Private credit deep dive. KKR's own real estate lender reserved 6.5% of its loan book, cut its dividend 60%, and put itself up for sale — on the same day Wall Street decided private credit was fine again. The loan book was never the story. The liability side is. Here is the trade that comes out of it.
Standing rule: no new call until the last one is marked to market. On July 24 I published "Private credit became AI's balance sheet". One trading session plus a weekend later:
| Call (7/24) | 7/24 price | Now (7/27 midday) | Move | Verdict |
|---|---|---|---|---|
| LONG ARCC — HIGH CONV | $18.77 | $18.95 | +1.0% | Working, small. Open. |
| LONG BXSL — HIGH CONV | $23.12 | $23.19 | +0.3% | Flat. Open. |
| LONG OBDC — SPEC | $10.78 | $10.93 | +1.4% | Working. Open. |
| LONG APO — SPEC (entry $112-118) | $121.11 | $124.25 | +2.6% | NEVER TRIGGERED. I set the entry too low and missed 2.6%. My error — see below, I am raising it. |
| FADE FSK (entry $11.00-11.40) | $10.63 | $10.86 | +2.2% | Losing on logic. Never triggered, and FSK is the best performer in the complex today. Being honest: the KKR-support narrative is stronger than I gave it credit for. |
| AVOID OWL | $9.32 | $9.60 | +3.0% | Losing. OWL is the single best performer in the group over the window. Thesis intact but the tape disagrees. |
| TCPC UNINVESTABLE | $3.33 | $3.21 | -3.6% | Working — the only short leg that worked. |
Honest read: the long leg is up ~+0.9% on average; the short leg is up ~+0.5% against me. The pair is net negative. The specific failure is that I paired quality longs against weak-balance-sheet BDCs when the actual weakness in this cycle is in weak-liability-structure vehicles. FSK has a sponsor writing $300M of checks. OWL is a manager, not a credit book. TCPC — the one with an eroding NAV, a departed CEO and federal valuation scrutiny — is the only short that behaved. I am replacing FSK with TCPC as the short leg and raising the APO entry. Detail in Section 7.
Monday, July 27, midday. The US–Iran fighting pause is holding into a third night. Brent is down 6.8% to about $90.25, WTI down 6.1% to $83.83. The 10-year Treasury is at 4.6345%, down ~4bp; the 2-year at 4.30%; the 30-year at 5.12%. Durable goods orders rose just 0.3% in June against a 2.1% consensus — a genuine miss that nobody is talking about because the chips are on fire (SMH -4%, AMD -8%, Teradyne -8%, Micron -5%).
And here is the number that frames this entire piece: XLF, the financials sector ETF, printed an all-time high of $57.05 this morning (last $56.76, +0.80%), taking out the July 16 closing high. Financials at a record. Simultaneously, per the Raymond James BDC Weekly Insight dated July 23, 2026, the listed BDC complex trades at an average 5.4% discount, with BlackRock TCP Capital at 0.47x book and BlackRock Capital Investment at 0.45x.
The consequence: the market is willing to pay a record price for the deposit-funded lender and 47 cents on the dollar for the fund-funded lender. That is not a credit judgment — banks and BDCs lend to overlapping borrowers. It is a funding-structure judgment. The market has correctly identified that in a redemption cycle, the liability side is what kills you. It has just applied that judgment with a shotgun instead of a scalpel.
| Asset | Price (7/27 midday) | Chg | Note |
|---|---|---|---|
| SPY | $737.59 | -0.18% | Gave back a +0.9% open |
| QQQ | $679.04 | -0.76% | Chip complex dragging |
| IWM | $291.67 | +0.17% | Rate relief helping small caps |
| XLF | $56.76 | +0.80% | All-time high $57.05 intraday |
| HYG | $79.29 | +0.08% | High yield not confirming any stress |
| BIZD (BDC ETF) | $12.52 | +1.46% | Outperforming SPY by 164bp today |
| 10Y UST | 4.6345% | -4bp | FOMC Wed 2pm; hike risk still priced |
| Brent | ~$90.25 | -6.8% | Iran pause; was >$100 last week |
| Gold | $4,108.91 | +1.39% | Dollar softer |
Deep-dive standard requires reading the actual document. I read KKR Real Estate Finance Trust's Q2 2026 earnings release, the accompanying 8-K reporting the material event, and the July 22 earnings call transcript, where IR director Jack Switala walked the loss line.
