Direct lending — and how AI disruption is reshaping risk across its loan books — has emerged as the defining story of the private credit ecosystem in the first half of the year. Nowhere has that tension been more visible than among business development companies (BDCs), the public market proxy for private credit. In recent months, rising redemption pressure and weaker performance within these vehicles have fueled concerns around mounting credit stress, particularly among software-exposed borrowers.
Through the second quarter of 2026, redemption requests at 10 of the 16 perpetual non-traded BDCs tracked by Fitch averaged 10.3% of shares, up from 9.7% in the first quarter. Most vehicles continue to operate with 5% redemption caps. Liquidity stress, however, is not the same as credit stress. Many market participants contend these headwinds are less a reflection of underlying fundamentals than of investor sentiment, arguing retail investors entered these vehicles without fully appreciating their structure.
After all, redemption gates exist precisely to prevent forced asset sales and protect remaining investors when withdrawals accelerate. That said, some BDCs have begun exhibiting signs of weakening credit quality. According to LCD, at the end of Q1, roughly 538 of the approximately 5,000 companies held by 170 BDCs showed signs of credit stress, reflected in higher non-accrual rates and declining fair values. Among the group, software companies accounted for the largest share of stressed borrowers.
While the BDC narrative has grasped headlines, signs of broad AI-driven stress in direct lending remain limited. That said, direct lending portfolios have accumulated notable software exposure, and lenders need to identify their borrowers that are most vulnerable to AI-driven disruption. Further, as a large universe of software borrowers navigate protecting their business models and driving new growth opportunities in an AI era, it will likely demand more proactive and granular portfolio monitoring.
Direct lending amassed significant dry powder in the early 2020s, coinciding with a wave of private equity investment in software companies under a markedly different cost of capital regime. According to LPL Financial, throughout 2022 and 2023, direct lenders funded roughly 40–70% of leveraged buyouts each quarter, a notable increase from a pre-pandemic share of 15–25%. Additionally, software and technology companies accounted for roughly 17% of BDC investments by deal count, second only to commercial services. As a result, the direct lending borrower universe today closely resembles the PE darlings of this era.
Notably, record-low borrowing costs during this period pushed software valuations to historic highs while enabling sponsors to finance acquisitions with generous amounts of leverage. For both lenders and sponsors, recurring revenue streams, sticky customers, high retention, and strong gross margins made software businesses a compelling investment pool.
These companies were viewed as highly durable businesses with predictable growth, underpinned by the belief that seat-based pricing could scale exponentially and exit multiples would remain resilient. As a result, many deals were underwritten against a growth vs. profitability framework (i.e. ARR rather than EBITDA.) This approach, however, leaves borrowers vulnerable to shifts in market perception, which is precisely the pressure now being brought to bear by AI.
In February, the launch of Claude Cowork triggered the so-called “SaaSpocalypse,” shocking software equity valuations as investors reassessed the durability of SaaS business models and the predictability of future cash flows in an AI-driven world. Although valuations have since recovered, software multiples have settled in the 5x–6x range for much of the market — a notable reset from the pandemic-era peaks that defined the sector. Ultimately, as the AI narrative unfolds, lenders must identify risks across their portfolios, use learnings to inform underwriting decisions, and proactively monitor credit indicators to assess priority areas.
For direct lenders, navigating AI-driven disruption will require developing methodologies for distinguishing fundamentally strong software companies from the broader universe. Across the industry, many lenders are approaching this via proprietary scorecards that rate underlying borrowers through AI risk-scoring frameworks based on a defined set of characteristics. Of course, across a portfolio of hundreds or thousands of borrowers, executing this exercise at scale and with the necessary precision requires a consolidated dataset of borrower fundamentals, qualitative insights, operating metrics, and more.
One of the most closely watched metrics across direct lending portfolios is the loan-to-value (LTV) ratio. A frequently cited point in recent years is that equity contributions in newer vintages have been historically high, with cushions still averaging roughly 48% even at the peak of deal activity in 2021. The challenge, however, is that software businesses are asset-light; the denominator in the LTV calculation is the value of a company’s earning power.
If AI disruption leads markets to assign lower growth expectations, margins, or terminal values to software companies, valuations fall, LTVs rise, and equity cushions erode. That said, direct lenders typically sit at the top of the capital structure through first-lien, senior secured positions and maintain exposure across large loan books spanning hundreds of borrowers. Losses would therefore require substantial value erosion that exceeds these seniority and diversification protections.
EBITDA is the input for many of the most important credit metrics, serving as the primary translation layer between business disruption and credit stress. Changes in operating performance can materially affect leverage and interest coverage ratios, reducing a company’s ability to service debt, compressing covenant headroom, and increasing the risk of default.
Here, AI disruption could hit EBITDA from multiple angles. Revenue erosion is the most direct channel, as AI-native competitors could disrupt incumbents, accelerate customer churn, or weaken pricing power during renewals. Companies may also face margin pressure as they increase investment in research and development to respond to a rapidly changing landscape.
