Autoresearch: private-credit defaults at 5-year highs while institutional fundraising sets a record

Generated by /autoresearch on 2026-08-18. Step-2 macro bucket #9 financials & the rate regime (thin vertical). 5 fetches. Priors capture skipped — unattended weekday shift. Context: vault/projects/stock-market

Summary

The private-credit book is printing the worst borrower-quality numbers since 2021 at the same time as the strongest institutional fundraising year on record. That is not a contradiction if the two clienteles have diverged, and the sourced numbers say they have.

Borrower quality (primary + named secondaries):

• Blue Owl Capital Corporation (OBDC) first-party: investments on non-accrual 2.8% of cost / 0.8% of fair value as of 2026-06-30, up from 2.0% / 1.0% on 2026-03-31. The 2.8% at-cost figure is the number the WSJ (and recaps) call a five-year high. (OBDC Q2 2026 release via Morningstar/PR Newswire; OBDC Q2 earnings presentation)
• WSJ analysis of quarterly reports (recapped 2026-08-10 by InvestmentNews; 2026-08-17 by FXStreet/Money Metals): loan defaults at funds managed by Ares, Blackstone, Blue Owl, and Golub Capital "touched their highest levels since at least 2021," exceeding the 2023 hiking-cycle spike. (InvestmentNews)
• Fitch: US private-credit default rate 6% through Q2 2026 (InvestmentNews: "a record 6%"; FXStreet: 6% at end-May, 14 defaults in May alone, concentrated in healthcare providers, industrials, business services).

The other side of the barbell:

• With Intelligence (via InvestmentNews): private-credit fundraising $119B in Q2 2026, H1 $190B (+53% vs H1 2025), already 80% of full-year 2025's $240B. Direct lending $73B in Q2. Ares Pathfinder III closed June 2026 at $8.5B — largest asset-backed finance fund ever raised.
• Retail is the stressed sleeve. Redemption requests from the top 10 non-traded BDCs averaged 13% of assets in Q1 and 14% in Q2 2026, forcing gates. '40 Act private-credit AUM ~$654B in Q1 (BDCs $561B) edged down Q4→Q1.
• Boston Fed (Fillat et al., via InvestmentNews/Axios): BDC loans structured as PIK rose from ~5.4% in Q1 2022 to 9.8% in Q1 2026 (peak ~9.85% Q4 2025). Fillat: "a sign of stress."

Concentration that cross-links this vault: defaults currently in healthcare and oil-price-exposed industrials; the feared next sleeve is software, 20%+ of many PC loan books — the same collateral agentic-ai-seat-erosion-to-saas-rerate and the July bdc-redemption-spiral-to-private-credit-repricing hypothesis already named. FXStreet (citing WSJ): "Analysts worry that the default bug could begin to spread into the software sector."

This does not graduate the BDC-short. Institutional LPs are still writing larger checks into specialty finance / asset-backed, which is exactly the "Goldman leaning in while retail gates" tension logged on 2026-07-14. What it does do is date the credit-deterioration leg with first-party 2.8% (OBDC) and a 6% Fitch print, and put a dated catalyst on the calendar: Jackson Hole, Warsh keynote 2026-08-28 (InvestingLive: first Jackson Hole as chair; "frame big-picture questions rather than offer near-term guidance"; hike/hold odds for the 2026-09-16 FOMC "close to even"). Higher-for-longer is the floating-rate borrower's problem; a cut is the BDC-NAV's relief valve. Either print is load-bearing for this cluster.

Findings

Theme 1 — OBDC 2.8% is first-party, not a WSJ inference

The 2.8% figure circulating this week is OBDC's own Q2 disclosure, not a journalist's reconstructed default rate. The 8-K/release: "Investments on non-accrual represented 2.8% and 0.8% of the portfolio at cost and fair value, respectively, as compared to 2.0% and 1.0% as of March 31, 2026." New commitments $319M vs sales/repayments $747M — net runoff. 229 portfolio companies, $15.0B fair value. Marc Lipschultz (co-CEO, on the manager call, via InvestmentNews): "across our direct lending strategy, credit health remains strong" / "no meaningful change in our watchlist compared with a year ago." The 2.8% and the "strong" quote are the same firm's, one week. Fair-value non-accrual (0.8%) is the number management wants used; cost-basis (2.8%) is the number the WSJ used. Both are true; they answer different questions (mark vs contractual).

