What it is
A private credit fund raises money from insurers, pensions and wealthy individuals, lends it directly to companies too small or too leveraged for the investment-grade , and holds those loans until they mature or blow up. The borrower is usually a mid-sized firm owned by a . The loan is usually floating rate, senior, and written as a single instrument by a single lender rather than syndicated to fifty.
The load-bearing sentence belongs to the Federal Reserve, in its February 2024 note on the asset class: "Private credit loans are illiquid due to the lack of a secondary market. There is limited market discovery." Everything difficult about private credit descends from that. No trading means no price. No price means a valuation. A valuation means someone's judgement, and that someone is being paid a fee calculated on the number they produce.
How big it is depends on who is counting and what they are counting. PitchBook, as cited by Franklin Templeton, put global private credit AUM at $1.97tn and the US at about $1.3tn as of 30 June 2025. S&P Global, using Preqin, says $2.3tn for 2025. The Alternative Credit Council says $3.5tn as of end-2024. These are not rounding differences; the widest and narrowest are nearly a factor of two apart, and the gap is definitional — whether you include asset-backed finance, real estate debt and consumer lending or only direct corporate loans. There is no Fed or IMF figure that settles it. For an asset class the IMF's April 2024 Global Financial Stability Report devoted a chapter to as "a rapidly growing asset class" posing potential financial stability risks, nobody official knows how large it is to the nearest trillion.
The mark is a choice
In a public market, two investors who disagree about what a bond is worth produce a spread, and both of them get to it. In private credit there is no mechanism that forces the two views to meet. This is not a subtlety. It is the central mechanical fact of the asset class, and 2026 has produced the clearest documented demonstration of it we have seen.
A 112-page filing in the First Brands bankruptcy, written by restructuring chief executive Charles Moore of Alvarez & Marsal, describes Katsumi Global — a Japanese trade-finance firm — trying to sell part of its First Brands factored-invoice exposure to Apollo. Apollo declined, Moore writes, "after correctly identifying the numerous red flags discussed herein, even without the benefit of the specific disclosure known to Katsumi" (FT Alphaville). The filing quotes an internal Apollo presentation from April 2024, seventeen months before the Chapter 11: "the cash flow of the business has never matched the Company's claims about its margins"; a claimed 20 per cent operating margin at a company acquiring sub-10 per cent EBITDA businesses; and a chief executive who "has controlled dozens of entities over the years, many of which have ended up in litigation or liquidation (or both). Simple Google searches reveal almost nothing about him." Apollo shorted the debt instead. Katsumi carried the exposure and tried to sell it. We wrote this up on 19 August: one institution reached a conclusion from public information and a search engine, the other did not, and nothing in the structure of the market obliged either mark to .
Collateral has the same problem one layer down. The SEC has sued three former Tricolor Holdings executives — founder Daniel Chu, former CFO Jerome Kollar and former senior finance director Ameryn Seibold — alleging that from 2020 until the September 2025 bankruptcy the subprime auto lender raised more than $1.9bn on faked documents and by pledging the same vehicles as collateral for more than one loan (Bloomberg). Kollar and Seibold have pleaded guilty and are cooperating; Chu's lawyer calls the case a "rehash" and says many allegations are inaccurate. The mechanism, as we set out on 19 August, is that verification here is contractual rather than physical. Nobody walks the lot. A servicer represents, an auditor samples, a trustee holds paper. A false description can run for five years, and did.
None of this requires fraud to matter. Look at an honest book. At Blue Owl Capital Corporation, in the second quarter of 2026 were 2.8% at cost and 0.8% at fair value — the fund has written the bad loans down and is telling you, correctly, that most of the loss is already taken. Whether the remaining marks hold is the entire question, and there is no outside price to check them against. income, where a borrower settles interest by adding to principal rather than paying cash, was around 10.7% of the fund's investment income. PIK is how a stressed loan avoids becoming a non-accrual.
When discretion is abused rather than merely exercised, the remedy arrives years late and costs less than the fees. The SEC's order against Fifth Street Management found the adviser improperly allocated $1,327,405 of rent and other overhead to its BDC clients; the associated valuation findings were that inflated marks allowed the funds to trade at higher prices than they otherwise might.
Even the count of defaults is a methodology choice. Proskauer's Private Credit Default Index reported 2.51% for Q2 2026 and 2.73% for Q1 2026, up from 1.84% in Q3 2025 — a measure that counts payment and financial-covenant defaults only. A broader Fitch methodology put roughly the same Q1 2026 period near 5.8%. Same market, same quarter, twice the number, and the difference is mostly whether you count the that let a borrower avoid formally defaulting.
The numbers right now
The hard data has turned. The prices have not.
Median non-accruals at the twenty largest listed rose from 2.0% of cost at the end of March 2026 to 2.8% at the end of June — the highest in roughly a decade, on an FT analysis of Solve data (FT). PitchBook data put 3.95% of loans at the ten biggest as no longer paying interest. FS KKR reported 7.1% of its book troubled. Fitch says private credit defaults hit a record in July. And the books are shrinking: PitchBook LCD found repayments and loan sales outran new commitments at vehicles run by KKR, Blue Owl and Apollo's MidCap Financial. A shrinking loan book is a lender choosing not to lend.
