NICK'S DATA:

58 amendment PIKs out of 189 loans parsed — roughly one-third of the portfolio by count. Cliffwater reports less than 1% by asset value. The discrepancy comes from how they define and bucket amendment PIK versus original loan balances. They also appear to average across the ~one-third of the book sitting in CLOs or PIVs they claim not to control.

TAKEAWAY:

A former Lincoln professional who has valued these exact portfolios — at the firm Cliffwater pays to mark its book — pushed back on every data point. Not because the data was wrong, but because the framework for how private credit “should” be marked doesn't use market prices. The defense is the methodology itself. If fundamentals haven't changed, the mark doesn't move. The question is whether fundamentals have changed and nobody is looking.

Cliffwater pays Lincoln International to produce third-party fair value marks on its portfolio. This is the same Lincoln International the source worked at. The same firm that touches every major BDC.

The fund hires the valuation agent. The valuation agent marks the book. The fund reports the marks to investors. The fund's NAV — which determines management fees, performance reporting, and redemption prices — is a direct output of those marks.

This is not a conflict of interest in the legal sense — third-party valuation is industry standard and required. But the incentive alignment is one-directional: no valuation firm has ever been fired for marking a book too high. Plenty have been fired for marking it too low.

The source's defense on every single data point was the same: “Has anything materially changed since underwriting? No. So the mark doesn't move.”

This is the core mechanic. Private credit is not marked to market — it is marked to model. The model says: if the borrower is current on payments and the original underwriting thesis hasn't been formally invalidated, the loan stays near par. Spreads can blow out. Comparable public bonds can trade at 85 cents. JP Morgan can walk away from a $6.3 billion deal in the same sector. The mark doesn't move.

The methodology itself becomes the shield. You can't argue the mark is wrong because the mark isn't trying to reflect what the asset would sell for. It's reflecting what the valuation model says the asset is “worth” based on inputs the fund influences.

The source confirmed Cliffwater has zero non-accruals and only amends and extends. When a borrower can't pay, the loan gets restructured — maturity pushed out, interest converted to PIK, covenants loosened. The borrower is now “current” again. The mark doesn't move because — per Step 2 — nothing has “materially changed.”

But the amendment is the material change. Converting cash interest to PIK means the borrower couldn't pay. Extending maturity means the original timeline failed. Loosening covenants means the business breached its commitments. Each of these is a credit event. None of them are treated as one.

Result: 58 amendment PIKs out of 189 loans. One-third of the portfolio by count has been restructured. Reported PIK exposure: less than 1% of asset value. The gap between those two numbers is the sausage.

The transcript reveals a specific mechanic: take a $100 million loan, amend it to PIK, then report only the accrued PIK interest (~$25 million) as “in PIK.” The original $100 million loan balance? Not counted.

This is how you get from 58 amendment PIKs (one-third of the book by count) to “less than 1% of AUM in PIK” (the number the source was given). The definition of what counts as “in PIK” excludes the loan principal — only the rolled interest accrual is bucketed.

They also appear to average across the ~one-third of the book sitting in CLOs or coordinated PIVs they claim not to control — diluting the numerator by inflating the denominator with assets that are structurally excluded from the distress metrics.

The source walked through how loans actually get made. A lead underwriter like Ares commits$300 million of a $500 million deal. They keep some on balance sheet, put some in their CLO, then sell off pieces at an OID of97 cents. They can flip at 97.5–98 and pocket the day-one arbitrage.

Cliffwater buys those freshly funded loans in a group and packages them into a CLO wrapper. Their name goes on the credit agreement. They fund capital day one alongside the BDC.

This means Cliffwater's portfolio isn't independently sourced — it's built from the same loan pipeline as every other BDC, purchased from the same arrangers, at the same terms, into the same sectors. The diversification is nominal. The correlation to the BDC complex is structural.

The source clarified something important: Cliffwater's CLOs are not CLO equity. They own the entire tranche — no A/B/Z sleeves, no waterfall. The CLO wrapper exists for three reasons:

• 1. Cheaper leverage — borrow at SOFR + 125bps against CLO assets vs. unsecured balance sheet debt
• 2. Compliance — structured vehicle satisfies regulatory requirements for certain co-investment partners
• 3. Isolation — if the CLO takes losses, it's ring-fenced from the rest of the fund

Effective leverage: ~2x (50% equity in the structure). Not the 10-12x some have claimed. But this creates its own question:

If the CLO is just a funding wrapper and Cliffwater owns everything inside it, why does ~one-third of the book in CLOs/PIVs get treated as a separate sleeve they “don't control” when reporting distress metrics? You can't use the CLO for cheaper leverage and use it as a shield against reporting what's inside.

