Khan Capital | February 2026
Key Takeaways
- Software is private credit’s largest single-sector bet: Outstanding SaaS loans grew from $8 billion in 2015 to over $500 billion by end-2025, representing 19% of total direct lending, and the underwriting assumptions that justified this concentration are being challenged by AI.
- AI strikes at the core of software creditworthiness: Per-seat licensing erosion, collapsing switching costs, and margin compression from AI-native competitors are undermining the recurring revenue, customer stickiness, and pricing power that made software the ideal private credit borrower.
- Back-leverage is contracting: JPMorgan’s decision to mark down software loan collateral and restrict lending against these assets is reducing private credit funds’ borrowing capacity, with $300 billion in bank back-leverage at risk of repricing.
- PIK income is masking deterioration: Over a third of software loan agreements now include PIK options, and PIK income approaching 10% of BDC investment income creates a liquidity mismatch between reported earnings and actual cash available for distributions and redemptions.
- Disruption risk extends beyond software: UBS estimates 25 to 35% of private credit portfolios face elevated AI disruption risk, including business services and professional services firms, with default rates potentially reaching 13% in an aggressive scenario.
For a decade, private credit managers treated software companies as the ideal borrower. Fat recurring-revenue margins, predictable cash flows, low capital expenditure requirements, and sticky customer bases made mid-market SaaS firms a natural fit for the leveraged lending model. By the end of 2025, outstanding loans to software-as-a-service companies had grown from approximately $8 billion in 2015 to over $500 billion, representing 19% of total direct lending. A third of all private credit funds had extended loans to the sector.
The thesis was elegant. Software companies generated 70 to 85% gross margins, enjoyed 90%+ net revenue retention, and operated with minimal cyclicality relative to traditional industrial borrowers. Private credit managers underwrote these loans at 5 to 7 times EBITDA, secured by the very characteristics that made the businesses attractive: the intellectual property, the customer contracts, and the recurring revenue streams that underpinned the valuation.
Then artificial intelligence arrived not as a tool for these companies to deploy, but as a force capable of rendering their business models obsolete. The collateral that secured hundreds of billions in private credit loans is being repriced in real time, and the lending structures built around it are facing a test their originators never modelled.
The $500 Billion Bet on Software Durability
The scale of private credit’s software exposure did not accumulate by accident. It was the product of a deliberate strategic shift during the low-rate era of 2015 to 2022, when managers competed aggressively for assets as capital flooded into the space. Software companies offered what the market valued most: predictability.
According to Bank for International Settlements data published in its March 2026 Quarterly Review, Business Development Companies (BDCs) alone extended over 15% of their loans to SaaS firms in 2025. The broader private credit ecosystem, including direct lending funds, CLOs, and separately managed accounts, pushed total software exposure to an estimated 26% of the market, according to Morgan Stanley analysis.
The underwriting rationale rested on three pillars: the durability of recurring revenue, the defensibility of switching costs, and the assumption that software margins would persist or expand over time. Each of these pillars is now under direct assault.
| Metric | 2015 | 2020 | End-2025 | Trend |
|---|---|---|---|---|
| Outstanding SaaS loans | $8B | ~$120B | $500B+ | 62x growth in a decade |
| SaaS share of direct lending | ~2% | ~10% | 19% | Largest single-sector concentration |
| BDC funds lending to SaaS | ~10% | ~20% | 33% | One-third of all private credit funds |
| Software loans with PIK option | ~10% | ~12% | 33%+ | Threefold increase in three years |
| AI disruption risk exposure | N/A | N/A | 25-35% | UBS estimate across portfolios |
How AI Strikes at the Heart of the Thesis
The AI threat to private credit’s software portfolio is not a general macroeconomic headwind. It is a specific, structural attack on the characteristics that made these companies creditworthy in the first place.
Recurring Revenue Under Threat
The per-seat licensing model that generates the predictable revenue streams lenders underwrite is being eroded by AI agents capable of automating the tasks those seats were purchased to perform. When an AI coding assistant can replace three junior developers, the enterprise customer does not renew three seats. When an AI customer service platform handles 80% of support tickets, the contact centre software vendor loses its volume-based pricing power. The revenue does not decline gradually; it steps down as contracts come up for renewal, and the $400 billion software sell-off in January 2026 reflected the market’s sudden recognition of this dynamic.
