Same Loan, Different Price

One loan. Three fiduciaries. Three prices.

Khoros — the software company formerly known as Lithium Technologies — has debt sitting in at least three publicly reporting business development companies. As its restructuring became imminent, Goldman Sachs BDC marked the paper at 50 cents on the dollar. Hercules Capital marked it at 53. Sixth Street Specialty Lending carried the same exposure at 77. That is a 27-point spread on the same instrument, on the same measurement date, disclosed by three regulated fiduciaries — each supported by a valuation policy, a committee, and third-party review. One quarter earlier, all three had the name in the low 90s.

Khoros is not an outlier. Pluralsight: Golub Capital at 97, Blue Owl at 83 — same day, same loan, a 14-point disagreement on a senior secured position. Auven Therapeutics: Oaktree at par, Barings at 78. Isagenix: three holders clustered at 86–88 while a fourth carried it at 98. Explore each in the instrument above.

These are not footnotes from obscure micro-cap lenders. They are large positions at some of the most sophisticated credit managers in the world, drawn from public Schedule of Investments filings — the most transparent corner of private credit. If this is what the visible slice looks like, the burden of proof sits with anyone claiming the unobserved remainder is cleaner.

This is where the private credit debate should begin. Not at the default-rate table. Here.

Everyone is looking at the wrong metric

The public discourse has been almost entirely a credit story, and the credit story has real data behind it. Fitch put the U.S. private credit default rate at a record 6.0% for the twelve months ended April 2026; its privately monitored ratings portfolio recorded 9.2% for 2025, up from 8.1% the prior year. Moody's estimates that distressed restructurings — debt exchanges and maturity extensions negotiated under duress — made up roughly 65% of all private credit defaults in 2025. Lincoln International's "bad PIK" measure, the share of loans amended mid-life into payment-in-kind because the borrower cannot generate cash, reached 6.4% of its loan universe by Q4 2025, up from 2.5% at the end of 2021 — a shadow default rate roughly triple the headline figure. The Financial Stability Board's first dedicated report on the asset class named valuation opacity a systemic vulnerability. The Southern District's U.S. Attorney was blunter still, warning that mismarking assets to generate fees sits squarely in prosecutors' sights.

All of this matters. But notice what the credit framing quietly assumes: that the marks underneath the reported NAVs are at least directionally correct — that the patient's chart reflects the patient. For anyone who values these instruments for a living, that assumption is precisely what the evidence calls into question. Defaults, PIK ratios, and redemption gates are downstream symptoms. The measurement system itself is the condition.

The paradox that gives the game away

PIMCO's credit research team documented something in 2026 that deserves far more analytical weight than it has received. Across BDC portfolios at year-end 2025, marks on the same instrument held by multiple managers were, on average, about five points apart. Eighty-three percent of shared loans priced within a two-point band — but the remaining 17% showed a distribution notably skewed to the upside. The outliers are not randomly scattered around a consensus. They are systematically optimistic.

Hold that against a second observation from the same body of work: since 2021, price dispersion within individual BDC portfolios has run an order of magnitude below the dispersion in the broadly syndicated loan market, despite comparable underlying credit risk. Through time, each portfolio's marks are remarkably smooth. Across holders, marks on identical paper diverge by five, fourteen, twenty-seven points.

ASC 820 defines fair value as the price a market participant would receive in an orderly transaction at the measurement date. For a Level 3 asset, the standard grants substantial modeling latitude. It does not contemplate two honest, contemporaneous fair value conclusions on the same instrument sitting 27 points apart. The conventional defense — "loans repay at par or they don't" — has genuine force for performing credits: if expected recovery is 98, marks of 97 and 99 are both inside estimation error, and neither manager is wrong. Where the defense collapses is at the tail, and the tail is exactly where PIMCO found the skew. When one fiduciary has cut a name from 93 to 50 and another still carries it at 77, they are not making the same judgment with different rounding. One of them will be vindicated by cash. The other is reporting a number.

Anatomy of a sticky mark

No conspiracy is required to explain any of this. Four interlocking mechanics produce a systematic lag between economic reality and reported NAV.

1. Quarterly cadence with no forced convergence. Valuation committees operate independently, and nothing forces marks toward each other except, eventually, realized cash flows. When a credit deteriorates, one fund moves first and the others lag — and during the lag, the slower funds are reporting NAV on last quarter's assumptions. That is not malfeasance. It is the structure of a market without continuous price discovery.

