Newsletter · 2026-07-29 · Ashutosh Agarwal

Golub Says Private Credit Defaults Are Running at Twice Normal - The Private Credit Boom (and Cracks) - Week of July 29, 2026

The Private Credit Boom (and Cracks) newsletter for the week of July 22–29, 2026. Golub Capital's co-CEO put a sourced number on the stress (Fitch's private credit default index running at a little more than twice normal), US direct lending issuance fell about 40% in a quarter as the syndicated loan market took large deals back, and federal grand jury subpoenas landed at two life insurers over how private credit holdings were disclosed.

The Private Credit Boom (and Cracks)

Week of July 22–29, 2026: Golub Says Private Credit Defaults Are Running at Twice Normal

TL;DR

An insider finally said the quiet part. David Golub, co-CEO of Golub Capital, one of the biggest direct lenders on earth, went on a podcast this week and said the Fitch index of private credit defaults is "running at 2x, a little more than 2x normal." Last week the warnings came from strategists and short-sellers. This week they came from inside the house.
The money going out the door has slowed hard. US direct lending issuance was $44.76 billion in the three months to May, roughly 40% below the near-$75 billion of the prior quarter. Meanwhile, the syndicated loan market (the banks) is winning deals back. One loan-market analyst put it plainly: "the pendulum has shifted to the BSL market's advantage."
The insurance angle got legal. Two life insurers, Delaware Life and Clear Spring Life and Annuity, reportedly received federal grand jury subpoenas over whether billions in private credit holdings were properly disclosed. Last week this was an analyst's spreadsheet argument. This week it's a prosecutor's.

What's new

1. A top-five direct lender admits stress is running at double the normal rate

This is the week's most important sentence, and it came from an operator, not a critic.

David Golub, co-CEO of Golub Capital, a firm his host described as "a direct lending powerhouse with more than $90 billion in AUM," sat down on Credit Exchange with Lisa Lee (July 24). Asked what moment the industry is in, he split the answer in two.

His own borrowers are fine. Golub publishes a quarterly index tracking how the companies it lends to are performing, and he said it shows "continued, reasonably good growth. It's not great… It's slower than it was two years ago, but it's resilient growth in the face of the uncertainties."

Then the industry view:

"What I'm seeing as I look out more broadly at the industry is a high level of credit stress. And the best indicator of this is Fitch. So, Fitch does a very nice index of private credit defaults. And it's running at 2x, a little more than 2x normal. So, that's, I think, a very good indicator that we're in a period of higher than normal credit stress. And we're seeing with that dispersion in manager performance."

Two things matter here. First, "2x normal" is a specific, sourced number from a lender with skin in the game, not a bear's estimate. Second, Golub immediately paired it with dispersion: "very big differences between how some private credit managers are doing and how market-leading firms with real competitive advantages are doing."

Why it matters: the bull case for the big listed managers has always been "we are the good ones." Golub just made that the official industry framing, and the flip side is that he expects some firms not to survive. He called it "a tectonic Darwinian moment," predicting "the development of a core group of private credit leaders" over the next ten years "at a scale that we've not previously seen before." His evidence that consolidation is already here comes from the private equity side: "the top 200 firms today are raising such a disproportionate share of the new capital being raised in the industry that they control 91% of the dry powder. These are U.S. data. So, 200 firms control 91%. That is not a fragmented industry."

Golub also handed us two other headlines.

On private credit trading, the practice of lenders selling private loans to each other, which some in the industry treat as heresy: "We've done over $2 billion of private credit trading in the first half of 26. It's a record level for us." He's been running a capital markets and sales-and-trading desk for a decade, and he's unapologetic: "Two of the big misconceptions. One is it's new and the other is it's bad… It's not new and it's not bad." That is a direct challenge to the "we hold every loan to maturity" marketing that the industry has leaned on for years.