The headline was a beat. The headline was irrelevant. KREF reported adjusted EPS of -$0.58 against a -$0.65 consensus — a 10.8% "beat" on the loss line — while revenue of $18.18M missed the $20.80M estimate by 12.6%. Beating on a smaller loss while missing on revenue is not a beat. It is a company earning less and reserving more.
Here is what was actually in the filing:
| Line item | Q2 2026 | What it means |
|---|---|---|
| GAAP net loss | -$121.8M / -$1.95 per share | Against a $433M market cap. One quarter destroyed 28% of the equity market value. |
| Distributable loss | -$36.4M | Even stripping the non-cash reserve, the business does not cover itself. |
| CECL allowance | $291.9M, up $119.8M in the quarter | On a ~$4.5B loan portfolio = 6.5% of the book reserved. |
| Composition of the hit | $75.1M credit loss provision + $44.8M fair-value write-down on a held-for-sale loan | The write-down is the tell: they marked a loan they are actually trying to sell. That is a transaction print, not a model. |
| Book value per share | $10.24 | Stock at $7.37 = 0.72x book. TTM EPS -$2.92, ROE -16.7%. |
| Liquidity | $721.6M | Real. This is not a liquidity failure — it is a solvency-of-earnings failure. |
| Dividend | Cut to $0.10/share | Yield now 5.45%, down from a double-digit print. A 60% cut. |
| Legacy office exposure | ~18%, targeting <10% by YE2026 | Getting from 18% to under 10% means selling ~$360M+ of office paper into a bid that does not exist at par. More realized losses are coming. |
| Board action | Strategic review committee formed — sale, merger, asset sales, or continue current plan | KKR-managed vehicle publicly exploring its own disposal. |
Every criticism of private credit valuation reduces to one complaint: the marks are models, not trades. The held-for-sale designation removes that defence. KREF is not modelling this loan — it is selling it, and it took a $44.8M haircut to clear. That is a real print in the CRE credit market, and it is the single best datapoint anyone has published this quarter on where private credit paper actually clears versus where it is carried.
Downstream consequence #1: Every other CRE-adjacent BDC and mortgage REIT now has a comparable transaction to be measured against. Expect auditor pressure on Q3 marks across the group.
Downstream consequence #2: KREF's 6.5% reserve rate lines up almost exactly with Fitch's 6.0% record private credit default rate and the ~5.5% sector BDC non-accrual reading. When a model-based measure (CECL), an event-based measure (Fitch defaults) and an accounting measure (non-accruals) all land within 100bp of each other, that is not coincidence. Six percent is the number. Anyone still quoting the industry's headline 1.6% default figure is quoting a number Moody's itself has publicly reconstructed — Moody's puts the true private credit direct-lending default rate at 4.7% once distressed exchanges are included, versus the 1.6% headline. The gap is amend-and-extend.
Here is what changed in 2026 that most sell-side private credit notes still have not internalised. The problem is not that the loans are bad. The problem is who owns them and how quickly that owner can ask for money back.
Note the asymmetry in that last bullet. BCRED's NAV went from ~$82B to $77.6B while meeting redemptions and while its inception-to-date return stayed at 9.0%. The fund is shrinking without breaking. That is a successful gate — and it is exactly why the equity of the manager (BX +1.0% today, $131.30) is behaving better than the equity of the levered credit books.
Today, July 27, Apollo and Ares announced distribution of private equity, private credit and infrastructure funds to European retail customers through Revolut, for as little as €1. Apollo also committed $1.5B into Keppel's offshore energy fund, and Blackstone's credit arm is reportedly near a deal for HSBC's Australian loan book.
Read those two facts next to each other. The largest retail private credit vehicle in the world capped withdrawals at 5% seven weeks ago. Today the same asset class opened a €1 retail door in Europe. That is not a credit signal — the underwriting on new 2026-vintage paper is genuinely better than 2021-22 vintage. It is a distribution-cycle signal, and distribution cycles top before credit cycles bottom.
A gated fund is a fund that cannot originate. If your quarterly repurchase queue is at the 5% cap, your entire deployable cash flow goes to the queue, not to new loans. BCRED alone is $77.6B. Add Cliffwater, HLEND, ASIF, OBDC II and the peer set, and you have removed tens of billions of annual origination capacity from the middle market — capacity that spent five years bidding spreads down.
Second-order effect: new-issue direct lending spreads widen. That is margin expansion for anyone with permanent capital who is still open for business.
Third-order effect: permanent-capital listed BDCs — ARCC ($13.6B cap), BXSL ($5.4B), MAIN ($5.0B) — cannot be redeemed. Their shareholders can only sell to each other. For the first time since 2021, the listed BDC has the better liability structure, and it is being priced at a discount to book while the vehicle with the worse liability structure is being sold to Europeans for a euro.