2. PIK accounting — and the taxonomy that matters. Payment-in-kind is not one thing. Structured PIK is underwritten at origination — common in venture and growth lending, priced into the OID and the stated yield. Amendment PIK — bad PIK — is a loan that was priced as cash-pay and converted mid-life because the cash is not there. Economically, a bad-PIK amendment is a default that has been deferred and capitalized. Yet GAAP accrues the PIK as income: net investment income rises as borrower liquidity deteriorates; management fees charged on gross assets grow on a principal balance inflating with capitalized stress; and BDCs, required to distribute 90% of taxable income, pay cash dividends on income that exists only on paper. The terminal phase is already visible: lenders foreclosed on $24.1 billion of debt in 2025 and another $15.2 billion in the first half of 2026 — $39 billion in eighteen months, nearly triple the preceding three years combined.

3. Discount-rate inertia. DCF-based fair value requires a market-participant discount rate. In practice it is calibrated at origination and nudged thereafter by reference to public credit spreads. As direct-lending spreads compressed through 2025, that calibration mechanically raised model values — at precisely the moment borrower fundamentals in the stressed cohort were weakening. The model rewards the market's crowding while the credit deteriorates underneath it.

4. Liability-management opacity. Lincoln recorded a 13% quarter-over-quarter jump in amendment activity in Q4 2025, including a 14% rise in maturity extensions and covenant holidays and a 31% rise in sponsor infusions. A bilaterally negotiated extension does not, by itself, trigger a revaluation — it preserves the loan's performing status and removes the very trigger that would have forced one. From an exit-price perspective, a borrower that cannot refinance at maturity without concessions has just produced material credit information. In practice, that information frequently never reaches the mark.

The mark is short a put — the part the models miss

Here is the claim I have not seen made squarely elsewhere, and the one I think matters most: the widest mark dispersion is not a disagreement about credit at all. It is a disagreement about whether to price the credit agreement.

Consider Pluralsight, the watershed case. Facing an unserviceable debt load, the sponsor executed a drop-down: crown-jewel intellectual property was moved out of the lenders' restricted collateral pool into an unrestricted subsidiary, new sponsor money came in secured by the transferred IP, and the proceeds paid the direct lenders their cash interest. The lenders received a coupon and lost their collateral. When the comprehensive restructuring came, roughly $1.3 billion of funded debt was written down, and recoveries were impaired by exactly the assets that had walked out the door.

A standard yield-based DCF prices a senior secured loan off probability of default and loss-given-default against a stable collateral pool under absolute priority. A permissive document breaks that math. If the sponsor holds the legal capacity to strip collateral and prime the lender, the lender is short a put option — and a conventional DCF assigns that put a value of exactly zero. The documentation data says this optionality is pervasive: so-called J.Crew blockers appear in roughly 23% of broadly syndicated loans but only about 2% of private credit software loans.

For valuation practice, the implication is a genuine methodological shift: legal optionality must become an explicit input, not a qualitative footnote. Where drop-down capacity, non-pro-rata amendment thresholds, or unrestricted-subsidiary basket capacity exist, the fair value of the debt should reflect the deducted value of the sponsor's priming option — an option-pricing overlay on the DCF, not a spread tweak. Two loans with identical cash flows and different documents are not the same asset, and they should not carry the same mark.

Where it concentrates: software and the LTV cliff

The stress is not evenly distributed. Software is the single largest sector exposure in private credit, and software underwriting leans on enterprise-value multiples and loan-to-value rather than fixed-charge coverage — the sponsor's equity cushion is the credit thesis. That makes these loans uniquely sensitive to something credit models handle badly: instantaneous multiple compression. AI disruption does not need to destroy a software company's revenue this quarter to destroy its forward multiple this quarter.

The cushion is the mark. Past 60% LTV, a software loan is no longer a yield instrument; it is the fulcrum security of a capital structure whose terminal value is genuinely uncertain, and it should be valued like one — continuously stress-tested against multiple compression, not carried on trailing recurring revenue.

The market has already voted

While the models deliberate, observable prices have been accumulating — and they form a ladder.

At the top sits stated NAV: 100 cents of reported value. One rung down sits the public market's clearing price for the same asset pools: the median listed BDC traded at roughly 0.74x NAV at the end of March 2026, the widest discount since October 2020. A rung below that sit the opportunists: when Saba Capital and Cox Capital tendered for shares of Blue Owl's non-traded BDCs, they priced at 20–35% below stated NAV — and Blue Owl's own attempted NAV-for-NAV merger collapsed precisely because shareholders refused to accept stated NAV as the unit of exchange. Notably, the board's defense leaned on a banker's opinion that the tender price was inadequate — an argument about the offer, not an affirmation that the NAV was correct. Those are different claims.

