Software is in Trouble

Together with Cognitive Credit

As specialist AI systems move beyond simple tasks to making meaningful contributions to end-to-end fundamental analysis, 2026 will see many credit investors formally integrate Gen AI into their research processes.

This represents a shift in the analyst role—from information gatherer to supervisor of automated systems that can synthesize data at scale. Firms that fail to adapt risk falling behind. In this complimentary whitepaper, find out how Gen AI can be applied to credit research and how the operating model is likely to evolve over the next several years.

Content includes:

● The essential components of modern credit investing

●  Practical applications of Gen AI in credit research

● The evolution of credit teams in a Gen AI era

● The irreplaceable role of human judgement

Introducing Cognitive Credit AI: We’re also excited to announce the launch of Cognitive Credit AI. Built exclusively on our proprietary financials and document databases, it delivers specialist AI with deep knowledge of credit markets, free from human bias and third-party opinion. Distill years of historical figures & management commentary into actionable intelligence in seconds. Learn more here.

Welcome back!

Well, the headlines have been nuts recently, and private credit backed software loans are under the microscope. But it’s not even about being under the microscope - it’s about the fact that those looking in from the outside are unsure about the quality of these loans and are causing mass hysteria around the dangers of private credit.

When large managers like Ares need to have town halls to reassure the team about market volatility (reported by Bloomberg), it’s not exactly the most comforting thing, even for the biggest credit bulls. This coupled with the constant news of quarterly redemption requests exceeding 5% are driving a lot of fear-driven headlines around “___ firm limiting redemptions”.

It is extremely important to note that leveraged lending is cyclical, with higher default rates during worse economic environments. Plus, the AI risk that impacts software companies within the lending universe is very real. But with everything comes nuance and bifurcation, as opposed to black and white outcomes.

The Algorithm on X is favoring people with poor and sensationalized views of the private credit market. Ultimately, all this comes down to 1) asset selection (as it always has) and 2) being underweight a specific industry.

Leveraging my perspective, plus the perspective of the 12+ credit professionals in software I talked to over the past week, I’m going to opine a bit:

While the end of my time doing direct credit investing was more geared towards secularly challenged industries, the beginning of my career was focused almost entirely on Software.

The 2010s era of Software LBOs were generally viewed as very straight forward stories.

For years, software companies with high EV multiples and strong recurring rev and retention numbers were extremely easy to lever to 7x+. As a result, these were rated B2 or B3, and have pricing in the L+325-450 range. Additionally, the 2L paper was well covered and generally at L+800, although some spreads were wider. Some of that 2L paper was actually a very good trade - as you’d clip a nice coupon and then eventually, as the credit performed, you’d get refinanced out.

In the 2010s, Managers increased their Software exposure meaningfully as PE Sponsors viewed many software companies as checking all the boxes as a near-perfect LBO candidate. Lending demand was robust, and distress was relatively limited.

A lot of LevLoan & Private Credit lenders had at least 12% of their portfolio in software.

Outside of rare cases, and situations like Blackboard, Software was pretty well insulated. For anyone who didn’t use Blackboard in college, it was a way to check grades and course materials for college students and instructors. They quickly lost to share to competitors with better usability.

But for many years, the risk and mitigants section for software companies focused primarily on deferred revenue Adj. EBITDA addbacks, as opposed to any AI or moat-related risk.

That’s changing rapidly and some pockets of software are in real trouble.

Every single investment memorandum needs to talk about AI now, but not every conversation over the past few years was centered around AI risk. I would even say there’s been a large degree of sleepiness regarding AI in credit in the years post ChatGPT. But to be fair, it wasn’t abundantly clear we were in a “oh shit” moment until Claude’s advancements last December.

Obviously, this trade of “the perfect LBO candidate” is starting to break for Sponsors & Lenders. Moats, beautiful retention numbers, use cases, and valuation are all simultaneously under seige while contract renewals are going to be the shoes to drop in 2027-2029.

This will coincide with nearing maturities, including $40B in 2028 that is heavily comprised of B- paper.

Additionally, another large private credit borrower group outside of software & healthcare, are "tech-enabled" service providers who have some level of rev predictability as well. A bunch of private credit loans classified as “business services” companies are actually Software too. Bloomberg previously went through a list of holdings among 7 major BDCs and found 250 investments aggregating $9B+ that should probably be labeled as software instead of “business services” or something. Those types of names might run into trouble too.

I'm a big defender of the leveraged finance markets, but it's really important to note that a lot of these changes impacting public software companies are probably even more severe in the LMM, MM, and UMM.

