Private Credit Quietly Became AI's Balance Sheet — And This Week It Started Selling the Risk to You
@dailyanalysts · Friday, July 24, 2026 · Standalone Private Credit / Hidden Debt Deep Dive · Prices verified intraday 12:00 ET (market open)
The one-line thesis: The $1.6–1.7 trillion private credit market has become the off–balance-sheet financing engine for the AI capex the public equity market just started punishing — and this week the biggest private-credit deal ever ($35B for Anthropic/Broadcom) began being sold down into banks, insurers and retail wrappers. That is not de-risking. That is the risk migrating from three originators into the plumbing of the whole system, at the exact moment Fitch's default rate sits at a record 6.0% and the 10-year threatens 5%. My highest-conviction way to play it: own the liquid, senior-secured, first-lien listed BDCs (ARCC, BXSL) into next week's Q2 prints; avoid the gated retail vehicles and impaired external BDCs (FSK, OWL, TCPC). Bifurcation, not blow-up — yet.
0. Scorecard callback — the July 20 private-credit trades
I set this bifurcation book on July 20 ("The Q2 Earnings Referee"). Four sessions later, honest marks (intraday 7/24):
- ARCC (LONG, entry $18.40–19.00) → $18.77. On track, in the zone. ✅
- BXSL (LONG, entry $23.00–23.60) → $23.12. On track, in the zone. ✅
- OBDC (SPEC, entry $10.40–10.90) → $10.78. On track, in the zone. ➖
- FSK (AVOID / fade toward $9.50) → $10.63, down from ~$10.73. Grinding my way. ✅
- OWL (AVOID) → $9.32, still pinned near its lows. Correct to avoid. ✅
What changed this week and I have to flag honestly: On 7/20 the alt-manager stocks (APO, ARES, BX, KKR) were falling on green tape — I read that as the market pricing "fee-stream doubt." Today they are the strongest thing in the complex: BX +2.6% ($127.68), ARES +2.6% ($124.37), KKR +2.0% ($97.87), APO +1.8% ($121.11). The catalyst was Blackstone's 7/23 statement that BCRED redemptions have "materially slowed." So the acute-panic leg has paused. My structural caution stands, but the tape is telling you the market currently believes "growing pains, not crisis." Respect it.
1. The catalyst that makes this urgent NOW
On June 9, Apollo and Blackstone closed a $35 billion financing to buy Broadcom custom AI chips (XPUs) for Anthropic — structured through an off–balance-sheet special-purpose vehicle so the debt never touches Broadcom's balance sheet, split across three tranches, with senior notes carrying Broadcom credit enhancement and a subordinated slice taking the equity-like risk (Bloomberg via Yahoo Finance). It is the largest private credit transaction ever assembled.
The new development — the reason I am writing this today: as of this week, Wall Street banks have begun trading the first pieces of that $35B package to a wider pool of investors (Bloomberg). Roughly $15B of it is being offered out. Read what that means slowly: the "private" credit deal is being un-privatized — distributed into bank balance sheets, insurance general accounts, and non-traded retail BDC/interval-fund wrappers. The three originators who underwrote the concentration risk are now laying it off onto everyone else.
Apollo's co-president Jim Zelter's lieutenants have framed this as Apollo "muscling in on Wall Street's turf with ~$300 billion of loans." Correct. But the flip side of originating $300B is that you eventually have to place $300B. Origination is a fee business; placement is a risk-transfer business. This week we crossed from the first into the second.
2. Why this is a genuinely big deal (the numbers that carry consequences)
| Data point | Level | Why it matters (the consequence) |
|---|
| Fitch U.S. Private Credit Default Rate | 6.0% record (April, held in May) | Up from 5.8% in Jan; FY2025 mid-market hit 9.2%. Morgan Stanley warns direct-lending defaults could reach 8%. Every 100bps of default on a ~$1.6T book is ~$16B of loss working through NAVs. |
| AI debt financed by private credit | >$1 trillion (Oliver Wyman) | Hyperscaler + neocloud capex is projected above $5T by 2030 (Goldman). Private credit is the marginal lender. If AI ROIC disappoints, PC is first-loss. |
| BCRED Q2 redemption requests | ~10% of NAV (vs 7.9% Q1) | Blackstone capped repurchases at the 5% quarterly gate, then said 7/23 the pace "materially slowed" and it will meet 100% of requests. Gate held — this cycle. New equity raised fell to ~$1B in the quarter. |
| 10-Year Treasury yield | ~4.69–4.71% | Highest since Jan 2025. BDC/PC books are floating-rate — high rates flatter income but crush the leveraged borrower. Push toward 5% and coverage ratios crack. This is Burry's "watch the long bonds." |
| PIK income share of BDC investment income | ~8%, non-cash loans at 14-yr peak | Payment-in-kind lets a stressed borrower "pay" interest with more debt instead of cash. Rising PIK + rising non-accruals = the mechanism that hides the true default picture until a mark breaks. |
Michael Burry — freshly vindicated on his TSLA/CAT/NVDA shorts in July's pullback — put it plainly on 7/23: "Watch the long bonds. Not sure how much longer PE and PC can hold their breath" (Benzinga). He is describing the exact collision at the center of this piece: $100-ish oil (now cooled to ~$89 WTI as Iran peace-talk hopes flicker) → sticky inflation → higher-for-longer rates → refinancing pressure on the most leveraged private borrowers, many of them AI infrastructure SPVs.
