Private credit firms are pouring billions into artificial intelligence and software companies, chasing the promise of recurring revenue streams and explosive growth. But a new research report from Kroll Bond Rating Agency is sounding an alarm that many lenders may be underestimating the unique risks embedded in these deals — risks that could redefine how the industry evaluates creditworthiness in the age of AI.

The report, released by KBRA via BusinessWire, represents one of the most comprehensive attempts by a ratings agency to frame the specific credit risks associated with lending to AI and software businesses. As private credit assets under management have ballooned past $1.7 trillion globally, the concentration of lending activity in technology — particularly in AI-adjacent companies — has become impossible to ignore. KBRA’s analysis arrives at a moment when the intersection of rapid technological disruption and leveraged lending is creating a volatile cocktail that demands closer scrutiny from institutional investors, fund managers, and regulators alike.

Private Credit’s Insatiable Appetite for Software and AI Deals

The surge of private credit into software and AI is not accidental. Software companies, particularly those operating on a Software-as-a-Service (SaaS) model, have long been attractive to direct lenders because of their predictable, subscription-based revenue, high gross margins, and relatively low capital expenditure requirements. These characteristics create what appears to be an ideal borrower profile: steady cash flows that can reliably service debt. Over the past several years, software has become the single largest sector in many private credit portfolios, with some business development companies reporting technology exposure exceeding 25% of their total assets.

But the arrival of generative AI has fundamentally altered the calculus. KBRA’s research highlights that AI is introducing a layer of complexity and uncertainty into the software sector that traditional credit analysis frameworks were never designed to capture. The speed at which AI can render existing software products obsolete, the capital intensity of building and maintaining AI infrastructure, and the uncertain path to monetization for many AI-native startups all represent material risks that lenders must now grapple with. The rating agency’s work underscores a critical tension: the very qualities that made software lending attractive — predictability and stability — are being eroded by the disruptive force of AI itself.

KBRA’s Framework: Dissecting the Layers of AI Credit Risk

KBRA’s report identifies several distinct risk vectors that private credit investors should be evaluating when underwriting AI and software companies. Among the most significant is what the agency describes as technology displacement risk — the possibility that a borrower’s core product or service could be supplanted by AI-powered alternatives in a compressed timeframe. Unlike previous technology cycles, where disruption played out over years or even decades, AI-driven displacement can occur with startling speed. A SaaS company that today commands strong retention rates and healthy net revenue retention could find its value proposition undermined within quarters, not years, as competitors integrate large language models or other AI capabilities into rival offerings.

The report also draws attention to the distinction between companies that are genuinely AI-native — meaning their core product is built on proprietary AI models and data — and those that are merely AI-adjacent, layering AI features onto existing software platforms. This distinction matters enormously for credit analysis. AI-native companies may face higher upfront capital requirements, greater uncertainty around product-market fit, and more volatile revenue trajectories. AI-adjacent companies, meanwhile, face the risk that their AI enhancements are easily replicable, offering no durable competitive moat. KBRA’s framework encourages lenders to rigorously categorize borrowers along this spectrum and adjust their risk assessments accordingly.

The Recurring Revenue Mirage and Covenant Erosion

One of the more provocative elements of KBRA’s analysis concerns the reliability of recurring revenue metrics in the AI era. Annual recurring revenue (ARR) has become the lodestar metric for private credit underwriting in software. Lenders routinely structure facilities around ARR-based leverage multiples, and the metric is often used as a primary covenant trigger. But KBRA’s research suggests that ARR may be a less reliable indicator of credit quality than the market assumes, particularly for companies exposed to AI-driven disruption. Customer churn could accelerate rapidly if AI alternatives emerge, and the stickiness that historically characterized enterprise software contracts may weaken as switching costs decline in an AI-enabled world.

Compounding this concern is the ongoing erosion of covenant protections in private credit deals. As competition among direct lenders has intensified, borrowers have extracted increasingly favorable terms, including covenant-lite structures, generous add-backs to EBITDA, and loose definitions of what constitutes recurring revenue. KBRA’s report implicitly warns that these relaxed terms could amplify losses in a downturn scenario where AI disruption triggers a wave of credit deterioration across software portfolios. The combination of potentially overstated ARR, thin covenant protections, and concentrated sector exposure creates a risk profile that many private credit portfolios may not be adequately pricing.

Valuation Pressures and the Denominator Effect

Beyond the direct credit risks, KBRA’s research touches on the broader valuation dynamics at play in the AI sector. Private credit lenders to AI and software companies often rely on enterprise value as a key metric for assessing loan-to-value ratios and recovery prospects. But enterprise valuations in the AI space have been inflated by speculative enthusiasm, with revenue multiples for some AI-native companies reaching levels not seen since the dot-com era. If sentiment shifts — whether due to a high-profile AI company failure, regulatory intervention, or simply a recalibration of growth expectations — the resulting valuation compression could leave private credit lenders with significantly less collateral coverage than they anticipated.

This valuation risk is particularly acute for lenders who have extended unitranche or second-lien facilities to AI companies at elevated leverage levels. In a scenario where a borrower’s enterprise value contracts by 40% or 50% — not an implausible outcome for a company whose AI product fails to achieve commercial traction — recovery rates could fall well below historical norms for software lending. KBRA’s analysis serves as a reminder that the favorable loss experience private credit has enjoyed in software over the past decade may not be a reliable guide to future performance in an AI-disrupted environment.

The Data Moat Question and Intellectual Property Risks

Another critical dimension of KBRA’s research involves the role of proprietary data and intellectual property in determining the creditworthiness of AI borrowers. Companies that possess unique, high-quality datasets may enjoy a significant competitive advantage, as the performance of AI models is heavily dependent on the data used to train them. Lenders who can identify borrowers with genuine data moats may be better positioned to underwrite durable businesses. However, assessing the quality and defensibility of a data asset is a fundamentally different exercise than evaluating traditional software metrics like customer count, contract length, or gross margin.

Intellectual property risks also loom large. The legal environment surrounding AI is evolving rapidly, with ongoing litigation over copyright, data usage rights, and model training practices. A borrower whose AI model is built on data that is later deemed to have been improperly used could face existential legal liability. KBRA’s report encourages private credit investors to conduct deeper due diligence on the provenance of training data and the robustness of IP protections — areas where many lenders currently lack specialized expertise.

What Comes Next for Private Credit’s AI Bet

KBRA’s research arrives at a pivotal moment for the private credit industry. The asset class has grown rapidly, in part by filling the void left by banks retreating from leveraged lending in the wake of post-financial-crisis regulation. That growth has been fueled by strong returns and relatively low default rates, particularly in software lending. But the agency’s work suggests that the industry may be approaching an inflection point where the risks associated with AI and software lending begin to diverge meaningfully from historical experience.

For institutional allocators — pension funds, endowments, insurance companies, and sovereign wealth funds that have poured capital into private credit strategies — the implications are significant. Portfolio concentration in AI and software, once viewed as a source of stability, may increasingly be seen as a source of systemic risk. KBRA’s framework provides a starting point for more nuanced risk assessment, but the ultimate responsibility lies with lenders and their investors to develop the specialized expertise needed to navigate a sector that is evolving faster than the underwriting models designed to evaluate it. The firms that invest in building genuine AI literacy within their credit teams will likely be the ones best positioned to distinguish between the transformative companies that will define the next era of technology and the overleveraged casualties of a hype cycle that has yet to fully play out.