Not everyone is exposed to the same software names though

Not every manager is the same though. It was reported that in mid-December (right before Claude’s material advancements) Apollo had been ahead of the curve in reducing its software exposure. Tbf, Software isn’t really something Apollo does on the PE side anyways.

Apollo's David Sambur warned in February that Software valuations need a "much needed reset". Apollo has zero software exposure in its PE business and <2% across its entire platform, according to Sambur. He was also quoted as saying: “Have we had groupthink where you’ve got 30% or 40% of buyouts being in software? In retrospect, it was a pretty big red flag."

There were also a bunch of funds (believe it was Diameter) making some noise around Internet Brands and comparables roughly a year ago around risk around AI displacing search traffic-driven revenue. Chegg was another obvious AI disintermediated name as well. Getty and Shutterstock too (more on that below).

But broadly, many Software companies are facing the following issues:

1. Anyone can build it at a much lower cost than historically. Allowing almost anyone to provide a comparable product in their suite. Suddenly, a lot of high-moat, limited-competition industries could be much more competitive. Features are suddenly commoditized in AI agents or AI-powered coding can quickly develop tools.

2. Margins should decline as a result and given higher AI-related costs. Historically, low capex was needed for software as well, that might fundamentally be changed.

3. Recurring revenue and retention are not as secure as they once were. This is more of a 2027-2029 story and may provide a false sense of comfort for investors in 2026.

4. Seat count business models are at risk and might need to change to usage based pricing.

5. For public names, a significant amount of cash flow is allotted towards stock based compensation, significant overhead reduction will be needed. Atlassian just announced they’re going to lay off 1,600 workers (10% of the company).

6. Many of these businesses should merge/consolidate, albeit it will be at uglier multiples. What comes to mind immediately are the domain service providers, while domain hosting is quite valuable, the ability to create and design a website has been commoditized by AI. $GDDY ( ▼ 1.88% ) has an interesting chart for example, roundtripping a strong 2024 and down -46% y/y. Getty and Shutterstock had to merge as AI-generated photos are a massive disruption to their businesses. Outside of photos from sporting events, the Grammys, or other real life events, many businesses are electing to just use AI-generated photos so they don’t even have to bother with Getty licensing fees or getting hounded around usage rights.

7. For companies that are highly levered, this risk is amplified. When now levering a company, one has to provide ample room to reinvest in the business. traditional software LBOs did not have this modeled out. The high leverage profile may inhibit a lot of these companies from reinvesting & staying competitive.

8. Ultimately, the focus of large and sophisticated organizations are going to be on how to effectively deploy agents for productivity and cost saving related gains. It’s not a given that the preferred agent of choice will be something developed by the Software provider. It’s probably going to be from Claude, OpenAI, or Gemini.

Who in the public markets loses if Software tanks?

I would actually argue “everyone” is a loser in the world of Software re-rating lower (given it was a high-growth, high-margin, part of index funds and all our of portfolios) but as it relates to some of what you see trading on the public markets here’s the “losers”.

1) Software Sponsors

Okay so some of these these large, software players are actually private, but I think they’re the biggest losers. Not Private Credit. Even though credit will lag Software. They’re obviously the first losers given where they sit in the cap stack and given the private credit docs are tighter than BSL docs.

What I can’t quite reconcile is how come everyone is bearish private credit because of software but not private equity? While some of these companies have taken their equity out of the business, this isn’t the case for everyone. A lot of Software oriented PE Sponsors are going to be in for a world of hurt and I think it will be much more severe than credit will face.

2) BDCs

We’ve seen a pain trade in BDCs already. This is unlikely to subside. There is clearly a bit of a mismatch where retail investors react fast and don’t fully understand that BDCs full of loans with clear maturities aren’t “stocks”. Sensationalized panic around private credit, redemption gating, coupled with the actual real situations of problems we’ll see, will result in retail being more likely to panic sell than institutions. We’ve already seen dividend cuts as a result of lower NAV.

3) CLO Equity

There are some CLO equity funds in the public markets, obviously these folks would be taking the first losses should there be impairments. Historically, CLO equity has been a great trade coming out of a crisis, but given the fluidity of AI risk I’m not sure this is the case anymore. BSL credit docs are weaker and CLOs, as amortizing funds, are not designed for workouts.

What do actual credit professionals think?

I chatted with a bunch of Credit Software folks to see what they’re thinking:

Basically, I got a mix of 1) the selloff is overblown, 2) underlying performance is fine, and 3) LTVs are still solid.

BUT - those views are just for where things stand over the past couple of weeks.

So here’s some deeper thoughts I consolidated:

• The rebound is probably just because we were oversold.