3. The second- and third-order effects most people are missing
First-order (widely discussed): higher defaults hurt low-quality BDCs and gated retail funds. Fine. Consensus.
Second-order (the risk-transfer channel): The syndication of the Anthropic/Broadcom deal
this week is the tell. Private credit's selling point was that patient, locked-up capital held the risk. But when originators distribute mega-deals into banks and insurers, the risk re-enters the regulated system through the back door. The Chicago Fed and Dallas Fed have both flagged that banks now finance private-credit lenders who finance data centers — a chain where a bank can be two steps removed from an AI SPV and still be exposed. The Financial Stability Board's May 2026
Report on Vulnerabilities in Private Credit names exactly this interconnection and the opacity of marks. Senator Warren has already pushed FSOC to probe the "AI debt bubble."
Third-order (the retail adverse-selection trap): As institutions trim non-traded BDC exposure (net flows across the non-listed complex turned sharply negative — Blackstone alone returned ~$3.7B via BCRED in an earlier quarter), the DOL's move to open 401(k) safe-harbor to private assets is inviting retail in as the marginal buyer of the paper institutions are handing back. When the sophisticated seller and the unsophisticated buyer trade at a mark the seller sets, that is textbook adverse selection. The non-traded wrapper — priced monthly at NAV, gated at 5% — is where I expect the next "mark break," not in the daily-priced listed BDCs.
And a fresh confirmation the credit cycle is turning in real assets, not just corporates: KKR Real Estate Finance Trust is weighing a sale or merger as loan losses mount (7/23). CRE credit is the canary that already sang; private corporate credit is the bigger, later bird.
4. The bifurcation map — who to own, who to avoid
| Ticker | Price (7/24) | Fwd P/E | Yield | Book quality | Call |
|---|
| ARCC (Ares Capital) | $18.77 | 10.4x | 10.2% | Largest BDC, diversified senior-secured, ROE 8.1%, EPS TTM $1.63 | LONG |
| BXSL (Blackstone Secured Lending) | $23.12 | ~8x | ~13% | ~98% first-lien senior secured, beta 0.43 — the "cleanest" book | LONG |
| APO (Apollo) | $121.11 | ~14x | ~1.8% | Origination machine + Athene insurance funding; HSBC just initiated Buy $145 | SPEC LONG |
| OBDC (Blue Owl Capital Corp) | $10.78 | 8.5x | 11.5% | ~25% disc to $14.41 NAV; cut base div 16% in Q1; insiders buying | SPEC |
| FSK (FS KKR) | $10.63 | — | ~15% | EPS TTM −$1.96, ROE −9.4%, elevated non-accruals, prior div cut | AVOID / FADE |
| OWL (Blue Owl Capital — manager) | $9.32 | 182x | 10.0% | Redemption overhang at OCIC/OTIC; UBS note helped trigger investor exodus; PT cuts (BMO $11, Oppy $15) | AVOID |
| TCPC (BlackRock TCP) | ~$3.33 | — | — | NAV cut to $7.07 (~53% disc); DOJ/SDNY valuation probe; class actions | UNINVESTABLE |
5. Actionable trades — levels, targets, invalidation
HIGH CONVICTION LONG ARCC (Ares Capital)
- Entry: $18.30–19.00
- Target: $21.50 (retest of the 52-wk high $23.20 region on a clean Q2)
- Invalidation: weekly close < $17.40 (the March low), or a Q2 non-accrual spike above ~2.5% of fair value
- Timeframe: 1–3 months · Catalyst: Q2 earnings in the coming days
- Why: Two independent signals agree — (1) largest, most-diversified, best-underwritten senior-secured book trading at ~10x forward on a 10%+ covered yield; (2) sector-wide redemption fear easing (BCRED 7/23) removes the forced-seller overhang. This is the "guilt-by-association" name that de-rates with the gated retail vehicles but does not behave like them.