Non-accruals matter because they are close to the only number here that is not an opinion. A fund can mark a loan wherever its process supports. It cannot book interest it is not receiving. An 80 basis point move in the median in a single quarter, across the middle of the distribution rather than at one unlucky fund, is fast.
On the liability side, the queue. Investors in non-traded BDCs asked for 12.4% of net asset value back in the second quarter of 2026, a record, against quarterly repurchase caps generally set around 5% of NAV. Only 38% of requests were met, leaving roughly $9.6bn queued; $12.7bn was withdrawn across the first half, $5.9bn of it in Q2. These figures come from a secondary compilation rather than a filing we have seen, so treat the decimal places lightly — but as we wrote on 19 August, the structure they describe is not in doubt. Once requests exceed the cap the fund pro-rates, everyone gets a fraction of what they asked for, and the rational response is to ask for more next quarter in order to get the same amount out. The queue feeds itself. Nobody has to default and no mark has to move.
The buyer of last resort also just declined a job. 's Sophos needs to refinance or extend more than $2bn of loans as soon as next month. It spent months trying to line up private credit backing; the effort faltered, and the company went back to its existing leveraged-loan holders, potentially offering a higher coupon, amortisation payments and a tighter (Bloomberg). Thoma Bravo has told lenders it will not inject fresh capital. It is the third Thoma Bravo credit in three months to need this treatment: creditors took control of Medallia in June, and a $5bn Proofpoint refinancing had to be sweetened in July. The direction of travel is backwards. Paper that was supposed to leave the public market forever is going back to it, to the one place it gets marked every day.
Now the prices. The listed BDCs trade within about 2-5% of their recent highs — Ares Capital 2.5% off, Blackstone Secured Lending 3.0% off, Golub's own BDC 2.2% off, this last while David Golub tells investors "We're in a credit cycle. Others denied it for a while. I don't think there's a lot of denial any more." over the same week were unchanged at 273bp. The only things doing any selling are the managers' own shares: over five days Blue Owl down 9.1%, Apollo 8.1%, KKR 6.2%, Ares 5.6%, Blackstone 4.7%. Equity holders of the firms that originate this paper are marking down the franchise while the credit market prices no deterioration at all. We are not going to resolve that for you. We are going to keep pointing at it.
Who funds it
Two marginal buyers replaced the banks, and both have features the banks did not.
The first is insurance float. US life insurers held roughly $849bn of private credit in 2024, about 14% of balance sheets, on a Chicago Fed figure relayed secondhand by a venture blog — treat it as indicative rather than precise. Better sourced is the direction: privately placed bonds reached 48.4% of total US life-industry bonds at year-end 2025, and a Marsh survey reported by Reuters in July 2026 found 57% of insurers planning to increase private credit exposure over the next 12-24 months, rising to 73% of life insurers and 81% of firms with more than $25bn of assets.
The model pairs long, sticky annuity liabilities with illiquid loans originated by an affiliated manager. It works precisely because annuity money does not run. The weak point is who sets the price, and the Walter case is what that looks like when it goes wrong. After receiving grand jury subpoenas from the Southern District of New York in February 2026, insurers controlled by 's TWG Group disclosed more than $20bn of loans to related parties whose connections had not been disclosed. Delaware Life agreed to swap $6.5bn of TWG-affiliated assets for unrelated investments; after the swap, it and a second Walter insurer still held more than $10bn of affiliated investments. S&P put Delaware Life on negative outlook after it restated its annual statements and reclassified a large volume of private credit as related-party. Meanwhile Walter had spent the summer negotiating a multibillion-dollar loan from Apollo secured on his Los Angeles Lakers stake, talks overtaken by a $12.5bn sale to Bob Iger and Josh Kushner (FT). We covered the sequence on 18 and 19 August, including a correction: we had initially read the $6.5bn swap as risk coming off, and it covers well under half of what is there.
The second buyer is retail, through non-traded BDCs and . Non-traded BDCs typically offer discretionary quarterly repurchases of up to about 5% of NAV, which boards may reduce or suspend. Interval funds must, under Rule 23c-3, offer quarterly repurchases of between 5% and 25% of shares at NAV. Both pro-rate when demand exceeds the cap. That is the mechanism currently queuing.
A third, newer tell: the funds are borrowing in the public bond market themselves. August 2026 set a US investment-grade issuance record at $145.2bn, beating August 2020's $136bn, the third consecutive record month; year-to-date supply of $1.46tn ran 8.5% ahead of 2020's pace. Among one Monday's twelve deals were two from private credit vehicles run by Blackstone and Blue Owl. We wrote about that on 18 August. A fund with public bondholders has covenants and leverage ratios calculated off the value of loans nobody marks daily.
How we got here
The asset class is largely a regulatory artifact, twice over.