On the walls coming down between private and public credit: "I think some of the distinctions that we've historically drawn, like the distinction between private credit and BSL, I think those distinctions are going to start to fade some. There's going to be maybe more of a continuum where previously there were two very separate categories of financing." (BSL means "broadly syndicated loans," the large, tradeable bank-arranged loans that private credit spent a decade taking share from.)

On the redemption gates that have caused so much bad press, the rule letting a non-traded fund limit withdrawals to 5% of the fund per quarter, Golub defended the structure but conceded the flaw: "if many investors all choose to elect for redemptions at the same time, not all of them are able to get what they want… maybe it's time to lab it, maybe improve it, come up with a different mousetrap. But in and of itself, there's nothing inherently flawed." He also forecast "persistence of some elevated levels of redemptions in non-tradeds."

2. New lending fell about 40% in a quarter, and the banks are taking deals back

Two separate podcasts, from opposite ends of the ideological spectrum, told the same story about deployment.

On Eurodollar University (July 22), the host laid out PitchBook data reported by Reuters: "US-focused direct lending issuance fell to $44.76 billion during the 3 months ended May 2026. That was approximately 40% below the nearly $75 billion recorded in the first quarter. Issuance to private-equity-backed borrowers dropped by nearly 37%, down to $28.5 billion."

That is the deployment engine stalling. And it isn't only because deals aren't happening, it's partly because the deals that are happening are going somewhere else.

On Know More. Risk Better. (July 23), an episode titled "BSL Divided: AI Overhang vs. Flight to Quality," a loan-market analyst walked through the competitive picture in unusual detail, and this fills a gap this newsletter has been flagging for three weeks running:

"So the dynamic has shifted and I feel like the pendulum has shifted to the BSL market's advantage."

The reasoning is specific. For anything above roughly $500 million, the syndicated market is now cheaper: "I can do your deal at a lower coupon, push out the tenor, and come in with some favorable documentation for you… it seems like it's a pretty compelling trade right now." Private credit's remaining edge is smaller and more structural: "sub a hundred, sub $75 million EBITDA companies," deals needing Canadian dollars, and deals needing very large delayed-draw facilities (money committed now but borrowed later): "you can only do a delayed draw in the loan market. That's, you know, call it 15 to 20% of the deal size."

New money in the loan market rose about 14% quarter-over-quarter to almost $75 billion in Q2. Roughly $5.5 billion launched in July alone.

Why it matters: for the last three years, the direction of travel was one-way, deals leaving the banks for private lenders. That has now reversed at the top end of the market. Every large deal that clears in the syndicated market at a tighter spread is a deal that doesn't produce fee income or a high-coupon asset for a direct lender. And it explains the deployment number: it isn't only that there's less to lend against, it's that private credit is losing the bidding on the good stuff.

One nuance worth holding onto: the same analyst noted that in 2025, when loans migrated from private credit back to the syndicated market, "42 or 43 percent overall in terms of volume" was tech software, and that flow is not expected to repeat, because the syndicated market no longer wants that risk. Software is the one place private lenders can't hand the problem back.

3. Investigators are now inside the insurance-and-private-credit link

Last week's issue flagged insurance balance sheets as the new locus of the bear case. This week it stopped being theoretical.

Per Eurodollar University (July 22): "Delaware Life Insurance and Clear Spring Life and Annuity received federal grand jury subpoenas concerning whether billions of dollars of private credit holdings were adequately disclosed as backing for affiliated ventures. The SEC… was also reportedly conducting a parallel investigation."

To be fair to the facts, and the host was: "No criminal charges have been reported, and an investigation doesn't establish wrongdoing."

But the subject of the inquiry is the thing to watch, whether private assets "carry commitments, relationships, or encumbrances that outsiders can't easily see." That is exactly the question the whole insurance-plus-private-credit model turns on. The host framed the open issue as: "where is the leverage?" pointing at borrower-level debt, fund-level subscription and NAV facilities (loans secured against a fund's own assets), warehouses, offshore vehicles, and guarantees, "an obligation that appears harmless until multiple claims, they all arrive simultaneously."