The liquidity data completes the picture. Investors sought roughly $20 billion of redemptions from direct-lending vehicles in Q1 2026; about 53% was honored; nine funds enforced their caps for the first time; and for the first time, perpetual non-traded BDC outflows exceeded inflows. Every investor who redeemed at NAV in that window received a price the public market simultaneously judged roughly 25% too generous — and every investor who stayed absorbed the difference. That is not an abstraction. It is a wealth transfer running through the valuation function.

None of this proves the private marks are wrong. Public BDC discounts embed leverage, fees, sentiment, and liquidity. But under ASC 820's calibration logic and Rule 2a-5, a persistent, observable 25-point gap between stated value and every price at which the market actually transacts is data — data that belongs in the valuation file, documented and reconciled, not silently set aside as "structural."

The steelman — and the bifurcation it reveals

Take the optimists seriously, because their evidence is real. The Cliffwater Direct Lending Index — more than 20,000 loans, roughly $550 billion — showed realized losses of 65 basis points at year-end 2025, below its long-term average of 100. Fitch reports that even amid record defaults, 2025 first-lien outcomes were mostly full paydowns, with remaining recoveries estimated at 70–90 cents. Houlihan Lokey's performing-credit index sat at a weighted average price of 98.65; Lincoln's at 99.0. And Moody's, decomposing 6,387 positions across 34 BDCs, showed that for loans originated at a discount, roughly three-quarters of the gap to par is entry basis — OID and fees priced at close — not post-origination deterioration. Much of what looks cheap is performing exactly as underwritten.

I accept all of it. But note what Moody's sharpest finding actually implies: post-origination deterioration concentrates in the near-par book, where loans entered with no embedded cushion — which is precisely where the dispersion cases live. Khoros entered near par. Pluralsight entered near par. The Moody's framework does not dismiss the alarm on these names; it aims the alarm directly at them.

The honest synthesis is bifurcation. The performing core — the large majority by count and value — is real, and the realized-loss data describes it accurately. Alongside it sits a stressed tail that different holders are recognizing at very different speeds, connected to the core by nothing except the averaging that hides it. The Cliffwater mean is not lying. The 27-point spread is not lying. Aggregate statistics are simply the wrong instrument for a distribution this skewed — and every consequential decision in this market right now, from redemption pricing to enforcement risk, lives in the tail.

What would actually fix it

Not daily marks. PIMCO is right that daily pricing is not price discovery — a manager who carries a deteriorated loan at 95 quarterly will carry it at 95 daily. The cadence is not the problem; the assumptions are. Four changes would be:

Cross-holder calibration disclosure. For any borrower held by three or more reporting vehicles, disclose the range of marks; where the spread exceeds ten points, each holder documents why its mark departs from the median. Convergence is not forced. Accountability for the outliers is.

Entry-adjusted reporting as standard disclosure. Moody's entry-basis decomposition should be a filing exhibit, not proprietary analysis. Investors deserve to know how much of a portfolio's discount was priced at close versus earned through deterioration.

LME-triggered independent review. A maturity extension, covenant holiday, or sponsor infusion is credit-relevant information by definition. Each such event should trigger a mandatory independent re-mark — not merely a modification-accounting memo. Lincoln's 31% quarterly jump in sponsor infusions is a rough census of how many of these events are currently flowing past the valuation function untouched.

Price the document. Blockers, drop-down capacity, and non-pro-rata consent thresholds become explicit, scored valuation inputs. The sponsor's priming put gets a price, and that price is deducted from the mark. This is the methodological frontier, and the firms that build it first will define it.

The alternative to voluntary reform is arriving on its own schedule. The SEC and DOJ have made mismarking a stated priority; the ratings infrastructure is under active investigation; and the Bank of England's system-wide exercise — reporting in 2027 — is mapping exactly how valuation opacity propagates through the system. The window in which the industry can fix its measurement problem itself, rather than have a fix imposed, is open. It is not open indefinitely.

The independent vantage point

I have no fund to defend, no book to talk, and no portfolio of marks to protect across quarters. That is not modesty — it is the structural condition that makes an independent opinion worth anything. Across twelve years and more than ten thousand valuation reports, the constant is this: the exit-price standard is not optional, the measurement date is not negotiable, and the question "what would a market participant pay for this today" is most uncomfortable exactly when the answer is lower than last quarter's mark.

Most of what the industry calls independent valuation is third-party concurrence — a review of the manager's model, run on the manager's assumptions, anchored to the manager's prior mark. A valuation committee that meets quarterly to bless a DCF with an origination-era discount rate is fully compliant with ASC 820 and Rule 2a-5, and structurally incapable of producing the convergence the exit-price standard assumes. Compliance was never the same thing as measurement.