Theme 2 — Two-speed market is now quantified, not anecdotal

The July hypothesis's step 1 (redemptions exceeding liquidity design) is still live and has a Q2 print (14%). Step 2 (credit deteriorating) is stronger: we now have a named BDC's own 2.8% and a Fitch 6%, not just KBRA DLD. Step 4 (manager fee-stream) is not confirmed — ARES just raised the largest ABF fund in history, which is a fee-tailwind in a different sleeve. The honest read is barbell inside the manager: retail perpetual-BDC fee base gated/shrinking, institutional specialty-finance fee base growing. Do not flatten OWL/ARES/BX into a single short.

Theme 3 — Warsh / Jackson Hole is the rate overlay, 10 days out

InvestingLive (2026-08, on the 2026-08-27–29 Jackson Hole symposium): Warsh delivers his first keynote as chair on 2026-08-28; after the July 29 hold (fed funds 3.50–3.75%) he said the speech will "frame big-picture questions rather than offer near-term guidance" and that the Fed is "not constrained by market prices." Three regional presidents dissented for hikes in July; ~half of participants penciled 2026 hikes at the June meeting. September 16 hike/hold odds "close to even."

For this cluster: no cut → floating-rate middle-market borrowers stay stressed (Fitch 6% has room to rise); a cut → BDC NAV relief and a possible retail-redemption pause. The warsh-higher-for-longer-to-brokerage-nii-rerate SCHW/IBKR chain and this PC-stress chain are now the same dated event with opposite signs. That is useful, not a conflict.

Software-as-next-sleeve remains unverified as a realized default cluster. Healthcare and oil-exposed industrials are the current ones (WSJ via FXStreet; Loparex named). Do not promote the SaaS-collateral cross-link from "the wiki can see it" to "it has printed."

Implications for the wiki

• Update bdc-redemption-spiral-to-private-credit-repricing: Q2 14% redemption, OBDC 2.8% primary, Fitch 6%, two-speed fundraising. Hold hypothesis. The software cross-link stays the un-closed gap.
• Update fomc-private-credit-outflows-alt-managers: OWL-specific 2.8% is now first-party; barbell (retail vs institutional/ABF) is the refinement, not OWL-idiosyncratic.
• Dated catalyst: 2026-08-28 Jackson Hole and 2026-09-16 FOMC.
• No new mechanism. No graduation. Thin vertical (financials) — this is a refresh of existing hypotheses, not a tenth AI-infra chain.

Provenance

• Fetched: OBDC Q2 2026 PR (primary).
• Fetched: OBDC Q2 earnings presentation PDF (primary, *.gov-equivalent IR).
• Fetched: InvestmentNews 2026-08-10 (WSJ recap + With Intelligence + Boston Fed PIK).
• Fetched: FXStreet / Money Metals 2026-08-17 (WSJ recap; Fitch 6%; sector mix). Treat as secondary — author is a gold dealer, not a credit analyst; numbers traced to WSJ/Fitch.
• Fetched: InvestingLive Jackson Hole / Warsh (search hit).
• Not fetched: the WSJ original (paywalled). Numbers used only where a primary (OBDC IR) or a named data vendor (Fitch, With Intelligence, Boston Fed) stands behind them.
• Grokipedia skipped — fast-moving credit prints, not encyclopedic.
• X pass not enabled.

Open questions

• Has software actually started defaulting, or is it still a feared sleeve? A single named software PC default in Q3 would convert the cross-link.
• Pathfinder III / ABF: is the institutional bid a quality rotation (out of sponsor direct lending, into asset-backed) or late-cycle reach-for-yield?
• Does the 2026-08-28 Warsh speech move September hike odds, or is it as empty as he advertised?