The same episode used Amazon as the canary. Amazon sold $25 billion of bonds this month in eight tranches to fund AI and cloud buildout. Orders peaked around $62 billion, then fell to roughly $41 billion once the banks tightened pricing, a book covered only about 1.6 times, against an average of roughly 4 times for US high-grade deals this year. Bank of America estimated Amazon's longest bonds paid an extra 18 to 21 basis points versus its existing debt. The host's read: "Amazon is not the epicenter of the credit problem. It is the lighthouse showing how far the change in financing conditions has already traveled."

Why it matters: if the marginal buyer of AI infrastructure debt is getting pickier with Amazon, the pricing power sitting behind leveraged, floating-rate, privately-held borrowers is worse than the marks suggest.

4. A real redemption number, and a Fitch review of the non-traded funds

John Cole Scott, president of CEF Advisors and chairman of the Active Investment Company Alliance, gave the cleanest data run on BDCs of the week on The NAVigator (July 24). He is an analyst and allocator, not an operator, but his firm tracks the sector's numbers professionally.

The performance picture, year to date: closed-end funds overall up just under 9%, equity funds up 15%, muni funds up almost 6%, taxable bond funds up 1%, "and, of course, BDCs are off negative 8% total return."

Underneath that, size is winning decisively. His large-cap BDC index was up 3% in the second quarter and down only 4% for the year. The small-cap index was up 2% in the quarter and down 11% for the year. "These larger credit platforms, though not all are favorites, are generally performing better as investors are rewarding size, scale, usually cheaper leverage and cheaper expenses."

On valuation: "BDCs ended at such large discounts that it's only been wider during actual recessions that were angry like COVID and the Great Recession."

Two hard data points:

Fitch reviewed 13 of the larger non-traded/private BDCs and assigned three a negative outlook and ten stable. Importantly, Scott says the negatives aren't about loan performance: "negative outlook doesn't mean that the investments are doing bad. They're just seeing more issues with the amount of redemptions and the way the portfolios roughly cycle each 20% a year."
A concrete gate example. At a $3 billion non-traded fund, Barings Private Credit, "investors asked for 11% and they maintain the five that they're obligated and required to give as per prospectus… which means if you were that investor, you got 46%."

That last number is the one to keep. Not a hypothetical, a 46% fill on a redemption request.

His base case into earnings: "slightly more elevated non-accruals in software and generically for the sector. We feel that the software is not going to be hairy and angry, but just becoming more of a real pervasive risk." (A "non-accrual" is a loan the lender has stopped booking interest income on, because it no longer expects to be paid, the sector's headline stress metric.)

And a mechanic worth flagging for anyone holding listed BDCs: if forced sellers among non-traded funds have to move loans, "the most likely buyer is one of the larger group of listed BDCs, which means they might get a good discount on a good loan." Distress at one end of the industry could be an accretive buying window at the other.

5. The operators' rebuttal: "way, way, way oversold"

Two lenders pushed back hard this week, and both are worth taking seriously because both put their own money behind it.

Steven Klinsky, founder of New Mountain Capital, which runs a 17-year-old credit arm and the publicly traded BDC New Mountain Finance Corporation, was on Insightful Investor (July 21). On the headlines: "I personally think it's way, way, way oversold."

His core argument is about where you sit in the capital structure. On the software repricing that has spooked everyone:

"If we didn't want to pay, buy software companies, you know, 21 times when someone else was bidding 20. But if you're the senior six, even if the multiples fall to 15, that may hurt the equity holder. But you're very safe at the six times. And if it falls to 10, you're still very safe at the six times."

In plain terms: if you lent six times a company's earnings and the buyout price collapses from 21 times to 10, the equity owner is wiped out and the lender is still covered. The equity absorbs the loss first, and there is a lot of equity to absorb it.

He is buying: "I'm on record for having been buying my own shares." And on pricing, he made a point that cuts against last week's bear framing while arriving at the same fact, spreads are widening: "when they talk about the golden age of private credit, that was kind of the golden age of private credit fundraising. And the spreads had gotten tighter and tighter. And I think the spreads are winding out again. So I think it's a better time to be a lender."