The most revealing moment in the conversation: the source's repeated insistence that “these are not public bonds — this is not how lenders work.”

This is the firewall. Public credit trades and reprices in real time. Private credit doesn't trade — so it doesn't reprice. The source's position is that this is afeature, not a bug. You underwrote the loan at par. The borrower is paying. The equity cushion was 60% at origination. The mark stays.

But the equity cushion was calculated using sponsor-marked enterprise values at peak-cycle EBITDA multiples. If the EV was inflated at origination, the cushion was always thinner than reported. And when the source says “nothing has materially changed,” it's because the inputs to the model — the same sponsor-influenced inputs — haven't been updated.

Marks justifying marks. The entire chain is self-referential. Public markets say 88 cents. The model says 100. The model wins because the model is the methodology, and the methodology is the product.

When Nick brought up JP Morgan walking on the $6.3 billion Qualtrics deal, the source said: “Deals get walked out on all the time.”

That's the point. If the largest arrangers are walking away from software deals, the deals that do close are the survivors. The spread data only reflects surviving deals. The marks only reflect loans that were funded. The loans that were walked away from — the ones where risk was too high even for the arrangers — don't appear in any dataset.

So when the source says “software spreads only moved 20 bps more than broad market,” that 20 bps is measured on the deals that closed. The real risk repricing is in the deals that didn't. And Cliffwater's existing software book — ~20% of the portfolio — was underwritten before the walking started.

Here is the thing people miss about private credit marks: they are not based on what the asset would sell for. They are not based on where a comparable loan trades. They are based on the assumptions that existed when the loan was originated — adjusted only for changes in credit spreads, not actual transaction prices.

The valuation model takes the original underwriting spread, compares it to current benchmark spreads, and adjusts the mark. If spreads widen 30 bps, the mark moves a few cents. If spreads don't move, the mark doesn't move — regardless of what is happening inside the business, inside the sector, or inside the borrower's ability to service the debt.

There is no price discovery. There is no bid. There is no offer. There is no market clearing mechanism. The “mark” is an output of a model whose primary input is the spread environment — and even that input is applied with lag and discretion.

This is why Cliffwater can mark up in April 2025 when the leveraged loan index marks down. The model says: spreads didn't move enough in my specific bucket. The market says: risk just repriced across the entire credit complex. The model wins.

ASC 820 — the GAAP standard for fair value measurement — classifies assets into three levels. Level 1: quoted market prices. Level 2: observable inputs (comparable trades, broker quotes). Level 3: unobservable inputs (models, assumptions, management estimates).

Virtually the entire Cliffwater portfolio is Level 3. There are no quoted prices. There are no comparable trades in sufficient volume. The “fair value” is whatever the model says, using inputs that the fund and its valuation agent select.

GAAP was not designed for a $3.5 trillion asset class that doesn't trade. The standard assumes that Level 3 is the exception — a small bucket of hard-to-value assets in an otherwise observable portfolio. In private credit, Level 3 is the portfolio. The exception became the rule, and GAAP has no mechanism to challenge it.

The auditor signs off on the methodology, not the marks. PCAOB reviews have found deficiencies in Level 3 fair value audits — including at Cliffwater's own auditor — but deficiency findings don't change the marks. The accounting framework is structurally incapable of catching what's happening because it was never built for a world where the entire book is unobservable.

Notice what happened in this conversation. The source — someone who has been inside the valuation process — was given a number: “less than 1% in PIK.” He believed it. He had no reason not to. That was the number Cliffwater provided.

It took Nick parsing every single SEC filing, line by line, to find the 58 amendment PIKs out of 189 loans. It took building a database from N-PORT and N-CSR filings to surface the 53 unreported non-accruals. None of this was in the fact sheet. None of it was in the marketing materials. None of it was volunteered.

This is the granularity problem. Interval funds report in aggregate. They give you a NAV, a return, a yield, and a one-page fact sheet. The underlying loan-level data is buried in SEC schedules of investments that run hundreds of pages. The format is designed for compliance, not comprehension.

