There's no SaaSpocalypse (Research)

How a private credit bubble got mistaken for an AI disruption story

Apr 07, 2026

👋 Hey, I’m George Chasiotis. Welcome to GrowthWaves, your weekly dose of B2B growth insights—featuring powerful case studies, emerging trends, and unconventional strategies you won’t find anywhere else.

This note is brought to you by Minuttia.

Most companies have no idea how they appear in AI search results.

The team at Minuttia built a free AEO report that shows how your brand appears across AI platforms like ChatGPT and Perplexity, based on real prompts, citations, and sentiment data.

You’ll get:

A visibility score and share of voice
Breakdown by topic, platform, and sentiment
A shareable report with clear next steps

It takes 30 seconds to request. Minuttia’s team does the rest.

Get your free AEO Report

Get the report. See how visible you really are. Then fix what’s broken.

A few weeks ago, the Financial Times published an article with a headline that caught my attention.

Apollo Global Management, one of the largest alternative asset managers in the world, had capped investor withdrawals from its flagship private credit fund.

The reason? Investor uncertainty about the impact of AI on the enterprise software industry.

Investors tried to pull roughly $1.6 billion from the Apollo Debt Solutions BDC, equal to 11.2% of net assets. Apollo honored just under half of those requests.

Apollo had already slashed holdings of riskier software loans and, according to their own letter to investors, “consciously chosen to create portfolios that are underweight software exposure relative to the broader private credit markets.”

Around the same time, Apollo’s John Zito went on record saying that software valuations in private equity are wrong. “All the marks are wrong,” he stated. The firm made bearish bets against software companies, including Internet Brands, SonicWall, and Perforce. Their target: reduce software exposure in credit funds from roughly 20% to below 10% of net assets.

Meanwhile…

Satya Nadella suggested on the BG2 podcast that applications “will collapse” in the AI agent era.
Forrester published a piece titled “SaaS as We Know It Is Dead: How to Survive the SaaS-pocalypse.”
And a16z published an essay arguing that AI will eat application software, framing this as “good news” for everyone except the incumbents.

The narrative, in short, is that software is cooked. That AI is about to consume the entire application layer. That SaaS as a business model is headed toward extinction.

I’ve followed this conversation closely for over two years. And I think the SaaSpocalypse narrative, while containing grains of truth, is significantly overblown.

Not because AI won’t change software. It absolutely will. But because the evidence being cited to support the “software is dead” thesis doesn’t hold up when you look at the actual numbers.

Let me explain.

Argument #1: Public SaaS Companies Are Posting Record Numbers

Let’s start with the most obvious place to look: public company earnings.

If software were truly being disrupted at the pace the narrative suggests, you would expect to see it in the financial results of major SaaS companies.

Revenue deceleration. Customer churn. Contracting margins. The usual symptoms of an industry under siege. Instead, here’s what the data shows for Q4 2025 and early 2026:

ServiceNow reported Q4 2025 subscription revenue of approximately $3.47 billion, up 21% year-over-year. Full-year subscription revenue hit $12,88 billion. Their 2026 guidance projects $15.5 billion in subscription revenue, representing 19-20% growth.
Cloudflare posted Q4 revenue of $614.5 million, a 34% increase year-over-year, and they’re expecting a total revenue of approximately $2,785 million in 2026.
Shopify reported Q4 revenue growth of 31% year-over-year and a 19% free cash flow margin.
Datadog grew Q4 revenue 29% year-over-year to $953 million. Customers with $1 million or more in annual recurring revenue grew from 462 to 603.
CrowdStrike grew subscription revenue 27% to just over $1 billion quarterly. Annual recurring revenue reached $4.24 billion. Operating cash flow hit a record $1.38 billion for the fiscal year.
HubSpot posted Q4 revenue of $847 million, up 20%. Full-year revenue reached $3.13 billion. Their customer base grew to approximately 289,000. (Great answer to those declaring HubSpot’s content strategy “dead,” don’t you think?)
Atlassian reported cloud revenue growth of 26% (around $928 million), while annual free cash flow reached $1.4 billion.
Monday.com grew revenue 25% in Q4. Enterprise customers spending more than $50,000 annually increased 34% from December. Net dollar retention rate in that cohort hit a historical high of 116%.
Salesforce posted $10 billion in Q4 revenue, up 8%. Their AI product, Agentforce, generated approximately $900 million in ARR in Q4 alone, growing 120% year-over-year.
And ServiceTitan, which went public in December 2024, reported a total revenue increase of 29% to $209.3 million in Q4 2025.