He was also impatient with the transparency complaint: "if you look at our BDC, we list, you know, kind of every position. What is the multiple to EBITDA? How is it moving? We have, you know, red, green, yellow indicators. I think it's pretty dang transparent."

Trevor Clark, founder and managing partner of TPG's Twin Brook Capital Partners (310-plus borrowers and $30 billion in AUM), was on InsuranceAUM.com (July 29). He gave the most detailed defence of underwriting discipline anyone offered this week, and, more usefully, named the specific things that went wrong elsewhere.

His funnel: "We're going to look at 1,500 small companies every year. We're then going to down select around 10% of those… We're then going to do two to three months of due diligence to confirm all that and then close on three to 4% of those companies." Position sizes of 1% or less, cash-flow lending only, financial covenants required.

On the software blowups, he was precise about the mechanism rather than blaming the sector: "software is not a bad industry. People got sideways because you did a bunch of recurring revenue-based loans or enterprise value loans that you didn't have the cash flows to actually meet the current cash burden of it. And that's how they got sideways."

On redemptions, he was blunt to the point of rudeness: "when you see headlines and you see people starting asking for redemptions based on that, I would argue that they didn't belong there in the first place." His analogy, buy a new car, drive it to McDonald's, come back and ask the dealer to buy it back: "at the end of the day, it's like who's surprised at this when you get paid for an illiquidity premium and then you want liquidity. When you get more liquidity, the premium goes away. Like this is like finance 101."

But he also pointed a finger at his own industry: "there are direct lenders who made a very specific choice in the last five years to lean really into retail capital formation… when you see people with 30, 40, 50% of their capital being in that bucket for this asset class, I'm going to say that some of these managers needed to kind of rethink that as well."

And he flagged two structural problems that are genuinely new and don't fit either side's script:

Negative selection bias in loan sales. "That worked when a big direct lender had $5 billion of assets under management. When lenders now have $30, $50, $100 billion, they're not selling loans to other lenders. So you've got a negative selection bias that's going to start to show up." Meaning: when the giants only sell what they want to be rid of, whoever buys is buying the rejects.
Longer holds mechanically manufacture covenant breaches. "You go from the average private equity hold time of an asset of four years to today, it's over six years… We're stepping up covenants every year. So a deal that you held on to three years versus six years, your covenants continue to get tighter and tighter. If you don't have that earnings, you're going to start to create, we could talk about defaults. A covenant default is not a payment default, but it definitely shows in perception a higher level of stress just from hold times."

That is an underappreciated point and it should temper how the market reads a rise in technical defaults this quarter. Some of the coming default headlines will be arithmetic from private equity's exit backlog, not deterioration in the borrower.

He agrees with Golub on the destination: "I think there's going to be fewer direct lenders two, three, five years from now because people are going to realize you didn't need to have 2,000 plus. I don't think there's any doubt about that."

The debate

The bear case, steel-manned. The problem isn't any single loan. It's that too much money chased a fixed number of borrowers and ended up concentrated in exactly the wrong industry at exactly the wrong time.

Tony Yoseloff, Managing Partner and CIO of Davidson Kempner, made this argument better than any of the doomsayers on Money Maze Podcast (July 23), and he is an operator running credit money, not a commentator:

"40% of private equity dollars have gone into software businesses the last several years. And, you know, low to mid-30s percent of direct corporate lending dollars have gone into software businesses because if your clients are private equity firms, you have to follow your clients. And, you know, one of the inverse positives, I guess negatives, of what's going on in the AI is that, you know, software is in peril right now. There really is no getting around that as an asset class."

The structural version: "If everyone's in the same asset class and your asset class ultimately outgrows the size of borrowers that are available to lend to, that's where you can start to have real issues."

And a sharp observation about the culture of private marks, from someone who runs both public and private books: the discipline of marking to market every day "sometimes gets lost in the private markets where maybe people still pretend in their heads that their investments were worth what they thought they were worth, you know, a year ago or two years ago."

