Public BDCs Address The Credit Risks Posed By Artificial Intelligence - Part I

Head-On

INTRODUCTION

In the last two weeks, which coincidentally coincided with the advent of the IVQ 2025 earnings season, there has been a sudden and widespread concern in the markets about how artificial intelligence (AI) might damage incumbent companies of every ilk, both public and private. Here is how Reuters framed the subject:

One Thing Leads To Another...

As our readers will know better than most, the anxiety has translated into severe concerns about the creditworthiness of many BDC-financed companies. Initially, companies in the "Software" segment were the focus of attention, but subsequently, the angst has expanded to any business where AI could change the landscape in the years ahead.

Many BDC stocks have dropped substantially in price on the back of these market speculations. The leading casualties have included Blue Owl Capital (OBDC), down (9%) since January 6, 2026; Blue Owl Technology Finance (-10.5%); TriplePoint Venture Growth (14.1%), and even the much beloved market leader Ares Capital (ARCC) is off (7.3%).

On every BDC earnings call, the managers have been almost universally proactive in addressing this new "wall of worry" in their prepared remarks. Likewise, the analysts have been just as keen to probe the matter, peppering the BDCs with questions - making this the subject "du jour". In this article, we've used our own AI assets to summarize how each of the 12 BDCs that have reported earnings and held a conference call tackled this thorny subject; we've identified the main themes and offered our own views. We're just at the beginning of getting our arms around this subject, but this represents a useful start.

Ares Capital Corporation (ARCC)

In the Q&A session, CEO Kort Schnabel addressed AI risk directly, noting that while AI is a disruptive force, Ares has constructed its portfolio to be resistant. He emphasized that they invest primarily in foundational infrastructure software that sits at the center of complex businesses, making it difficult to replace. He argued that AI cannot replicate proprietary data, which serves as a "data moat" for their portfolio companies, and noted that many of their companies serve regulated end-markets like healthcare and financial services, where trust in new AI solutions will take a long time to build.

Schnabel acknowledged that specific areas are at risk, specifically "single-function software apps," content creation tools, and data summarization businesses, but stated ARCC has very small exposure to these areas. He also highlighted the upside, noting that their portfolio companies are actively working to augment their products with AI solutions. Finally, he pointed to the structural protections of being a lender, noting that with loan-to-values (LTV) on their software book at 37%, there is a massive amount of equity cushion before the debt is impaired.

Capital Southwest Corporation (CSWC)

CEO Michael Sarner revealed that the company formed an "AI committee" approximately a year ago and created a specific segment in their investment committee process to rate AI risk for every potential deal. He noted that this scrutiny has real-world consequences, citing a recent deal they passed on because they could not get comfortable with how AI might impact the business valuation over a five-year horizon.

However, management also views AI as a potential benefit. Sarner noted that for some portfolio companies, particularly in financial services, AI is being used to increase efficiency. Ultimately, the firm is incorporating AI analysis into its internal operations as well to improve organizational efficiency.

Gladstone Investment Corporation (GAIN)

During the Q&A, management downplayed the direct competitive risk of AI to their specific portfolio. President David Dullum stated that most of their companies are using AI to enhance efficiency, such as product design at their portfolio company, Schylling.

.

Dullum distinguished GAIN's portfolio from those holding "tech companies" that might face direct competition from AI. He clarified that they do not have that specific type of exposure in their portfolio and, therefore, view themselves more as beneficiaries of the technology rather than victims of disruption.

Golub Capital BDC, Inc. (GBDC)

CEO David Golub addressed the "SaaSpocalypse" head-on, agreeing with market sentiment that AI is advancing rapidly and that some software companies are vulnerable. To mitigate this, Golub focuses on enterprise-critical platforms with sticky, embedded workflows and proprietary data sets that AI cannot easily replicate. He explicitly stated they avoid software focused on content creation, analytical overlays, or basic tools, predicting there will be "a lot of losers" in those areas.