That’s ten public SaaS companies. Let me put the numbers side by side so you can see the full picture.

Let’s start with revenue growth and free cash flow:

And let’s also see customer traction and retention:

Every single company growing between 20% and 34% year-over-year (with Salesforce at 8% on a $38B base, which is its own kind of impressive).

Several posting record quarters. Most expanding free cash flow margins into the 20-57% range. Multiple hitting all-time highs on customer count and contract value.

Does this look like an industry on its deathbed? Because I don’t see it.

Let’s move on to the next one.

Argument #2: AI Companies Run on Enterprise Software

Here’s something that doesn’t get enough attention in the SaaSpocalypse discourse.

The very companies building AI products are deeply dependent on traditional enterprise software.

Anthropic, the company behind Claude, launched “Claude Apps“ in January 2026 with integrations for Slack, Figma, Asana, Atlassian, Notion, Stripe, and Zapier.

Not competing with these platforms. Integrating into them!

Claude Code can now read your Slack messages and take actions based on them. The entire premise is that these enterprise tools are the foundation that AI builds on top of.

Anthropic’s Claude models are also integrated directly into Salesforce’s Agentforce platform, running through Salesforce’s virtual private cloud for regulated industries like financial services, healthcare, and cybersecurity.

Netflix, Spotify, and Salesforce itself are Claude Code customers. These are massive, sophisticated technology companies choosing to layer AI on existing enterprise infrastructure, not replace that infrastructure with something new.

As Fortune noted in February 2026:

AI agents from Anthropic and OpenAI aren’t killing SaaS.

The integration model has become the norm. Enterprise software incumbents are deepening partnerships with AI providers rather than being displaced by them.

Author’s Note: I call these companies the industrial software complex.

Every company, regardless of how cutting-edge its product is, needs a CRM, a project management tool, a communication platform, and a financial system.

OpenAI uses Salesforce. Anthropic integrates with Atlassian. Perplexity runs on the same stack every other growth-stage company uses.

The AI revolution isn’t replacing the industrial software complex. It’s running on top of and because of it.

Argument #3: Vibe Coding Is Not Ready for Prime Time

One of the core assumptions behind the SaaSpocalypse thesis is that AI can now write software. And therefore, traditional software companies will be displaced by AI-generated alternatives that are cheaper, faster, and eventually good enough.

The reality is far more complicated.

The term “vibe coding“ was coined by OpenAI co-founder and renowned engineer Andrej Karpathy in a viral tweet that got over 4.5 million views.

Andrej Karpathy@karpathy

There's a new kind of coding I call "vibe coding", where you fully give in to the vibes, embrace exponentials, and forget that the code even exists. It's possible because the LLMs (e.g. Cursor Composer w Sonnet) are getting too good. Also I just talk to Composer with SuperWhisper

11:17 PM · Feb 2, 2025 · 6.85M Views

1.43K Replies · 3.62K Reposts · 33.5K Likes

He described a future where developers “fully give in to the vibes, embrace exponentials, and forget that the code even exists.” It’s a compelling vision. And for prototyping, side projects, and internal tools, there’s real value in it.

But for production enterprise software? The track record is sobering.

Enrichlead, a startup built entirely with Cursor AI and zero handwritten code…

leo@leojrr

my saas was built with Cursor, zero hand written code AI is no longer just an assistant, it’s also the builder Now, you can continue to whine about it or start building. P.S. Yes, people pay for it

4:34 AM · Mar 15, 2025 · 99.9K Views

77 Replies · 26 Reposts · 604 Likes

… was attacked within days of launching…

leo@leojrr

guys, i'm under attack ever since I started to share how I built my SaaS using Cursor random thing are happening, maxed out usage on api keys, people bypassing the subscription, creating random shit on db as you know, I'm not technical so this is taking me longer that usual to

9:04 AM · Mar 17, 2025 · 2.29M Views

612 Replies · 315 Reposts · 6.15K Likes

Maxed-out API usage. Subscription bypasses. Database corruption.