His most uncomfortable point for private credit investors: public equity portfolios have offset software weakness with AI winners, which is why the S&P is near a high. A private credit book has the software exposure with none of the offset.

The louder bears added scale and timing. Ed Dowd, a former BlackRock portfolio manager, said on Market Disruptors (July 22) that "private credit is currently frozen," sized the market at "somewhere between $2 and $3 trillion," and argued it "was the incremental driver of credit growth in the US in '24 and '25," with the corollary that "the growth of the private credit industry has been so big and so fast that marginal credit and new credit created has probably been to credits that aren't that sound."

Danielle DiMartino Booth, CEO of QI Research and a former Dallas Fed staffer, put the timing on the calendar on Hedgeye Podcasts (July 24). Her point was that a lot of 2021–22 paper was written short: "a lot of them were done with 5-year maturities. So we're talking about a 2026, late 2027 event for a lot of this paper, whether it be high yield or commercial real estate that was not underwritten based on having a good long decade… We're not there. We don't have that much time." On the collateralised loan obligation market she was terse: "CLOs aren't any healthier today than what they, they're just bigger." And on sequencing: "It starts with credit and it ends in the stock market, which is kind of always the last to wake up to reality."

On Rebel Capitalist News (July 21), an episode billed as a BlackRock insider's warning, though what's actually in it is host George Gammon narrating and reacting to a research note, the most useful idea was about a statistic people are misreading. The average credit quality of the public junk bond market has improved. That sounds like good news. It isn't: "the dog crap is moved from this market that is at least transparent, has moved into the black box." Public junk looks healthier because the weakest borrowers left it. Worth flagging because the title will circulate: there is no BlackRock insider speaking on that episode.

The bull case, steel-manned. Nothing that made private credit attractive has actually changed; what changed is who owns it and how loudly they complain.

Trevor Clark's version: "Coming out of the GFC, people started to realize, wow, I can move from kind of public credit to private credit. I can get this quote unquote illiquidity premium. And oh, by the way, my losses in private credit are actually better than in public credit. And non-mark to market and floating rate, all those. Guess what? None of those facts, they haven't changed. Not a bit. What has changed? Well, you had a lot of capital get raised in that asset class over the last 15 years."

The numbers on that side are not trivial. On In The Harbor (July 24), the discussion put bank loan defaults at 1.3% and private credit at roughly 1.6%, against a yield pickup of roughly 200 basis points, bank loans around SOFR plus 300, private credit around SOFR plus 500. The arithmetic: "if the difference between the bank default is 1.3 and private credit is like 1.6, so it's only 0.3%, but I'm making two points more in yield, I'm a winner." (SOFR is the floating benchmark rate US loans price off.)

Same episode, on the frauds that generated the worst headlines, the argument that these were crimes, not credit misjudgements. On Tricolor: the assets weren't properly registered with public lien filings, "so because there wasn't a UCC designation put on some of the assets that were backing those loans. They were able to pledge the assets multiple times." Why would a lender skip that step? Cost: on small subprime auto loans, each filing "might be $1,500." The verdict: "that's still nefarious, you know, bad boy behavior by the executives over there. They know what they were doing." Underwriting standards don't protect you from someone pledging the same car twice.

The pull-quote of the week, because it captures both sides at once, an operator conceding the stress while insisting the industry will consolidate rather than break:

"What I'm seeing as I look out more broadly at the industry is a high level of credit stress… Fitch does a very nice index of private credit defaults. And it's running at 2x, a little more than 2x normal." David Golub, co-CEO, Golub Capital, on Credit Exchange with Lisa Lee, July 24

Where the two sides actually agree. This is the useful part. Both camps now agree on three things: defaults and stress are running above normal; the software concentration is real and unhedged; and the industry is going to shrink in number of participants. What they disagree on is whether that's a cycle clearing out weak hands (Golub, Klinsky, Clark) or the first innings of something that takes the insurance system and the banks with it (Dowd, DiMartino Booth). Nothing published this week resolves it. The resolution starts with BDC earnings, which begin now.