Ultimately, median EBITDA and revenue growth across the PE-backed universe has slowed since pandemic-era highs. Current levels alone do not appear sufficient to trigger broad concern, and AI still holds significant potential to meaningfully expand EBITDA for borrowers that successfully leverage the technology to drive growth. That said, as lenders look to understand EBITDA across the portfolio, transparency into how sponsors report and adjust EBITDA has become paramount.
Throughout the SaaS lending boom, EBITDA definitions were often complicated with add-backs and forecasts in order to more accurately reflect a business’s nuance. However, these definitions can often leave reported earnings looking healthier than reality, making these adjustments essential to track across a portfolio.
Further, this dynamic often leads lenders to maintain their own views of EBITDA — often through bespoke calculations — to better understand risks across the portfolio. Equally important is the ability to cross-reference reported EBITDA in financials against compliance certificates to spot any deviations.
Additionally, EBITDA is ultimately a lagging indicator, often reflecting operational changes only after they have occurred. For software borrowers, more granular operating metrics such as net revenue retention, gross retention, and more can serve as earlier signals of potential credit distress. The advantage, therefore, lies with lenders that can not only seamlessly analyze EBITDA performance across their borrowers, but also systematically collect, centralize, and monitor granular KPIs to identify risks further upstream.
The true test of these pressures will come in the refinancing market. According to Allianz, roughly 30% of outstanding software loans mature by 2028, compared to 22% for the overall leveraged loan market, with 46% of software debt coming due within the next four years compared to 35% for the broader universe. Private credit likely faces a similar dynamic, and, if slower growth, weaker pricing power, and lower enterprise value multiples converge, refinancing conditions could tighten, liquidity could become more selective, and credit events could begin to rise.
There are already signs that lenders’ appetite for software risk is becoming more selective. For example, The Financial Times recently reported on the struggles private equity-backed cybersecurity firm Proofpoint faced extending its debt. The new debt priced at a yield of almost 9.3%, roughly 4.5 percentage points above the secured overnight financing rate, up from 3 percentage points on the company’s existing facility. This reflects the broader dynamic exhibited on software loans throughout 2026; spreads have widened, and lenders can demand higher yields and tighter covenants.
Ultimately, for borrowers approaching maturity, access to refinancing will likely depend on lenders’ conviction in the durability of their business models. Those viewed as more exposed could face wider spreads, tighter access to capital, lower leverage tolerance, and more restrictive covenant packages.
For borrowers with a longer window, sponsors have the opportunity to strengthen business models through AI-driven operational improvements, and lenders will need to map upcoming maturities against the pace of AI disruption, assessing whether technology risk is likely to emerge within the life of the loan and their existing structural protections.
Ultimately, the AI debate comes down to the core mandate of direct lending: preserving capital and being repaid in full and on time. For lenders, the economic consequences of AI will be measured not by falling software valuations or slowing EBITDA growth, but by what happens when stressed borrowers enter restructurings, workouts, or liquidations. By that measure, the asset class has historically performed well, maintaining low default rates and limited principal losses.
That said, headline default rates appear to be ticking higher, with recent estimates ranging from roughly 2% to 6%. Yet, private credit defaults are notoriously difficult to measure, as illustrated by the wide gap between those estimates. Differences in borrower universes, methodologies, and definitional nomenclatures make headline figures difficult to compare. Borrowers can also defer the “official” recognition of distress by capitalizing interest through PIK, extending maturities, or raising incremental capital without ever triggering a technical default.
More importantly, defaults are an event, not an outcome. What ultimately matters to lenders is how much capital is recovered once a borrower enters distress. By that measure, the data remains encouraging. In the first half of the year, Aksia analyzed more than 60,000 of its proprietary private credit transactions, including roughly 40,000 fully realized loans between 2013 and 2025.
The study found a cumulative loss frequency of 4.3% and a cumulative loss ratio of just 1.1%. Applied just to first-lien loans, those figures were even lower, roughly 3.6% and 0.8%, respectively. Given that direct lending loans typically mature over a four-year time frame, a headline 1.1% loss ratio translates into an annualized principal impairment of roughly 20 to 30 basis points.
All is to say — despite legitimate concerns around AI’s impact on private credit returns and potential capital losses, the current picture isn’t sounding the alarm bells. However, averages can be misleading, and not every portfolio faces the same outlook. As Howard Marks famously noted, “Never forget the six-foot-tall man who drowned crossing the stream that was five feet deep on average.” AI-driven losses will depend on the composition of lenders’ individual loan books, including sector concentrations, borrower business models, and documentation quality.
Explore our Navigating AI-Disruption in Direct Lending Report for a full look at how firms can seamlessly understand portfolio-level EBITDA growth, create custom alerts for early warning indicators, automate covenant monitoring, score borrowers on AI exposure, and more.
Get updates in your inbox
Learn how Chronograph can streamline your private capital investment monitoring and diligence