Golub also outlined a framework for how AI risk materializes, suggesting it starts with slower growth leading to lower equity valuations, rather than an immediate credit meltdown. He argued that for AI to cause credit losses, it would have to completely replace incumbent products quickly, which he views as the least likely near-term scenario. He concluded that Golub Capital will likely continue to be a leading software lender, potentially benefiting if the BSL market retreats and pricing improves for private lenders.

Gladstone Capital Corporation (GLAD)

President Bob Marcotte addressed AI largely in the context of infrastructure and data centers. He clarified that GLAD does not directly invest in data centers, calling it a "big boys game," though they see some spillover spending in their manufacturing portfolio companies (e.g., bus bars, HVAC services). Marcotte expressed caution regarding the sustainability of the AI infrastructure boom. He noted they are "very cautious" about the reliance on that end of the market and are not investing in companies with a significant reliance on the continuation of that specific investment spend.

Hercules Capital, Inc. (HTGC)

CEO Scott Bluestein defended the portfolio vigorously, characterizing AI as a "net positive" for their business. He argued that their portfolio consists of innovative, entrepreneurial businesses led by technical founders who are integrating AI into their products, rather than "legacy providers" who lack proprietary data and are most susceptible to disruption.

HTGC emphasized its defensive underwriting standards, noting they avoid pure-play AI or data center GPU financing. Instead, they target software companies with specific domain expertise, hardware moats, or regulated customer bases. Bluestein also highlighted their conservative structure, targeting LTVs less than 20% and short loan durations (less than 24 months), which materially derisks the portfolio against long-term technological shifts.

Oaktree Specialty Lending Corporation (OCSL)

Management described a tightening of underwriting standards, prioritizing software businesses with "multiple control points," "data gravity," and a coherent AI roadmap. Co-CIO Raghav Khanna noted that AI has "raised the quality bar," leading to a higher pass rate on new opportunities.

CEO Armen Panossian highlighted that the primary risk isn't necessarily immediate performance degradation, but "refinanceability risk." He explained that if a sponsor fears long-term AI disruption, they may refuse to inject the equity capital needed to refinance the debt at maturity. To mitigate this, OCSL avoids "cov-lite" arrangements in the middle market and has minimal exposure (2%) to Annual Recurring Revenue (ARR) loans that often lack cash flow visibility.

PennantPark Floating Rate Capital Ltd. (PFLT)

CEO Art Penn differentiated PFLT by emphasizing their low exposure to software (4.4%) compared to peers who may have 20-30%. He stated they "stuck to their knitting" by avoiding the high-leverage, covenant-lite, ARR-based loans prevalent in the market.

Penn explained that their limited software exposure is focused on enterprise software integral to customers in heavily regulated industries like defense, healthcare, and financial services, where safety and security concerns mean change happens slowly. He argued this conservative positioning is a "significant differentiator" from peers facing scrutiny over their software books.

PennantPark Investment Corporation (PNNT)

Similar to PFLT, CEO Art Penn reiterated that software comprises only 4.4% of the PNNT portfolio and is structured with reasonable leverage and covenants. He noted that they debate AI as a "help or hindrance" in every investment committee meeting.

Penn shared a core underwriting philosophy: they ask, "If this company goes away, who really cares?" If the answer is affirmative, it implies the company has a defensible moat or market share that AI is less likely to destroy immediately. He acknowledged that while technological change is often overestimated in the short run, it is underestimated in the long run, but their short average maturity (approx. 3 years) helps defend against long-term shifts.

Prospect Capital Corporation (PSEC)

President Grier Eliasek highlighted that PSEC has avoided the software volatility by having significantly less exposure (2.8%) compared to the BDC average of 22%. He emphasized that they never participated in the Annual Recurring Revenue (ARR) loan market, viewing it as risky due to the lack of cash flow exit strategies.

When asked for a view on the future of AI and software, CEO John Barry exercised "intellectual modesty," stating he would not prognosticate on the sector because "I really don't know what's happening or going to happen with AI, software. And I don't think anybody else does".

Saratoga Investment Corp. (SAR)

CIO Michael Grisius noted that while they have expertise in SaaS lending, they are evaluating AI's impact on "every single deal" they look at. If the conclusion regarding AI's impact is "hard to say," they will steer away from the deal.