Stocks in play

Blackstone (BX) reported Thursday. Per Steve Eisman's read on The Real Eisman Playbook (July 24): earnings per share of $1.52 versus $1.21 last year and $1.36 expected, beating "on higher transaction revenues and strong performance fees thanks to better-than-expected realizations," with fundraising "very strong at $68 billion."

Bull: $68 billion of fundraising in a quarter when the asset class was being called a bomb on half a dozen podcasts is a real answer to the "flows are broken" thesis. Realizations came in better than expected, which addresses the return-of-capital complaint directly.
Bear: Eisman's caveats are the ones to watch, "slow base management fee growth and only modestly positive performance in certain asset classes like real estate and credit." Base management fees are the durable, recurring part of the earnings stream; transaction fees and realizations are not. And on the AI exposure: "on the conference call, management bragged that it is the biggest financier of AI data centers. Given the intensity of the current AI debate, that may or may not prove to be a great bet."
Next catalyst: Eisman's timing call is the one to diarise, "Private Credit's problems with software won't really start to matter until next year when the refinancing cycle begins."

Ares Capital (ARCC) reports today, July 29, and John Cole Scott flagged it as the event that opens the season: "the earnings season kicks off. The largest big one is Aries Capital on the 29th."

Bull: Scott's whole framework says large-cap BDCs are being rewarded, his large-cap index is down only 4% year to date versus 11% for small caps, and current discounts have only been wider in genuine recessions.
Bear: His own base case is "slightly more elevated non-accruals in software and generically for the sector." As the biggest BDC and the season opener, ARCC's software marks will set the tone for every name behind it.
Next catalyst: Today's print, and specifically the non-accrual rate and any change in the payment-in-kind share of income.

Golub Capital BDC (GBDC): David Golub did not discuss GBDC's numbers, marks or non-accruals on the podcast; treat the read-through as firm-level, not fund-level.

Bull: His borrower index shows "resilient growth," and $2 billion of loan trading in the first half is a genuine differentiator in a market where liquidity is the complaint.
Bear: He himself says industry defaults run at twice normal with wide manager dispersion, an argument that only works for you if the firm lands on the right side of it, which the market will want proof of.
Next catalyst: the next quarterly borrower index publication, and BDC earnings.

New Mountain Finance (NMFC): Steven Klinsky said on the record he has been buying his own shares. Insider buying by a founder in a sector trading at recession-level discounts is a data point, though it is not a substitute for the numbers.

Bull: Klinsky's senior-at-six-times argument, plus per-position disclosure with red/amber/green flags.
Bear: He is talking his own book, and he did not give current non-accrual or mark data.
Next catalyst: BDC earnings season.

Wells Fargo (WFC): CEO Charlie Scharf on Bloomberg Talks (July 22), confirming Wells advised on Apollo's Broadcom AI Center deal. Asked directly whether he sees credit quality concerns in private markets:

"There are some that do it really, really well, that have the experience, that have been through cycles, that have the analytics, that have the rigor. Those are the ones that we try and do business with… And we feel really good about the credit support that we're providing for them. There are others out there that, you know, people, you know, potentially should be more concerned about, but we don't have broad exposure to that."

And on what happens in a downturn: "if credit does turn down, it's just like in the banking universe. Some do it really well. Others do it less well… And that'll be shown up away from us."

Bull: Wells is monetising private credit as a fee-paying client rather than competing with it, and Scharf says the bank is still leaning into AI data-center financing: "We are… But you've got to be selective about who you're financing and what the structures are."
Bear: "That'll be shown up away from us" is exactly what every bank CEO said about every credit exposure before every credit cycle. He gave no sizing of the bank's lending to non-bank financial institutions.
Next catalyst: any disclosure of exposure to non-depository financial institutions, which is where bank-to-private-credit risk actually sits.

Apollo (APO) was named only in passing this week, as the sponsor of the Broadcom AI Center deal Wells Fargo advised on. No Apollo executive gave substantive credit or fundraising commentary on any episode we found. Nothing new to underwrite.

