misread. “Some software issuers will face stress, but professional services may be more exposed.”
Concentrated in the half that was underwritten in the boom
The report’s defense is that stress sits in a specific vintage and a specific sector rather than across the book. That’s true and it’s not comforting. The 2021 and 2022 vintages were the largest the industry ever originated, and a sixth of funds with non-accruals over 5% is a lot of funds. Amendments falling to 10.6% is good news only if it reflects fewer stressed borrowers rather than lenders who’ve stopped granting them, and the report’s own description of tighter terms and refused extensions suggests the latter. Small’s point about professional services deserves more attention than it’ll get. A staffing or consulting business with 6x leverage is more exposed to agentic AI than a vertical software company with a sticky install base, and almost no one is marking it that way yet.
4. Apollo Rebrands the Yankees
Apollo Debt Solutions told investors its $2.6bn financing to the Yankees’ owners represents “AI-resilient leisure,” an area it sees as protected from generative AI. The fund cut software by about $750mm in the second half of last year, bringing it to 10.5% of the portfolio, roughly half the share at peers with NAV above $4bn. Apollo launched a $5bn sports capital initiative last year and the Yankees deal is its largest US sports investment.
PIMCO’s latest note puts the software problem in prices. Loans to software companies recovered about a third of their spread widening over the summer, then gave part of it back after early September. The average bid for software loans sits about 8 points below non-software peers, versus 10 in June. About $62bn matures in 2028 and 2029. Amend-and-extend costs range from 25 to 150 bps depending on perceived AI resilience.
An 8-point gap and a $62bn wall
Apollo’s framing is marketing, but the underlying move was right and early. Cutting $750mm of software in the second half of 2025, before the February selloff, is why the fund can tell investors it’s at half the peer exposure. The PIMCO numbers describe what everyone who didn’t cut is facing. Eight points of discount on software paper is a markdown the quarterly NAVs haven’t fully taken, and $62bn of maturities in 2028 and 2029 is a refinancing wall that will price each borrower by its AI story. A 150 bps step-up on an amend-and-extend for a borrower already at 1.5x coverage is the mechanism by which an 8-point gap becomes a default. The Yankees don’t have that problem. Most of the 2021 software book does.
5. Cox Pivots From Buying to Lending
After two tender offers that drew about $5mm of combined interest, Cox Capital is launching a pilot to lend against gated BDC shares. Eligible investors with $1mm or more in a fund over $1bn can borrow up to 65% of NAV at a fixed rate of roughly 10%, with a two-to-four-year target term and repayment expected mostly from quarterly redemption proceeds. The firm is using a nearly $200mm vehicle, separate from its $150mm secondary-purchase fund, and plans to extend the program to wirehouses and advisers if demand materializes. “There’s really no reality where redemptions are exactly equal to or less than liquidity available,” said John Cox.
Lending against the queue
Cox learned that retail investors won’t sell at a 26% discount, or a 12.5% one, when the gate pays out 75% within a few quarters. So he’s offering them the other thing they want, which is cash now without crystallizing the loss. A 65% advance at 10% on a position the fund pays out at par over two years is a decent trade for Cox if the NAV holds. If the NAV takes the markdowns the BDC discount paper is predicting, the collateral shrinks and the borrower is paying 10% on a shrinking asset. The more interesting part is the structure. Cox is building a fund-finance product for retail investors, which means the same leverage-on-illiquid-assets dynamic that banks are retreating from in Australia is being offered to individuals in the US at a higher rate.
6. Refinancings Get Done, With Existing Lenders
Sound Inpatient Physicians is close to a $1bn refinancing with existing lenders including Monarch and PGIM. Prospect Capital told bondholders it will redeem $263mm of notes due next month while it continues to shop a refinancing. Pivot Energy closed a $500mm bilateral facility with Nuveen. Blackstone and AlbaCore lent £140mm to Network Plus. Partners Group provided €300mm for BU Bregal’s acquisition of MDT Technologies.
The lender of record is the lender of last resort
Sound Inpatient getting $1bn from Monarch and PGIM rather than new money is the pattern to watch. When a borrower refinances with the lenders it already has, it usually means the market didn’t offer a better bid and the incumbents preferred extending to taking a workout. That’s fine when the credit is sound and it’s how problems get rolled when it isn’t. Prospect redeeming $263mm while still hunting for a refinancing shows the other version. The fund is paying down with cash it would rather keep because the new deal hasn’t come together, and every month without one makes the next bond more expensive.
Deals of Note
EOS Corrugated - Barings and CVC Credit Partners offering ~5.5x at as tight as E+475 for Sagard’s sale of the €25mm EBITDA business
Sound Inpatient Physicians - ~$1bn refinancing with existing lenders including Monarch and PGIM
New York Yankees - Apollo’s $2.6bn financing to the owners, billed as “AI-resilient leisure”
Pivot Energy - $500mm bilateral corporate facility with Nuveen Energy Infrastructure Capital
Network Plus - Blackstone and AlbaCore lent £140mm to the infrastructure services provider
MDT Technologies - Partners Group €300mm facility backing BU Bregal’s majority stake purchase
Prospect Capital - Redeeming $263mm of notes due next month while shopping a refinancing
RoC Skincare and Ceribell - MidCap Financial senior facilities of $80mm and $60mm
Cox Capital lending pilot - Up to 65% LTV at ~10% fixed on gated BDC positions of $1mm+, from a ~$200mm vehicle
The Reality Check
A $38.9bn quarter with banks refinancing private credit at 2.5 to 1 is the number that undercuts every “concentrated stress” argument this week. The asset class was sold as countercyclical and it is currently losing share in the first real stress test it has faced. That’s not because demand collapsed. It’s because the lenders are preserving liquidity for redemptions and revolver renewals, and a manager who can’t deploy can’t earn the fees that justify the structure.
Small ball at large-cap pricing is how deployment pressure expresses itself. A €25mm EBITDA parts maker at 5.5x and E+475 is a structure that belongs to a different borrower. The lower middle market had the best terms in the industry two years ago because the big managers didn’t bother with it. They bother now, and the size premium has gone to zero in a year. Whoever holds those loans in 2028 will learn what a thin cushion at 5.5x does in a slowdown.
The trade body’s defense and PIMCO’s numbers can both be true. Stress is concentrated in the 2021 and 2022 vintages, those vintages are the biggest the industry ever wrote, and the software piece of them trades 8 points behind with $62bn due in 2028 and 2029. Apollo cut early and can call the Yankees AI-resilient with a straight face. The managers who didn’t cut are facing 150 bps amend-and-extends on borrowers who can’t cover the old coupon.
Cox lending at 10% against gated shares is the week’s most honest product. The retail investors won’t sell at a discount, the funds won’t pay out faster than 5%, and the banks are tightening the lines behind the funds. Someone was going to monetize that gap. A 65% advance on a NAV the public market thinks is overstated is a reasonable bet for the lender and a reminder that the leverage being removed from the top of the structure is being reintroduced at the bottom.