Investors don't get position-level marks. They don't get amendment histories. They don't get PIK conversion dates. They get: “less than 1%.” And unless someone builds a forensic parser and goes loan by loan, that number is never challenged. The opacity is not a side effect — it is the product.

The source treated the spread environment as evidence that the system is working. Loans are paying SOFR + 500 to SOFR + 700. Borrowers are current. Coupons are flowing. What's the problem?

The problem: at SOFR 3.6%, S+700 is an all-in cost of capital of 10.6%. S+500 is still 8.6%. These are not investment-grade corporations. These are middle-market companies — software firms, healthcare services, business services — with $50-250M of EBITDA, often loaded with 5-7x leverage from the sponsor's LBO.

A company paying 9-11% on its debt needs to generate unlevered returns well above that just to service the interest. For the vast majority of these businesses, that math does not work. A software company growing 5-10% with 70% gross margins but 6x leverage is spending more than half its free cash flow on interest alone. A services business at 40% FCF conversion is underwater. The cost of capital cannot be unlevered on most of these businesses.

When SOFR was near zero, S+600 meant a 6% coupon. Manageable. When SOFR went to 500+ bps at peak, that same S+600 became an 11%+ coupon. Even now at 3.6%, S+600 is still 9.6% — a cost of capital that most middle-market companies cannot earn through on an unlevered basis. The contractual spread didn't change. The economic burden nearly doubled from the zero-rate era. And the marks didn't move because — per the methodology — “nothing has materially changed.”

The coupons are flowing today because companies are choosing between paying cash interest and dying. Many are converting to PIK — which Cliffwater redefines as less than 1%. Others are drawing on revolvers. Others are selling assets. The coupon is current right up until the moment it isn't. And when it stops, the amend-and-extend machine catches it before the mark does.

The source tried to describe Cliffwater's CLO structure as something nuanced — different from ARCC's equity tranches, different from traditional CLO waterfalls. But when pressed on what Cliffwater actually owns, the answer kept coming back to the same thing: they own the whole loan.

No A-tranche. No B-tranche. No Z-sleeve. No waterfall. Cliffwater buys the entire tranche and sits on it. That is a unitranche — a single layer of debt with no structural subordination beneath it and no structural protection above it.

This matters because unitranche has a specific risk profile. In ARCC's CLO equity model, defaults hit the equity sleeve at 50 bps but management fees from senior tranches more than offset the loss. There's a buffer. In Cliffwater's unitranche model, every dollar of loss hits the fund dollar for dollar. There is no fee income from senior tranches absorbing the first loss. There is no subordination.

And yet Cliffwater markets itself with a Sharpe ratio above 11 on its enhanced fund. A unitranche portfolio with 2x CLO leverage, no structural protection, 20% software exposure, SOFR + 700 cost of capital to borrowers, and one-third of the book in amendment PIK — and the risk-adjusted return is better than Renaissance Technologies.

The Sharpe ratio is an output of the marks. If the marks don't move, volatility is near zero. If volatility is near zero, any positive return produces an infinite Sharpe. The “11 Sharpe” is not evidence of superior risk management. It is evidence that the marks are not reflecting risk.

Put it all together:

• > The fund hires the valuation firm
• > The valuation firm uses a methodology that doesn't reference market prices
• > Marks are based on origination assumptions and spread changes — not actual prices or bids
• > GAAP Level 3 was designed as the exception — here it is the entire portfolio
• > When a borrower can't pay, the loan gets amended — not marked down
• > PIK is defined narrowly enough to exclude the vast majority of restructured loans
• > No granularity is provided until someone parses every SEC filing by hand
• > One-third of the book sits in structures used to dilute distress reporting
• > Equity cushions are calculated on sponsor-marked EVs at peak multiples
• > Spread movements in public markets are dismissed as irrelevant
• > Survivorship bias in the origination pipeline inflates apparent sector health
• > SOFR + 700 cost of capital is unsustainable for most borrowers — but coupons flow until they don't
• > Practically the entire book is unitranche — no structural protection, every loss hits dollar for dollar

The output: a fund that has never had a losing month in 41 consecutive months. Zero reported non-accruals. Less than 1% PIK by their definition. NAV that goes up when the leveraged loan index goes down.

The source — a person who built these models at Lincoln — didn't dispute the data. He disputed the framework. His position: this is how private credit is supposed to work. The marks aren't wrong because the marks aren't trying to be market prices. They're trying to be fair values under a model that the industry designed, the regulators accepted, and the valuation firms execute.