Secondary market buyers of private credit fund interests are assigning discounts above the stated NAVs of funds that imposed quarterly withdrawal caps in March and April 2026. That spread is the market’s judgment that the marks have not fully captured the probability of future credit deterioration in portfolios concentrated in AI-vulnerable software borrowers.
The Marks-to-Market Problem
Private credit funds do not mark assets to market on a continuous basis. Valuations are updated quarterly at most, typically using discounted cash flow models informed by the fund manager’s assessment of borrower performance. In a period when borrowers are still current on interest payments—which all three gated funds have indicated is the case—those marks are unlikely to move materially. The models are backward-looking. The AI-displacement risk that LP bases are pricing is forward-looking.
The result is a structural disconnect. LPs price the risk they can see forming. Marks reflect the performance that has already been realized. Secondary buyers bridge the gap by transacting at discounts that reflect where they expect marks to eventually land. In a normal credit cycle, that discount narrows over time as marks catch up. In a scenario where the revenue stress is still developing, the discount can widen for extended periods before it resolves.
The Architecture Behind the Exposure
CEPR co-director Eileen Appelbaum traced the structural chain in April 2026. PE firms acquired life-insurance and annuity businesses over the prior seven years, using policyholder reserves as a low-cost, long-duration funding base for proprietary private credit funds. The credit funds deployed into PE-owned portfolio companies, concentrating in mid-market application software businesses during the 2022–2024 period. The leverage multiples—typically six to eight times EBITDA—made sense when the borrowers’ subscription revenue looked predictable. The AI-disruption question is whether that revenue remains predictable through 2028.
The fund disclosure structure does not help answer that question. Software appears as a sector allocation in most fund letters. The breakdown between infrastructure software (which AI disrupts more slowly) and horizontal application software (which faces near-term AI substitution risk) is not reported in standard format. LPs cannot measure their AI-displacement exposure from the available data.
What Three Gates Tell the Market
Two perpetual private credit vehicles capped quarterly withdrawals in March 2026. A third followed in April. The sequence is its own signal: three gates in six weeks at major perpetual vehicles is an unusual concentration of liquidity management action. None of the funds disclosed material credit losses alongside the announcements. Each gate announcement generated secondary market repricing and a widening of discounts. Each new gate triggered additional redemption activity from LPs who had been watching.
The Two-Category Breakdown That Matters
Private credit’s AI-displacement risk is not uniformly distributed. Portfolios that lent to horizontal application software—enterprise productivity, project management, CRM, document automation—carry the highest concentration of AI-substitutable borrowers. These are categories where LLM-based tools are already technically capable of partial substitution, and where enterprise buyers’ switching calculations are changing in real time.
Portfolios that concentrated in infrastructure software, deep-vertical SaaS with regulatory lock-in, or physical-asset-backed lending outside of software are materially less exposed. The aggregate statistics on private credit AI exposure obscure this differentiation, which is one reason the LP community cannot make portfolio-level decisions without doing primary research that the funds should be providing.
The structural arguments from fund managers—covenant quality, private workouts, direct-lender advantages over public high-yield—are accurately described. They remain arguments about process. The outcome question—how many software borrowers experience revenue declines large enough to trigger covenant events, and how quickly—depends on enterprise software revenue trends through 2028. Those trends are not yet known, and they are not disclosed by the fund structures. NAV prints over the next two quarters are the first externally observable data point in that sequence.
Source: Private Credit Fund Redemptions Climb Sharply, Some Caps Now in Place
Daniel