Read-throughs

Listed BDCs (ARCC, BXSL, OBDC and peers). Two forces, pulling opposite ways. Against them: John Cole Scott's expectation of "slightly more elevated non-accruals in software and generically for the sector," and the fact that Fitch has put three of thirteen large non-traded BDCs on negative outlook, a sentiment overhang even though the driver is redemptions rather than loan losses. For them: the size premium is working (large-cap index −4% year to date versus small-cap −11%), discounts are at levels "only been wider during actual recessions," and if non-traded funds become forced sellers, the listed vehicles are the natural buyers "at a good discount on a good loan." The trade inside the sector is size and scale, not yield.

Non-traded / semi-liquid BDCs. The Barings Private Credit example (11% requested, 5% honoured, a 46% fill) is now the reference point. Expect more marketing language to change: one podcast noted "a lot of funds are changing their marketing to say, no longer say semi-liquid." Golub's own forecast is "persistence of some elevated levels of redemptions." Watch for the second-order effect: forced portfolio turnover at funds that never intended to sell.

Insurance balance sheets. This is now the highest-consequence thread. Federal grand jury subpoenas at Delaware Life and Clear Spring Life and Annuity over disclosure of private credit holdings, with a reported parallel SEC inquiry, moves the argument from "these marks look generous" to "regulators are asking where the encumbrances are." On TFTC (July 27), Dave Collum described the model bluntly: private equity firms buying insurers, replacing the balance-sheet manager with themselves, "and then pushing it into their own products, which are then taking private equity risk and private credit risks." That's polemic, not analysis, but it's the same structure the subpoenas are probing.

Banks and regional lenders. No dedicated episode on regional bank exposure this week. The read has to come from Scharf, and it is a two-tier one: the money-centre banks are partnering with the top-tier managers and earning fees, while explicitly declining to be near the second tier. Klinsky adds a competitive angle from the other side: "Traditional banks are great… the regulation is loosening up. So I think they have more borrowing, more lending capacity than they could have under a different administration." Combined with the pendulum swinging back to the syndicated market, this is the clearest picture yet that banks are regaining ground at the large end.

Syndicated loans and CLOs. Genuinely mixed. New money is coming back, Q2 new-money issuance up about 14% quarter-over-quarter to almost $75 billion, roughly $5.5 billion launched in July alone. But CLO formation, which is the main buyer of these loans, has slowed: "when you're not printing 12 to 14 billion of new issue CLOs a month, it does change the technical dynamic." Retail flows are curiously split, high yield and syndicated loan funds seeing outflows year to date, CLO funds seeing inflows. Repricings and resets are keeping existing vehicles able to reinvest, which supports the market, "but it's not to the degree of ramping a new $500 million vehicle."

Data centres, AI financing and asset-backed lending. The AI buildout is quietly restructuring the whole leveraged finance market, and it is moving toward public markets, not private ones. CoreWeave did a data-centre institutional loan of over $3 billion, described as "the first data center institutional loan transaction in this modern era," and is back with another GPU-backed deal. A QTS refinancing "launched, sized at $1 billion and wound up being upsized to $3.25 billion" and secured investment-grade ratings, making it attractive to CLOs. Dwyer Omega cleared at SOFR plus 250 at a price of 99.75, tightened twice. In high yield, data-centre deals were roughly a quarter of second-quarter issuance. There is even a power angle: a large buyout loan for Pathfinder Power, LS Power buying assets from Constellation, "drew a pretty broad base because obviously the data centers are sucking up a ton of power."

The investor logic is worth understanding: this is how credit investors hedge their own AI exposure. Asked whether buyers like these deals because the collateral is tangible, the analyst said: "I feel like, yes, this is a way to sort of play the other side of the AI trade." You own the physical infrastructure rather than the software companies AI might eat. But the honesty was equally clear: "people are still wrapping their brains around how these deal works and you kind of take it a deal at a time… it's a very different prism through which you look at, assess risk."

