The Software Lending Cohort That Built Private Credit’s AI Problem

Private credit’s AI-displacement problem has a vintage. The loans that are generating the most LP concern are those extended between 2022 and 2024, when mid-market application software companies could support six-to-eight-times EBITDA leverage multiples on the strength of subscription revenue growth that appeared durable. The AI environment of 2026 is testing whether that durability holds.

What the 2022–2024 Vintage Looked Like at Underwriting

Enterprise software was the consensus safe harbor for private credit underwriters in 2022. Subscription revenues were highly visible, churn rates were low, and the SaaS business model had proven resilient through the 2020 COVID disruption. Private credit funds—many of them backed by PE-controlled insurance capital—deployed aggressively into the category. Mid-market software companies owned by PE sponsors borrowed at multiples that would have been difficult to justify in more cyclical industries. The covenants were set against the assumption that software revenue would compound at 15% to 25% annual rates.

None of that underwriting modeled a scenario where generative AI tools, available at decreasing marginal cost, allowed enterprise buyers to build or replace software functionality without purchasing additional seats. The risk category did not exist as a modeled scenario in most 2022-era credit committees.

The CEPR Structural Analysis

Eileen Appelbaum of the Center for Economic and Policy Research published a structural analysis in April 2026 situating the software lending exposure within the broader architecture of PE-controlled insurance capital. Over the prior seven years, major PE firms had acquired life-insurance and annuity businesses, gaining access to stable, long-duration policyholder reserves. Those reserves funded private credit vehicles with limited disclosure and no mark-to-market discipline. The credit vehicles lent the reserves—through multiple intermediary layers—to the same PE-owned software companies whose revenue assumptions the AI environment is now testing.

The structural concern is that the disclosure opacity insulates the chain from the feedback that would typically prompt mark-downs: the fund marks infrequently, the insurance entity operates outside standard investment-grade guidelines, and the policyholder at the far end of the chain has no direct line to the credit quality of the underlying loans.

Three Gates, No Confirmed Losses

Two of the largest perpetual private credit vehicles imposed quarterly outflow caps in March 2026. A third followed in April. All three communicated the caps as orderly liquidity management. None disclosed material credit impairment. Secondary buyers of fund interests assigned discounts above stated NAVs—pricing the probability that the marks will eventually reflect the AI-displacement risk that the LP base is worried about.

The secondary discount widens with each gate announcement, because each announcement adds information about the direction and pace of LP sentiment. The market’s best current estimate of where marks will eventually land sits in those secondary transaction prices, not in the quarterly NAVs that fund managers have published.

Portfolios That Built in Defenses Before 2022

Not every private credit book carries the 2022–2024 software lending profile. Funds that had already moved toward infrastructure software, vertical SaaS with regulatory and workflow lock-in, and physical-asset collateral before the peak lending period are in a materially different position. These portfolios still face LP questions about AI risk, but the underlying borrower characteristics—higher switching costs, more regulatory dependency, physical collateral backing—provide genuine credit defensibility that horizontal application software borrowers do not have to the same degree.

The structural arguments from the manager community—tighter covenants than public bonds, private workout mechanics, direct lender relationships—describe process advantages that are real. They do not determine the credit outcome. That outcome depends on how many 2022–2024 vintage software borrowers experience AI-driven revenue pressure severe enough to trigger covenant events, and on what timeline. NAV prints over the next two quarters and any shift toward AI-risk disclosure in LP letters will be the first externally observable data points in that determination.

Source: Private Credit Fund Redemptions Climb Sharply, Some Caps Now in Place