How The BDC Credit Reporter Is Tackling The Potential Impact Of AI On BDC-Financed Companies

February 18, 2026

In an opinion piece today Paul Davies in Bloomberg, citing from an earlier article in...Bloomberg, warned:

Artificial intelligence fears have ripped through stock and bond markets, but investors in loans and private credit are still playing catch-up. Part of the problem is working out which areas of the debt market are exposed to disruption and which aren’t. Because fund managers apply industry categories in varying ways, some companies that are really software sellers are misclassified as anything from retailers to food producers, according to Bloomberg News.

That's because many BDCs - very reasonably and innocently - label portfolio companies by the industry they serve rather than by the products they originate. Moreover, few companies only do one thing or operate in one segment, so categorization is always going to be a challenge.

In any case, Mr Davies goes on to point out the following:

Within the relatively transparent world of BDCs, at least 250 loans to software firms worth more than $9 billion were categorized as some other kind of business by one or more of those often publicly-traded vehicles that invest in private credit, Bloomberg News found. A pricing-software business called Pricefx was labelled “business services” by one lender, while Restaurant365, which supplies software to eateries, was called “food products” by another. Such idiosyncratic classification doesn’t make the business more likely to be blown up by AI — but it sure doesn’t help investors.

At the end of the article - after making the point that Private Credit and BDC exposure to "Software" may be larger than reported, Mr Davies makes this suggestion:

For investors, the choice is run away from the risk as many are doing in stock markets — or really dig into the stresses their debt funds actually face.

Reporting For Duty

This is a long way round for the BDC Credit Reporter to let you know that we're digging in, and at the best possible time: just as BDCs are disclosing their most recent portfolios in Week 3 of IVQ 2025 earnings season. Frankly, we are not convinced there is a "SaSS" or any other kind of "apocalypse" headed this way, but that's besides the point. What matters is what the markets believe. Only over time will we be able to tell if all the anxiety of the past few weeks was appropriate or not. We can help the process of resolving this existential conundrum by amping up our fundamental research on this thorny subject.

Coming Up

Here's what we're proposing to do in the weeks, months and quarters ahead:

For every public BDC (sorry, we don't have the resources to tackle the non-traded players as well) we'll be identifying every "Software" related portfolio company. If Software is roughly one-fifth of all holdings, that should be somewhere between 1,000-1,600 companies. We'll rely on each BDC's categorization at first, but we'll also be identifying any other entities that seem to fit the mold.

Sort Function

In an ironic twist, we'll be using our own preferred AI tools to evaluate which companies are most "at risk". For full transparency, here is the initial prompt we are using to systematically assess the situation:

Every time I enter the name of a company, all of which happen to be financed by BDCs, review their product offerings, industries targeted, and business operations, and try to determine if developments in artificial intelligence might cause their customers to switch away to firms like Perplexity, ChatGPT, Grok etc. to save money and improve their operations. Rate the risk as either LOW, MEDIUM or HIGH. If possible, find any discussion of this subject in the public record.

Here is a link to an example of the AI risk analysis we'll be compiling, using a portfolio borrower of Ares Capital (ARCC). We imagine both our prompt and the output will evolve over time.

Evolution

We will also be tracking the changing cost and fair market value of all these "Software" companies, using the database resources of Solve, from the IVQ 2025 on - which seems like an auspicious starting point - throughout 2026 and - possibly beyond. (To save us all from information overload, we will be undertaking an AI risk analysis only for company investments greater than $5mn - our standard materiality threshold).

Different Strokes

This is important because there is a huge variance in BDC exposure. Many players have almost no "Software" companies to worry about while some have taken on - thanks to the way they classify businesses by industry served - much more than 20%. The table will also show over time the changing relative weight of "Software" investments. We wouldn't be surprised to see some BDCs - even if not experiencing any credit troubles - de-emphasizing "Software" investments in order to boost their stock price. After all there are many other loan waters to fish in. The popularity of "Software" lending has been due to the prevalence of such borrowers; the excellent credit track record they have enjoyed and the decent terms enjoyed. That last factor might continue but the first two may not.

Write About

Furthermore, we'll be compelled, till this issue resolves itself, to discuss the AI elephant in the room in all of our future BDC Credit Reporter articles, based on what we can tell from the public record about a specific business.

FINAL THOUGHTS

We have four concluding comments.

First, we need to re-iterate that a BDC with a substantial exposure to "Software" borrowers will not necessarily suffer a corresponding level of credit losses as a result. The reverse is also true. A low percentage of Software investments does not guarantee a clean credit sheet. As always there remain plenty of credit reefs out there for borrowers to wreck themselves on besides AI. Unfortunately for us, and for anyone involved with BDCs, a "granular" approach is going to be necessary.

Second, whether a "Software" company's risk of being disrupted - or even put out of business, as many have warned - can be truly determined in advance is debatable. We offer our rating system only for "educational purposes" and everybody should draw their own conclusions.

Third, after hearing from a fourth of the public BDC universe where IVQ 2025 results are concerned - and identifying all the current crop of non-accruals in yet another table that we'll be sharing with our readers - we notice that there are very few "Software" companies in serious credit trouble. So far we've identified over 60 BDC-financed companies that are non-performing and only 3 appear to be "Software" related.

Of those, one is in trouble due to the downturn in the freight sector, not because its technology is obsolete because of AI. However, a valid argument could be made that both Pluralsight and Astra Acquisition fit the AI risk mold. Read the latest articles about each of these, here and here. The latter filed for Chapter 11 and has emerged debt-free and with a new name and format. The BDCs involved seem to have lost everything, not helped by the fact that the company's EBITDA dropped (88%). The former is a famous name now owned by its BDC lenders since 2024 which has not yet turned the corner as hoped and could yet result in over $100mn in realized losses.

Fourth, we'll be initially focusing on "Software" companies whose debt is coming to maturity in 2026. The crisis of confidence may impact even businesses that are performing normally and make refinancing difficult. We expect incumbent BDC lenders to either extend those loans with support from the sponsors or - in extreme circumstances - undertake debt-for-equity swaps. Far less likely is a wave of bankruptcies, but this is uncharted territory.