HIGH CONVICTION LONG BXSL (Blackstone Secured Lending)
- Entry: $22.80–23.40
- Target: $26.00
- Invalidation: weekly close < $22.00
- Timeframe: 1–3 months · Catalyst: Q2 print Aug 6
- Why: ~98% first-lien senior secured — the highest-quality book in the listed complex and the lowest-beta (0.43). If the whole theme re-rates up, BXSL captures it with the least downside if it doesn't. The direct read-through from BCRED's slowing redemptions is most positive for the Blackstone-managed listed vehicle.
SPECULATIVE LONG APO (Apollo) — the origination-engine play
- Entry: $112–118 on any pullback (chasing $121 is late)
- Target: $145 (HSBC's fresh initiation PT)
- Invalidation: daily close < $105, or evidence Apollo is stuck holding unsyndicated AI paper
- Timeframe: 1–3 months
- Why + the honest risk: Apollo is the winner of the AI-debt origination super-cycle and has the Athene insurance balance sheet to fund it. But it is a two-sided bet: the same $35B deals that generate fees also concentrate risk on Apollo's own books until placed. This is SPECULATIVE precisely because the syndication now underway is the test. If the paper places cleanly, APO re-rates; if it clogs, APO leads the complex down. Size accordingly — half position.
AVOID / FADE FSK and OWL
- FSK: fade rallies toward $11.00–11.40, target $9.50, invalidation weekly close > $11.80. Negative trailing EPS and elevated non-accruals mean any rate or default surprise hits here first.
- OWL: avoid outright. A manager trading at 182x forward earnings whose flagship non-traded funds are gated and losing assets is the wrong side of the fee-stream trade. Sell-side is cutting PTs, not raising.
- Pair for the risk-averse: long ARCC + BXSL vs short FSK + OWL isolates the bifurcation and hedges out a broad rate move.
6. Three scenarios into the Q2 BDC earnings wall (Aug 4–6)
Earnings calendar: TSLX Aug 4 · OBDC Aug 5 · FSK & BXSL Aug 6 · ARCC in the same window. This cluster is the referee.
Bull — 35%: Q2 prints show non-accruals flat-to-down QoQ at the quality names, NAV stable, dividends covered; BCRED's slowing redemptions confirm the panic peaked. ARCC → $21.50, BXSL → $26, the complex re-rates 10–15%. Trigger: ARCC/BXSL non-accruals < 2% of fair value + no dividend cut.
Base — 45%: Bifurcation confirmed — ARCC/BXSL hold NAV and coverage; FSK/OWL show rising PIK and slipping marks. Quality drifts up modestly, junk drifts down. The pair trade (long ARCC+BXSL / short FSK+OWL) is the cleanest winner. Trigger: mixed prints, PIK share ticking up sector-wide.
Bear — 20%: A surprise mark break — a second dividend cut, a >300bps non-accrual jump, or the syndication of the $35B AI paper stalls and forces wider spreads. 10Y presses 5%, Burry is right, and the whole complex gaps down 15–25% with the gated funds leading. Trigger: any BDC prints a double-digit NAV write-down, or a headline that a major AI-SPV tranche can't be placed.
7. My opinion, stated plainly
Consensus has settled on "growing pains, not crisis," and this week's tape agrees. I think consensus is right about the timing and wrong about the location of the risk. The listed, daily-priced, senior-secured BDCs are cheaper than they should be because they are marked to a market that can't distinguish them from the gated retail vehicles that genuinely deserve the discount. That is the mispricing, and it is my highest-conviction long.
But the syndication of a record AI-infrastructure loan into banks, insurers and retail wrappers — happening right now — is exactly how a "concentrated, contained" problem becomes a distributed, systemic one over the next 6–12 months. Private credit didn't eliminate the risk of financing a speculative capex boom on debt. It relocated it somewhere with fewer daily marks and a 5% monthly exit door. When the AI revenue math is finally tested, the daily-priced BDCs will have already told you the truth; the gated funds will still be pretending. Own the ones that can't lie to you.
8. Bottom line & action steps
- Buy ARCC $18.30–19.00 (HIGH CONV, tgt $21.50, inval weekly <$17.40) into Q2.
- Buy BXSL $22.80–23.40 (HIGH CONV, tgt $26, inval weekly <$22) — cleanest first-lien book.
- Buy APO $112–118 on a dip (SPEC, tgt $145, inval <$105) — origination winner, watch the $35B syndication as the tell.
- Avoid/fade FSK and OWL; TCPC uninvestable. Run long-quality / short-junk as a pair if you want the rate hedge.
- Watch three gauges: (1) does the Anthropic/Broadcom paper place cleanly? (2) 10Y — 5% is the line that breaks the leveraged borrower; (3) Aug 4–6 non-accrual & PIK trends.