The founder, Leonel Acevedo, learned in real-time what happens when nobody understands what the code actually does.

And there are several more horror stories like this one.

The data paints an even starker picture:

According to a CodeRabbit report, AI-generated code creates 1.7 times more issues than human-written code.
A 2025 Veracode study found that 45% of code samples failed security tests and introduced exploitable bugs.
And, a survey of 18 CTOs, 16 reported production disasters directly caused by AI-generated code.

I’ve experienced this firsthand since I built an AI search tracking tool to monitor AI search visibility across platforms.

Don’t get me wrong. The tool works. But I can’t even count the number of times I had to intervene because the tool wasn’t performing as expected, moments before an important client or sales call.

Get your free AEO report

The classic “it works, but I don’t know why” problem that makes AI-generated code risky at scale. So, after getting my hands dirty with AI, I’ve landed here:

For prototyping and experimentation, these tools are extraordinary. For mission-critical enterprise applications where compliance, data integrity, and uptime are non-negotiable, they are not there yet.

You can argue with this all you want, but that has simply been my experience so far.

Let’s move on to the next one.

Argument #4: AI Disruption Is Bottom-Up, Not Top-Down

The most important nuance missing from the SaaSpocalypse conversation is where AI disruption is actually happening. And the answer is: primarily at the bottom of the market, not at the top.

AI is genuinely disrupting consumer and prosumer tools.

New AI-first design tools can generate production-ready interfaces from text prompts, which matters if you’re a freelancer or solopreneur using Canva.

Simple, single-purpose SaaS products are being absorbed into AI workflows. But enterprise software operates in a completely different universe. And the reasons are structural, not sentimental:

Integration complexity: A mid-market B2B company doesn’t use Salesforce in isolation. Salesforce connects to their marketing automation platform, their billing system, their customer support tool, their data warehouse. Replacing Salesforce means replacing or reconfiguring every integration point.
Compliance dependencies: Regulated industries like healthcare, financial services, and government don’t just use software. They certify it. Their audit trails, access controls, and data governance policies are built around specific platforms. Swapping those platforms isn’t a technology decision. It’s a regulatory project that requires legal review, board approval, and months of procurement approvals.
Institutional knowledge: Business logic, custom workflows, training materials, tribal knowledge about “how we do things” accumulates in enterprise software over the years. That institutional memory has genuine monetary value.

The market spending data confirms this bifurcation. Gartner forecasts worldwide software spending to grow 14.7% in 2026, reaching $1.4 trillion. Software is now the fastest-growing segment of the $6.15 trillion total IT spend. This simply means that enterprises aren’t cutting software budgets. In fact, the opposite is happening.

Salesforce added 6,000 enterprise customers while the rest of the world talked about an AI bubble. Over 40% of Agentforce Q2 bookings came from existing customer expansion.

What we’re witnessing isn’t the death of software. It’s a reshuffling.

Bottom-end, commoditized, single-purpose SaaS tools are vulnerable. Enterprise platforms with deep integrations, compliance frameworks, and institutional lock-in are not only surviving but growing faster than they were before AI entered the conversation.

The Counter-Argument: What the Bears Get Right

I want to be fair here. The bear case isn’t entirely without merit. And ignoring the legitimate concerns would be intellectually dishonest.

The a16z piece I read during my flight from Lithuania back to Greece a few weeks ago, “Good News: AI Will Eat Application Software,” makes a sophisticated argument.

The traditional SaaS model of selling seats and collecting recurring revenue is changing. Software historically “ate labor” by making workers more efficient. AI takes that further by potentially replacing workers altogether.

A16z points to portfolio company Salient, which replaces debt collectors with AI rather than selling debt collection management software. The business model shifts from “make humans more productive” to “replace the humans entirely.”

That’s a real shift. And it will affect certain categories of software, particularly those that exist solely to manage manual processes that AI can automate end-to-end.

