Private credit under zero-trust — "mark-to-myth," and who holds the bag

Web-verified 2026-06-09. Structured + edges + sources: macro-private-credit-marks.json. The third self-marked asset class, completing the pattern with bank HTM (cost) and AI stakes (fair-value). Overlay; not used in the proofs.

The self-marked NAV — the BDC price-NAV gap is fact; "inflated" is the strong reading

• Private credit ~doubled to ~$2T in five years; loans are manager-valued, illiquid, not exchange-priced — the manager earns fees on the mark it sets.
• 2025 divergence: the public S&P BDC Index fell ~13% while private (non-traded) BDCs reported ~+8–10% — either health deteriorated or NAVs are inflated/lagged (both adverse).
• Nov 2025: the SDNY flagged whether advisers "cherry-pick prices" for higher fees as an enforcement focus.
• Counter-view recorded: Apollo's Marc Rowan rejects the systemic-risk framing.

The marks got tested in 2025 — and failed instantly

• First Brands (auto parts): marked ≥100¢ months before bankruptcy → ~33¢ now; pitched as a $6B loan with ~$1B cash; ~$2.3B receivables "missing."
• Tricolor (subprime auto): AAA ABS at par before bankruptcy → lower bonds 12¢; ~$1B default.
• Same lesson as bank HTM and AI marks: a chosen value held at par until the forcing event, then a near-instant repricing to a fraction.

The quality is deteriorating — STRONG

• PIK (interest deferred onto principal) ~8.8% of investment income; "bad PIK" ~6.4% of loans (Q4-2025) — ~3× 2021.
• ~40% of borrowers have negative free cash flow (vs ~25% in 2021); US default rate ~5.8% (Jan 2026); loan-term-change requests 2.5% → 6.4% YoY.
• The FSB issued a private-credit vulnerabilities report (May 2026).

Who holds the bag (the decisive part)

The risk is migrating off banks and sophisticated funds and into the least-able-to-bear holders:

• Insurance–PE nexus: Apollo's Athene (>$300B, growing direct-lending/ABF) and KKR's Global Atlantic — AM Best finds ~⅕ of their investments are loans to affiliated funds (related-party). Insurance-linked capital put ~$180B into private credit in 2025 (up from $120B in 2023), funding annuities with illiquid manager-marked credit.
• Retailization + 401(k): interval/tender/BDC wrappers >$400B (end-2024), doubling by 2029; the Trump Executive Order (Aug 7 2025) opens 401(k)s to private credit/equity/crypto. Critics (Better Markets) call it pushing opaquely-marked, illiquid risk onto Main Street.

Takeaway

The same private credit that is the marginal lender to the AI buildout (PIMCO/Apollo/Blue Owl — Oracle's $72B, CoreWeave, Meta's $27B Hyperion) is self-marked, deteriorating, and migrating into annuities and retirement accounts valued at marks the holders can't independently check. This is the corpus's answer to who ultimately holds the bag.

Where this sits in the larger pattern (updated 2026-06-11)

Manager-set private-credit NAVs are the third of five self-marked numbers the project now formalizes: bank securities at HTM cost (macro-bank-htm-marks), AI stakes at fair-value / equity marks (fin-google-amazon-anthropic-meta), private-credit NAVs (here), insurance liabilities at offshore-captive marks (spec-insurance-bermuda), and — newest — AI-compute depreciation carried on a chosen useful life (fin-ai-depreciation-debttrap). The Z3 self-marked-value theorem proves they are one defect whose gaps correlate under a common factor (no diversification); the cross-sectional analysis (macro-cross-sectional-analysis) then measures that factor — ~91% of US credit-spread variance is a single principal component, exactly the "no netting across the marked classes" the theorem predicts. Private credit is not a diversifier of the AI / bank / insurance risk; it is the same trade wearing a different label.

← Research index · structured data: macro-private-credit-marks.json · macro-private-credit-marks.md