Switching Costs Collapsing
The conventional wisdom held that enterprise software was inherently sticky: the cost of migrating data, retraining staff, and integrating new systems created a natural moat around incumbent vendors. AI is dissolving that moat. Large language models can now facilitate data migration, generate integration code, and train users on new systems at a fraction of the historical cost. The switching cost premium that lenders embedded in their valuation models is compressing in real time.
Margin Compression
Perhaps most critically, AI is lowering the barriers to entry in software markets that private credit managers assumed were oligopolistic. Open-source AI models, low-code development platforms, and the ability of large enterprises to build bespoke internal tools using AI are all compressing the pricing power that justified the leverage ratios at which these loans were originated. A company underwritten at 6 times EBITDA on the assumption of stable 75% gross margins faces a fundamentally different credit profile if those margins contract to 55%.
The Collateral Problem
When a traditional industrial lender faces a deteriorating borrower, the collateral package provides a recovery floor: real estate, equipment, inventory, and receivables have tangible liquidation value. Software lending operates on a different calculus. The collateral is the business itself: its intellectual property, its customer relationships, and its contracted revenue.
AI disruption undermines each of these. Intellectual property loses value when competing capabilities can be replicated by foundation models at near-zero marginal cost. Customer relationships weaken when the product they are paying for becomes commoditised. Contracted revenue becomes less reliable when renewal rates decline and usage-based pricing models replace fixed subscriptions.
JPMorgan’s decision to mark down software loans held as collateral by private credit funds and restrict lending against those assets is the clearest signal yet that the banking system recognises this dynamic. Banks lend money to private credit funds using the funds’ loan portfolios as collateral through back-leverage arrangements. According to Federal Reserve data, Wall Street lenders had provided roughly $300 billion in such financing as of mid-2025, with JPMorgan alone carrying $22.2 billion of exposure.
When JPMorgan marks down the collateral, it directly reduces what private credit funds can borrow. For funds that relied on back-leverage to enhance returns, this creates a cascading effect: lower borrowing capacity, reduced ability to fund new originations, and in some cases forced deleveraging that requires selling assets at discounted prices.
The PIK Time Bomb
Embedded within the software loan portfolio is a structural feature that amplifies the AI disruption risk: payment-in-kind (PIK) arrangements. More than a third of private credit agreements with software borrowers at the end of 2025 included the option to switch to PIK, a threefold increase over three years.
PIK allows borrowers to defer cash interest payments by adding the owed amount to the loan principal. In a growing company, this is a rational financing tool: the borrower conserves cash for expansion, and the lender earns compound interest on a rising principal balance. In a company facing revenue decline from AI disruption, PIK becomes a mechanism for deferring recognition of distress.
The numbers are striking. PIK income now accounts for approximately 10% of total investment income for major BDCs, with roughly half of those PIK arrangements concentrated in the technology sector. When PIK income exceeds 10% of total income, portfolio-level liquidity stress becomes a genuine concern: the fund reports income it has not received in cash while simultaneously needing cash to meet distributions and redemptions.
For a BDC paying a 10% dividend yield funded partly by PIK income, the AI disruption of its software borrowers creates a triple threat: the PIK income may never convert to cash if the borrower deteriorates, the underlying loan value is declining as the business model erodes, and investors are simultaneously requesting redemptions that require actual liquidity the fund does not have.
What the Market Is Misunderstanding
The consensus view treats AI disruption risk as a sector-specific problem: software companies are vulnerable, but the rest of the private credit portfolio is fine. This misses two critical dynamics.
First, the AI disruption risk extends well beyond pure software companies. Business services firms, which account for roughly 30% of BDC holdings, are equally exposed: staffing companies face AI-driven automation of white-collar tasks, data analytics firms compete against AI-native alternatives, and professional services businesses see their billable-hour models threatened by tools that can perform the same work in minutes. UBS estimates that 25 to 35% of the entire private credit market faces elevated AI disruption risk, and in an aggressive disruption scenario, default rates in US private credit could climb to 13%, significantly higher than the 8% stress projected for leveraged loans.
Second, the repricing of software collateral has systemic implications for the entire back-leverage architecture. When JPMorgan marks down software loans, other banks follow. When back-leverage contracts, private credit funds reduce new originations. When originations decline, the mid-market borrowers that depend on private credit for financing face a capital drought: not because their own businesses are disrupted by AI, but because the lending ecosystem that funds them is contracting in response to disruption elsewhere in the portfolio.