The other side of that trade is where credit is actually being priced painfully. The same episode described one over-levered borrower getting a second lien bond done "at 12% at 95% for like 13.3% yield," on a company levered around 7.7 times. That is what the marginal cost of capital looks like right now for a stressed issuer.

Finally, a ratings note. On Marketplace (July 24), Moody's was reported as warning that heavy borrowing by the largest AI spenders "threaten credit quality," that lending to them is getting riskier and will get more expensive. When the rating agencies start saying that about investment-grade hyperscalers, the pricing for everyone below them moves too.

Retail wealth and evergreen structures. Thin coverage, but one relevant note: on Market Matters from New York Life Investment Management (July 27), Jeremy Held of Bow River Capital argued that evergreen funds should be understood as long-term diversified vehicles rather than semi-liquid products, with the real benefits being immediate deployment and diversification rather than liquidity. That is the same repositioning Golub and Clark are making from the credit side: the industry is collectively walking back the liquidity promise.

What changed vs last week

1. The bear case got an inside witness. Last week's lead was PIMCO's Lotfi Karoui saying "a default cycle in credit has begun," a strategist's call. This week David Golub, running a $90 billion-plus lending business, put a sourced number on the same phenomenon: Fitch's private credit default index at "2x, a little more than 2x normal." The bull-versus-bear argument is no longer about whether stress is elevated. Both sides now concede it. The argument has moved to whether it's dispersion (some managers are bad) or contagion (the structure is bad).

2. A three-week-old gap finally closed: we got the BSL-versus-private-credit comparison. For three consecutive issues this newsletter has flagged the absence of any manager or analyst comparing new-loan pricing between private credit and the syndicated market. This week it arrived, and it went against private credit: "the pendulum has shifted to the BSL market's advantage," with private credit's remaining edge confined to sub-$75 million EBITDA borrowers, delayed-draw structures and non-USD needs.

3. The insurance thesis escalated from analysis to law enforcement. Last week, Nick Nemeth's argument about insurer balance sheets was a modelling exercise: thin real capital, Level 3 assets, surrender risk. This week, federal grand jury subpoenas at Delaware Life and Clear Spring Life, with a reported SEC parallel. Same thesis, entirely different risk profile for anyone holding an insurance-linked alternative manager.

4. The redemption story got a specific, citable number. Last week we had Brad DeMong's range, redemption requests running from "high single digits to up to 40%" per quarter against a 5% design. This week: Barings Private Credit, 11% requested, 5% honoured, 46% fill. Plus Fitch's review of 13 large non-traded BDCs with three on negative outlook explicitly for redemption pressure rather than credit.

5. Deployment got quantified, and it's worse than the qualitative commentary implied. Nobody last week put a number on new lending volume. This week: US direct lending issuance of $44.76 billion in the three months to May, about 40% below the prior quarter, with private-equity-backed issuance down nearly 37% to $28.5 billion. That is a fee-and-yield problem for every manager in the space, independent of credit quality.

6. Blackstone's data-centre exposure went from a permitting story to an earnings story. Last week the news was Blackstone walking away from a Prince William County, Virginia project after a court voided rezoning. This week Blackstone reported and, per Eisman, "management bragged that it is the biggest financier of AI data centers." The exposure is now a stated strategic identity, not an incidental one.

7. Partial convergence on spreads, opposite conclusions. Last week Karoui argued the illiquidity premium was "too thin, needs to reset." This week Klinsky said "the spreads are winding out again. So I think it's a better time to be a lender." Both observe the same widening. One calls it a warning, the other calls it an entry point. That's the entire debate in miniature.

Where last week's view stands unchallenged. Nobody this week contradicted the software concentration thesis, in fact Yoseloff strengthened it with the sizing (40% of PE dollars, low-to-mid-30s percent of direct lending dollars into software). Nobody contradicted the mark-dispersion concern. And nobody disputed that non-traded vehicles are under redemption pressure, the operators simply argued the investors shouldn't have been there.