Congress created business development companies in 1980, through the Small Business Investment Incentive Act, which amended the Investment Company Act of 1940 to give public investors a way to fund the long-term growth of private US businesses (Ares Capital). The bargain: distribute at least 90% of taxable income and the fund is not taxed at entity level, so the dividend yield is the product. The leverage limit was an asset coverage ratio of 200%, roughly 1.0x debt to equity. The Small Business Credit Availability Act of 2018 let BDCs elect 150% coverage instead — roughly 2.0x — with either shareholder approval or a required majority of independent directors plus a one-year delay.
The supply side came from the banks leaving. After 2008, Dodd-Frank and Basel III capital requirements plus the 2013 interagency leveraged lending guidance made middle-market leveraged loans expensive to hold and awkward to originate, and pushed banks toward distributing rather than keeping them. Direct lenders took the share, with the — one lender, one instrument, several risk layers combined — as the product that made execution simpler than a syndication.
The precedents for how non-bank lenders end are on the shelf, though the packet gives us fewer numbers than we would like. CIT Group, whose funding was wholesale rather than deposit-based, filed Chapter 11 on 1 November 2009; the $2.3bn of TARP preferred stock it had received in December 2008 was expected to be wiped out with the rest of the equity. It restructured, survived, and was eventually absorbed by First Citizens in 2022. American Capital was sold to Ares after the 2015-16 oil crash; we do not have its loss or non-accrual figures and are not going to guess at them. And Fifth Street, above, is the precedent for what happens when a manager's discretion over its own marks is tested by a regulator.
Why it matters to this crash
This is the corner of the system where the cycle's credit losses are accruing with the fewest observable prices, at a scale that now rivals the public markets it replaced. That is the whole reason it is a crash point rather than a sector.
It also refuses to stay in its lane. When the SEC's enforcement director David Woodcock described a collapsed subprime auto securitisation by saying the defendants "violated the integrity of our private credit markets," the regulator collapsed the boundary between corporate private credit and in a single sentence. First Brands ran the same trick as Tricolor — collateral pledged more than once, invisible because verification was documentary. The insurance leg puts annuity savers behind marks that an affiliate may have set. And the newest capital is arriving with the post-2008 safeguards read out of it: the SEC's staff agreed last month with Latham & Watkins that some data centre debt is not an asset-backed security and therefore falls outside Dodd-Frank risk retention (CNBC), days before Nvidia signed memoranda with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to build "compute financing platforms" aimed at mobilising more than $500bn. It is a staff opinion rather than a rule, and it is not yet clear the platforms will securitise anything at all. It is still the skin-in-the-game rule being read narrowly at the exact moment the largest financing wave in a decade is being structured.
The official sector has been flagging the asset class since the IMF's 2024 chapter, and the FSB has since published a report on vulnerabilities in private credit focused on bank interlinkages, lenders' credit exposures and the data gaps. The industry's counter-case circulates too: a Managed Funds Association research note headlines a 1.6% default rate from that same February 2024 Fed note as evidence of no material financial stability risk. That is a lobby group summarising a research paper, not 's conclusion, and it is a 2024 number being used to describe a 2026 book.
We have had private credit near the top of our risk list for consecutive cycles while the vehicles that hold the loans traded within a few per cent of their highs. That gap is why this entry exists.
What would make this dangerous
The queue turning into a run. Pro-ration is self-reinforcing: an investor who asks for 5% and receives 2% learns to ask for 12% next quarter. Non-traded BDC repurchase programmes are discretionary and can be reduced or suspended by the board, which converts a liquidity problem into a confidence problem overnight. Queues stretching four to eight quarters end in one of two places — confidence returning, or forced sales into a market with no bids and no reference price. Watch the Q3 and Q4 2026 fulfilment percentages, not the headline requests.
The insurance leg going first. Policyholder liabilities cannot be pro-rated. If a large annuity writer hits a downgrade spiral or a regulator forces an unwind, it has to sell assets that nobody outside the firm has ever priced. The Walter probe is the live test: $10bn-plus of affiliated investments still sitting on two insurers' balance sheets after a $6.5bn swap, with prosecutors inside.
Leverage rebuilding the bank connection the model was supposed to remove. The statutory ceiling is now 2.0x debt to equity for BDCs that elect it, the funds are issuing public investment-grade bonds, and the Fed has documented a further channel in its note on bank credit lines to private credit funds. An unlevered asset class funded by locked-up money does not transmit. This one increasingly is not that.
The refinancing wall pushing paper back into daylight. Sophos is the template: if private credit keeps declining sponsor refinancings, borrowers return to the , get a daily mark, and every private holder of a comparable credit acquires an inconvenient comparable. One Sophos is repricing. Twenty is a repricing event.
And the masks coming off. PIK income near 10.7% of investment income at one of the largest funds, a sector figure loosely put around 8% by commentary we cannot source to primary data, and a default rate that doubles depending on whether liability-management exercises count. Each of these makes the reported loss rate lower than the true one by an amount that nobody outside the funds can measure. That is not a prediction of disaster. It is a statement that if disaster arrives, the first accurate number will arrive after it.