This is how a technology shock in one segment of private credit becomes a credit contraction across the broader mid-market economy.
Investor Implications
The collision of AI disruption with private credit’s structural architecture creates a set of risks that cut across asset classes and require careful positioning.
Private Credit Allocations: Investors with exposure to BDCs or semi-liquid private credit vehicles should scrutinise their manager’s software and technology concentration. Funds with greater than 20% software exposure face materially higher NAV risk than the published marks suggest. The divergence between reported non-accrual rates and the market’s assessment of true credit risk is widening, and the funds most vulnerable are those that originated aggressively during the 2021 to 2022 vintage years.
Alternative Asset Managers: The publicly listed managers, Blue Owl (OWL), Ares (ARES), and Apollo (APO), face a fee revenue contraction as redemption-driven outflows reduce assets under management. Software companies’ stocks collapsed by almost 30% between October 2025 and February 2026, and BDCs with high software exposure have underperformed by roughly 5 percentage points since October.
Leveraged Loan Markets: If private credit managers are forced to sell performing loans to meet redemptions or restore leverage ratios, the resulting supply could pressure spreads in the broadly syndicated loan market. This creates dislocation opportunities for institutional credit investors with available capital, but it also means that the AI disruption premium is being transmitted from private to public credit markets.
Technology Sector Financing: The tightening of private credit underwriting for software borrowers will reduce available capital for mid-market technology companies at precisely the moment many of them need to invest in AI capabilities to remain competitive. The irony is acute: the technology that is disrupting lenders’ existing portfolios is the same technology that borrowers need funding to adopt.
Conclusion
Private credit built its largest single-sector concentration in software on the assumption that recurring revenue, high margins, and customer stickiness were durable characteristics. Artificial intelligence is challenging each of these assumptions simultaneously, and the $500 billion in outstanding software loans now sits at the intersection of technological disruption and financial engineering.
The problem is not that every software borrower will default. Many will adapt, invest in AI integration, and emerge stronger. The problem is that the lending structures were built for a world of gradual, predictable change, and AI is delivering rapid, discontinuous disruption. Collateral values are being repriced faster than quarterly NAV adjustments can capture. Back-leverage is contracting as banks reassess their exposure. PIK income is masking deterioration that has not yet surfaced in official loss statistics.
For investors, the key insight is that AI disruption risk in private credit is not a future concern: it is being priced into the market now, through JPMorgan’s collateral markdowns, through the BIS’s public warnings, through the widening gap between BDC stock prices and reported NAVs. The question is not whether the repricing will happen, but whether the managers holding $500 billion in software loans have the underwriting discipline, the balance sheet resilience, and the intellectual honesty to navigate it transparently. History suggests that in credit markets, the answer to that question is only ever revealed after the losses have materialised.
Sources: BIS: Private Credit’s Software Lending Meets AI Disruption · CNBC: Private Credit Worries Resurface as AI Pressures Software Firms · CNBC: JPMorgan Marks Down Software Loans · Yahoo Finance: UBS AI Disruption Analysis · Alternative Credit Investor: PIK Usage in BDCs · iCapital: PIK in Private Credit · Federal Reserve: Bank Lending to Private Credit · Prime Buchholz: Software Stress and AI Risk · S&P Global: AI Disruption Spills Into Private Credit
Related Reading: The AI disruption reshaping private credit’s software portfolio is part of a broader transformation in alternative asset markets. For the full account of how redemption gates and liquidity stress are unfolding across the industry, see The Private Credit Crackup. The earlier structural concerns about lending standards and default cycles are explored in Private Credit Faces Its First Real Test. The semi-liquid fund structures that amplify redemption dynamics are analysed in The Rise of Semi-Liquid Funds: Private Markets’ $4 Trillion Gamble, and the broader push to open private markets to retail investors is examined in Democratising Private Markets: The Retail Revolution and Its Risks. For the foundational growth story behind the asset class, see The Rise of Private Credit: From Niche to $1.7 Trillion. The implications for the publicly traded managers are analysed in KKR, Apollo, and the Alternative Asset Manager Model Under Pressure. See also Khan Capitals’ May 2026 coverage: $725bn hyperscaler AI capex cycle; Apple after Cook: Ternus and the AI question.
Leave a Reply