• It’s hard to reinvest into AI-powered software workflows given leverage.

• We’re not going to see AI impact software numbers in 2026, but there’s going to be a hit come 2027 and when it comes to refinancing in 18-24 months.

• Software credit professionals are having to spend a lot of time defending even the best, most mission-critical names.

• Going to investment committee on newer software opportunities are a much harder situation now.

• Many of the Software analysts we talked to begin with are underweight to begin with.

• Software exposure in European leveraged finance markets is much lower than in the U.S.

• Private credit docs are a lot stronger than broadly syndicated loan docs.

• Many software companies have given flash reports. but some of these reports lacked substance, despite the positive topline numbers.

• Some management teams seem complacent, or are conveying that they have a strong handling of the situation without really saying why.

• Enterprise adoption of AI tools will be slow and take some time.

If you’ve followed me for a while, you know one of my bigger pet peeves with the private markets, is how poor financial disclosures are post-underwrite. The limited transparency in financial reporting and MD&A is a big problem for managers that are asleep at the wheel.

Most finance professionals are non-technical, don’t understand coding, and lack some of the technological business model skills that ex-IB analysts/PE Associates who pivoted to tech after their 2-5 year stint in finance lack. I still think you can play a bit of catchup if you’re a software specific investor or you follow closely on FinTwit, but many finance professionals kinda waved through understanding software companies historically. It’s still quite hard to fully understand the software and true defensibility of software that you aren’t actually a user of.

The companies that are able to deploy an AI and use that for pricing power and retention will be much more valuable. We’re going to see a massive bifurcation in what type of software companies are valuable and which ones are screwed.

Which software companies live:

So I’ve been accidentally been quite bearish across this newsletter. But the reality is a lot of software loans and equities will be fine. Here’s some of the characteristics of the companies that could come out okay (according to me and the software credit analysts I spoke to):

• Companies that can leverage AI across their distribution will be best positioned.

• The pockets that seem more attractive include Cyber, ERP, and some vertical players.

  • There is some debate though regarding cyber, but I am in the pro-cyber camp.

• Horizontal is in trouble, vertical software should be more insulated.

• A regulatory element. Or something that involves high value risk transfer.

• It needs to be mission critical, cost of failure needs to be catastrophic.

• It needs to be a data of record.

• Scale - larger companies with larger profit potential will be able to weather the storm, while the LMM and MM will be in more trouble.

Right now, software companies have lengthier contracts, are growing HSD or double digits, and have good retention numbers. The problem will come in 2027 and onwards.

I can’t really imagine standing idle while the industry dramatically evolves, but if you cover software companies and haven’t circled up your group of winners, losers, and too-hards then you should probably do that now. You should also do that for any software public equity exposure you have - i did a big writeup on that specifically here.

Transparency is a problem in private equity and credit deals:

Look, I think one of the many mismatches (beyond potentially sophistication) between alternative asset professionals and the retail investors they want giving them money is the lack of disclosures. Obviously, beyond the name, pricing, and marks for loans, there’s limited disclosures around the actual underlying investments for Public BDCs and other private credit financings.

My two cents - if there was more disclosure around what the actual underlying software companies were, then there would be more certainty around which managers are more skillful than others.

For people who have seen how the sauce is made, a lot of financings make sense as underwrites, are well thought out and diligenced, and get past rigorous IC processes. From my lens, many of the loans underwrote by top investors are going to be fine, but an increased default cycle is only natural given the high leverage profile and historically ebbs and flows in default rates depending on the economic and interest rate environment. Although, I don’t think many private credit professionals are actually negative private credit….so there is some natural bias.

On the public side, it’s weird to me how companies are focusing on share buybacks instead of AI investment…when you see that it shows that many software teams aren’t actually worried.

Ultimately, this comes down to the same things it’s always come down to. Who is the more skilled investor, who can differentiate the winners and losers, and who isn’t afraid to go after hairier situations that have better pricing. Still, I genuinely believe your base case needs to include Claude and a swarm of agents being extremely intelligent and autonomous and whether that is a good or bad thing for your underwrite.

Until next time.

My Latest:

• Want to read more thoughts on software? I wrote on public equity software bifurcation in the WSR Investing Club here 

• Our Comp Report: Access the largest Compensation Report on Wall Street here 

• Our Job Board: Find Buyside and Sellside jobs on Buyside Hub here

• Recruit for Credit Jobs with HYH Premium - sign up below:

Access our latest HYH Premium piece, Portability Provisions: When the Debt “Travels” With the Company (and Why Everyone Suddenly Cares) here.