They view AI as a "double-edged sword," acknowledging that some portfolio companies are incorporating AI to improve their credit profiles. Grisius also noted that while they remain active in SaaS, the majority of their recent deals have been outside the software space.

Sixth Street Specialty Lending, Inc. (TSLX)

CEO Bo Stanley provided a detailed framework, arguing that AI "levels the playing field on development costs," which will reduce pricing power and revenue growth for many software companies. However, he views this largely as an "equity valuation problem" rather than a credit problem. He noted that while enterprise value multiples have compressed significantly, credit spreads remain resilient.

Stanley argued that the true "moat" for a software company isn't the code (which AI makes cheaper to write) but the incumbency advantage—owning the customer relationship, data integration, and regulatory complexity. TSLX believes existing companies that are well-managed can use AI to lower their own costs and defend these moats. They also noted that 40% of their portfolio has enterprise software exposure, but they map this by end-market (e.g., healthcare, business services) to better reflect the underlying demand drivers.

Summary of Main Themes

1. "Data Moats" and Incumbency as the Primary Defense A dominant theme across almost all BDCs (ARCC, GBDC, HTGC, OCSL, TSLX) is the reliance on proprietary data and deeply embedded workflows as a defense against AI. The consensus is that while AI makes writing code cheaper (lowering barriers to entry), it cannot easily replicate historical customer data or complex, mission-critical integrations. BDCs are prioritizing companies that are "systems of record" rather than simple tools or content generators.

2. Equity Risk vs. Credit Risk Several managers (GBDC, TSLX, OCSL) distinguished between the risks to equity holders versus lenders. The argument is that AI may slow growth and compress valuation multiples (an equity problem), but as long as the companies remain cash-flow positive and retain their core customers, the debt remains money-good. OCSL added a nuance to this, warning of "refinanceability risk"—where the debt is performing, but equity sponsors refuse to refinance at maturity due to long-term valuation fears.

3. Avoidance and Low Exposure A significant subset of BDCs (PFLT, PNNT, PSEC, GLAD) differentiated themselves simply by having minimal exposure to the software sector (often <5%). These managers emphasized that they "stuck to their knitting" by avoiding the trendy, high-leverage, Annual Recurring Revenue (ARR) loans that are now under scrutiny, preferring instead to lend to cash-flowing industrial or service businesses.

4. Regulated Industries as a Shield Many BDCs (ARCC, PFLT, PNNT, HTGC) are taking refuge in highly regulated industries such as healthcare, defense, and financial services. The thesis is that regulatory requirements for safety, security, and data privacy will slow the adoption of AI disruptors, providing a longer runway for incumbent portfolio companies to adapt or pay down their debt.

VIEWS

Not surprisingly, no BDC concedes that a significant credit downturn will occur to BDC-financed software companies generally and to their own exposure in particular. However, we were impressed by the willingness of several players with significant investments to discuss the inherent risks constructively and not just dismiss the market jitters as unreasonable.

We were also reminded that the stock markets - and the financial journalists who shape sentiment - may have only very recently awakened to the potential impact of AI technology on virtually every segment of the economy, with “software (whatever that means) at the forefront, but the professional lenders have been contending with the complex issues involved for some time now. The same can be said for the private equity community, which tends to know what it’s doing. We get the impression there’s been no blind accumulation of new borrowers by BDCs in recent years, but a thoughtful assessment of the risks and opportunities.

Of course, only time will tell whose credit assessment was spot on and whose was not. That’s why we’re beginning to track at the BDC Credit Reporter every material BDC exposure to the AI phenomenon with a view to identifying changes in valuations by BDC and by segment over time. We'll share some of our findings on these pages.

Of course, investors who are truly nervous about this subject would be best off focusing on BDCs that have eschewed lending to technology-related businesses that might get “disrupted” in the years ahead. We are less than a third of the way through earnings season, but we can say with some confidence that most of the lower middle market-focused BDCs and a few targeting the core middle market have little in the way of borrowers at risk. Very roughly, about a quarter of the public BDC universe is a “safe harbor” in this regard.