Fourth-order effect, and the one that actually matters for portfolio construction: if you believe this, you should not be short the private credit complex. You should be long the permanent-capital slice and short the redemption-exposed slice. That is a very different trade from "short private credit."
The largest private financing ever executed closed on June 9, 2026: Apollo led a $35 billion capital solution for the Broadcom AI XPV platform, in partnership with Blackstone and leading global banks, funding Anthropic's compute expansion. Critically, the structure was designed to keep the debt off Broadcom's balance sheet. As of this month, banks have begun trading and syndicating pieces of it to a wider pool — insurers, retail wrappers, other funds.
Blue Owl has committed at least $50B to data center financing. Moody's has flagged hyperscaler AI capex at ~$785B in 2026 heading toward ~$1T in 2027, with roughly $460B of direct debt and $1.2T of off-balance-sheet lease commitments.
Quantifying the mechanism: US banks' direct private credit loan exposure is around $285B per Moody's, inside a broader ~$1.2-1.5T of bank lending to non-depository financial institutions. The FSB's May 2026 Report on Vulnerabilities in Private Credit identifies subscription lines as the dominant channel. Translation of that plumbing into a tradable statement: a 6% loss rate on a $1.8T market is roughly $108B of losses; that is absorbable by fund equity. A funding-line withdrawal against levered funds is not, and that is the tail that matters. It is not the base case, but it is why you size this sector at 3-5% of a book, not 10%.
Price/NAV data as of the Raymond James BDC Weekly dated July 23, 2026; prices and fundamentals as of midday July 27.
| Ticker | Price | P/NAV | Yield | TTM EPS | ROE | Verdict |
|---|---|---|---|---|---|---|
| ARCC | $18.95 | 0.95x | 10.2% | +$1.63 | +8.1% | OWN. Non-accruals ~1.2-1.5% vs sector 5.5%. Forward P/E 10.4. Q2 print Wed. |
| BXSL | $23.19 | 0.88x | 13.3% | +$1.91 | +7.1% | OWN. ~98% first lien, beta 0.43. Cheapest quality name in the group. |
| MAIN | $54.03 | 1.59x | 6.0% | +$4.75 | +14.3% | Best operator, worst price. ROE is real; 1.59x book is not a margin of safety. |
| GBDC | $12.91 | 0.89x | 10.3% | — | — | Fine. Nothing to add over BXSL. |
| OBDC | $10.93 | 0.74x | 11.5% | +$0.71 | +4.8% | SPEC LONG. 24% discount to $14.41 NAV. Insiders bought at $11.21-11.92. Q2 Aug 5. |
| FSK | $10.86 | ~0.63x | 15.8% | -$1.96 | -9.4% | Non-accruals 8.1%; rev -15.7% YoY. Cheap for a reason, but KKR is supporting it. Downgrading my fade to WATCH. |
| NMFC | $6.90 | ~0.65x | ~14% | — | — | Non-accruals 2.6% Q1, dividend cut 22%. Avoid. |
| TCPC | $3.21 | 0.47x | 21.5% | -$1.49 | -19.1% | SHORT / UNINVESTABLE. Rev -27.4% YoY. CEO departed amid markdowns and federal scrutiny. $265M cap. Q2 Aug 6. |
| KREF | $7.37 | 0.72x | 5.5% | -$2.92 | -16.7% | WATCH ONLY. Strategic review is the only reason to own it. See Section 7. |
The dispersion is the whole opportunity. MAIN at 1.59x book and BCIC at 0.45x book is a 114-point spread inside a single, allegedly homogeneous asset class. Markets that wide are markets that have stopped doing credit work. The discipline is simple: buy the names where reported ROE is positive and non-accruals are below the 5.5% sector line, at a discount to book. That is a two-name list: ARCC and BXSL.
| Ticker | Price | Today | Note |
|---|---|---|---|
| BX | $131.30 | +1.0% | Q2 beat, DE $1.52/sh. PT raises: TD Cowen $145, Oppenheimer $140, GS $127. Best manager, fully priced. |
| APO | $124.25 | +1.3% | Originated the largest private financing ever and syndicated it. Morgan Stanley OW, PT $164. Fee engine, not credit holder. |
| KKR | $100.73 | +1.4% | $16B Kuwait pipeline deal w/ BX & Brookfield. But owns the KREF manager — reputational drag. |
| ARES | $127.36 | +0.7% | Record fundraising. Fine, no edge. |
| OWL | $9.60 | +1.6% | TTM P/E 169x, fwd 187x, EPS $0.13, ROE 3.9%. Down 54% from its $20.68 52-week high. PT cuts: BMO to $11, Oppenheimer to $15. Retail fundraising reportedly evaporated. |
My opinion on OWL, stated plainly: Blue Owl is the wrong pairing in this cycle. It is simultaneously the most retail-distribution-dependent large manager and the largest data center private credit lender ($50B+ committed). That is funding that can leave, against collateral that cannot be sold. A 54% drawdown does not make it cheap at 187x forward earnings. I would not own it above $7.75.
