AI is destroying the collateral beneath three trillion dollars in private credit. The loans are marked-to-model. The losses exist but aren't recognized. Ninety-four percent of downgrades are distressed exchanges — borrowers renegotiating terms they can no longer meet.
The Cockroach covered the symptom — six major funds hitting redemption gates in a single quarter, over ten billion dollars in withdrawal requests, managers paying out less than half of what investors demanded. That is a liquidity story. This is a credit story. The gates are closing because what sits behind them is deteriorating — and the accounting framework that governs private credit was designed to make deterioration invisible until it becomes undeniable.
The Collateral
Software companies represent roughly twenty-five percent of all private credit direct lending. For some managers, the concentration is far higher — Blue Owl's exposure reaches fifty-five percent in certain fund segments. Including broader technology and business services, industry-wide exposure approaches forty percent. These loans were underwritten on assumptions about recurring revenue, high gross margins, and low churn — the defining characteristics of enterprise SaaS.
Generative AI is systematically dismantling each of those assumptions. When AI can generate, modify, and maintain code at a fraction of the cost of human engineering teams, the competitive moat around mid-market SaaS products narrows. When AI agents can replace the workflows that SaaS products automate, the addressable market contracts. When enterprise buyers can build custom tooling with AI in days rather than evaluating vendors for months, the switching costs that justified premium pricing evaporate.
Morgan Stanley warned in March that default rates in private credit could reach eight percent, with pressure concentrated in AI-vulnerable software sectors. CNBC reported on March 25 that AI disruption is now explicitly linked to private credit defaults — not as a theoretical risk but as a current driver of portfolio stress. Blackstone's flagship BCRED fund marked down its position in Medallia, a SaaS company, as an early example of the pattern.
The loans backing these funds are not defaulting in the traditional sense. The borrowers are not missing payments. They are renegotiating terms — deferring interest, converting cash interest to payment-in-kind, extending maturities. This is the shadow default: the economic substance of distress without the legal recognition of failure.
The Opacity
Private credit loans are marked-to-model, not marked-to-market. There is no public exchange, no daily price, no bid-ask spread that forces a fund manager to acknowledge when a loan has lost value. The manager estimates the value using internal models — assumptions about discount rates, recovery rates, and borrower performance — and reports that estimate as the net asset value.
This is not fraud. It is the agreed-upon convention of the asset class. But it creates a structural asymmetry: losses accumulate in reality while remaining invisible in the accounting. A SaaS company whose product is being commoditized by AI may still be making interest payments — for now — while its enterprise value declines month over month. The loan backing that company sits on the fund's books at par. The gap between the mark and the reality widens in silence.
Jay Clayton — the former SEC chair, now the U.S. Attorney for the Southern District of New York — warned in late 2025 that sketchy marks in private market valuations have drawn prosecutorial attention. The concern is not theoretical. When multiple managers hold the same credit but mark it at different values, the dispersion itself becomes a signal that someone is wrong. Apollo responded by moving toward daily NAV reporting with third-party valuations — an implicit acknowledgment that the current system's opacity is becoming a liability.
The problem is architectural. Mark-to-model works when asset values are stable and managers have no incentive to inflate. Neither condition holds when AI is destroying the competitive position of a quarter of the loan book and fund managers face a reflexive spiral — marking loans down triggers redemptions, which forces asset sales, which triggers more markdowns. The incentive to delay recognition is enormous. The accounting framework permits the delay.
The Distressed Exchange
Morningstar DBRS reported that ninety-four percent of all private credit downgrades to D or SD in the twelve months ending February 2026 were distressed exchanges — seventeen rating downgrades in that period, nearly all of them borrowers renegotiating loan terms rather than outright defaulting. Separately, Fitch calculated a private credit default rate of 5.8 percent trailing twelve months through January 2026, with sixty percent of those defaults involving interest payment deferrals or PIK conversions.
A distressed exchange is the financial equivalent of a patient being kept on life support. The borrower cannot service the debt on its original terms. The lender agrees to modify — lower the interest rate, defer payments, extend the maturity — because the alternative is a formal default that forces a markdown on the books. Both parties benefit from the fiction: the borrower avoids bankruptcy, the lender avoids recognizing the loss.
But the loss exists. A loan that has been restructured to defer interest payments is worth less than a performing loan at par. The PIK toggle — where cash interest is converted to additional debt — means the borrower owes more over time while generating less revenue. The distressed exchange delays the recognition of loss. It does not prevent the loss.
When ninety-four percent of downgrades take this form, it means the private credit market has developed an institutional preference for hidden distress over recognized failure. Every distressed exchange is a shadow default — a credit event in economic substance that does not appear in the default statistics, does not trigger a fund-level markdown, and does not show up in the NAV that investors use to evaluate their holdings.
The Discovery
The gates force discovery. When investors redeem faster than the quarterly caps allow — and every major fund tested its caps in Q1 2026 — the fund manager must choose between selling assets to meet withdrawals or telling investors their money is locked.
Selling assets requires finding a buyer. Finding a buyer requires disclosing the quality of the collateral. Disclosing the quality means acknowledging the marks. This is the moment when the shadow default becomes visible — not because the borrower finally fails, but because the liquidity mechanics of the fund force the manager to price what was previously priced by model.
Blue Owl's permanent halt of redemptions on its OBDC II fund — followed by the liquidation of roughly a third of the portfolio to repay a Goldman Sachs credit facility — is the clearest example. The fund did not gate temporarily and wait for conditions to improve. It stopped redemptions entirely and began selling assets. The liquidation price of those assets will reveal what the marks did not.
Ten billion dollars in Q1 redemption requests across major managers is the market's demand for price discovery. The managers are resisting — honoring less than half of requests on average, citing the need to preserve capital and avoid forced sales. The resistance is rational from the manager's perspective. Forced sales into a buyer's market would crystallize losses that mark-to-model accounting currently obscures.
But the resistance cannot hold indefinitely. Every quarter of gated redemptions builds pressure. Every distressed exchange defers but does not eliminate the underlying deterioration. And the force that is destroying the collateral — AI commoditizing the products and business models of mid-market SaaS — is not cyclical. It does not mean-revert. The competitive advantage that justified these loans at origination is being permanently eroded.
The Same Event
The six hundred and fifty billion dollars flowing into AI infrastructure this year and the three trillion dollars at risk in private credit are not separate stories. They are the same economic event viewed from opposite sides.
The capital building AI infrastructure is building the tools that destroy the competitive position of mid-market SaaS companies. The destruction of those competitive positions erodes the collateral beneath private credit loans. The erosion of collateral triggers distressed exchanges, which hide behind mark-to-model accounting, which delays recognition until redemption gates force discovery.
The AI capex boom is the cause. The private credit crisis is the effect. The shadow default is the transmission mechanism — the space between when the damage occurs and when the accounting acknowledges it.
The Cockroach counted the gates. The Shadow Default names what is behind them.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)