The referee arrives this week. ARCC reports Wednesday July 29 at 12:00pm ET — two hours before the FOMC decision at 2pm. Then OBDC on August 5, TCPC and FSK on August 6.
Trigger: ARCC prints Wednesday with non-accruals at or below 1.5% of fair value, NAV flat-to-up versus ~$19.80-19.94, and reaffirms the $0.48 dividend. Fed holds. BCRED's Q3 repurchase queue comes in below the 5% cap.
Path: The "private credit is breaking" narrative loses its last data leg. Quality BDCs re-rate toward book. ARCC to $21.50 (1.08x book), BXSL to $26. BIZD +12-15% over eight weeks. The 5.4% sector discount compresses to 2%.
Tell to watch: ARCC's spillover income line. A spillover above $1.30/share is dividend armour and the single most bullish thing the print can contain.
Trigger: ARCC non-accruals tick to 1.5-2.2%, NAV down 0.5-1.5%, dividend held. TCPC and FSK print further NAV erosion on August 6.
Path: Bifurcation is confirmed in the data. Quality grinds up 6-10% over three months; the impaired tail makes new lows. ARCC $20.25-20.75, BXSL $24.50-25.25, TCPC $2.40-2.70, KREF drifts to a deal or to $6.50. The pair trade — long ARCC+BXSL / short TCPC+OWL — carries the return, not beta.
Tell: whether PIK income as a share of BDC investment income (currently ~8%) rises again. Above 10% and the base case degrades toward bear.
Trigger: Any one of: (a) ARCC non-accruals above 2.5% of fair value on Wednesday; (b) the Fed hikes Wednesday — still meaningfully priced per CME FedWatch — sending the 10Y through 4.85% toward 5%; (c) the $35B Broadcom/Anthropic syndication reprices materially wider than issue as it trades out; (d) a second BDC announces a strategic review or dividend cut.
Path: Discounts widen across the board. The whole complex -15% to -25%. ARCC to $16.50, BXSL to $20, TCPC below $2, KREF to $5.25 (its April low). Alt manager multiples compress a further 20%.
Why 20% and not lower: a rate hike into a durable-goods miss of 0.3% versus 2.1% expected would be a policy error the credit market prices instantly. That is a live risk on Wednesday, not a tail.
Equal-weight both legs. This is the cleanest expression of the entire thesis: long permanent capital with positive ROE at a discount to book, short redemption-exposed structures with negative ROE and unsustainable distributions. It is market-neutral to the FOMC outcome, which matters more than usual with a live hike risk on Wednesday. Invalidation for the pair: the long leg underperforms the short leg by more than 6% over any rolling two-week window.
Private credit equity should be 3–5% of a diversified book, not 10%. The base case is fine and the yields are real, but the tail — a bank funding-line withdrawal against levered funds, per the FSB's subscription-line channel — is a genuine, if low-probability, gap risk with no liquidity to exit through. Take the 10-13% yield on ARCC and BXSL as the core; express the negative view through the short leg rather than by underweighting the sector; and do not add on Tuesday. Wait for Wednesday's ARCC print.
Private credit's honest loss rate is 6%, confirmed three independent ways this quarter, and 6% on senior-secured paper yielding 10-13% is a survivable cycle, not a crisis. The failure is not in the loans; it is in the wrappers. KREF's Q2 filing — a $291.9M allowance on a $4.5B book, a $44.8M held-for-sale write-down that produced a real transaction print, a 60% dividend cut, and a board that has hired itself out for sale — is what the bad end of this market actually looks like when it stops pretending.
And the good end is on sale at 0.88 to 0.95 times book with positive returns on equity, on the same day the financials sector ETF made an all-time high. That is the trade. Buy ARCC and BXSL, own APO as the arranger rather than the holder, short TCPC against them, and do not touch KREF above $6.80 or OWL above $7.75.
Nothing here is investment advice. Opinions are my own and clearly marked as such; every factual claim is sourced above. Positions can and do change — invalidation levels are stated so you can hold me to them.