The valuation compression is also real and worth acknowledging. Public SaaS median EV/Revenue multiples peaked at 18.6x in 2021 and now sit around 6-7x. That’s a 62% compression.

In early February 2026, over $1 trillion in market capitalization was erased from software stocks in a single week. Private credit default rates reached 5.8% through January 2026, the highest since the index launched.

Apollo’s (which we mentioned in the intro) concerns about software loans aren’t irrational. They’re a response to real repricing and real defaults on leveraged software companies.

The firms that loaded up on debt during the 2021 era of 18x (well, actually in certain cases more than 30x, but never mind!) revenue multiples are now facing a market that values them at 6x. That math doesn’t work for every borrower.

And the pricing model shift is genuine. IDC predicts that by 2028, 70% of software vendors will refactor pricing away from per-seat models toward consumption, outcomes, or organizational capability metrics. This transition will create winners and losers.

But here’s where the narrative breaks down:

Repricing is not the same as displacement. Valuation compression does not mean revenue destruction.

Private credit nervousness about overleveraged software companies is a financial markets story, not a technology story.

And the fact that SaaS pricing models are evolving doesn’t mean SaaS is dying.

The companies I listed earlier aren’t growing at double-digit rates because they’re riding on borrowed time. They’re growing because enterprises have real problems that require real software to solve, and AI is making that software more valuable, not less.

The Real Danger

So if AI isn’t killing software, what should we actually be worried about? Here’s the thing:

The real threat to the software ecosystem isn’t AI. It’s private credit.

Over the past decade, private credit has become the dominant financing mechanism for software acquisitions. What was an $8 billion market in 2015 has ballooned to roughly $500 billion in outstanding SaaS loans today. Software now accounts for ~19% of all private credit deployment, making it the single most concentrated sector in the asset class.

Much of that capital was deployed during 2020-2022, when interest rates were near zero (read: ZIRP era) and software multiples were astronomical. Global PE transaction volume hit $1.2 trillion in 2021, a 111% increase over 2020.

Deals were done at breathtaking leverage ratios. Vista Equity Partners and Elliott Management took Citrix private for $16 billion at roughly 15x debt-to-EBITDA. Zendesk went private for $10.2 billion. Cornerstone OnDemand for $5.2 billion. Entry multiples reached a new high of 11.8x EBITDA in 2025, well above the 2010-2022 average of 9.1x.

Now interest rates have risen, public market multiples have compressed from 18.6x to 6-7x EV/Revenue, and those deals priced at peak valuations are sitting on debt they can barely service.

The average PE holding period has stretched to a record 5.6 years, well beyond the typical 3-5 year exit window.

An estimated $1 trillion in unsold PE assets is piling up, creating what analysts are calling “zombie companies,” which may be surviving but are unable to service debt interest with operating profit.

The numbers are starting to show the strain. Private credit defaults hit a record 9.2% in 2025, according to Fitch Ratings. In early 2026, $17.7 billion in US tech loans dropped to distressed trading levels in just four weeks, the most since October 2022. Total tech distressed debt reached $46.9 billion by early February.

Marathon Asset Management’s Bruce Richards put it bluntly. Software companies in private credit portfolios are “about 10 times levered.” His prediction: a 15% default rate for leveraged software by 2027-2028, when maturity walls hit. Not a possibility, he clarified. A mathematical certainty.

Apollo’s John Zito was equally direct (sorry for repeating his words):

I literally think all the marks are wrong. I think private equity marks are wrong.

His recovery estimate for small-to-medium PE-backed software companies in the wrong AI positioning: 20-40 cents on the dollar.

And the redemption crisis is already here. Across every major private credit fund, investors are rushing for the exits:

Across half a dozen funds tracked by the Financial Times, investors submitted $11.7 billion in redemption requests. Only 66%, or $7.8 billion, were honored. Blue Owl suspended quarterly redemptions entirely and sold $600 million of its portfolio to a Goldman Sachs credit facility just to provide liquidity.

This is the part that gets lost in the SaaSpocalypse conversation:

The software companies posting 20%+ revenue growth I discussed earlier are public companies with healthy balance sheets. The companies in trouble are a different population entirely: PE-backed firms acquired at peak multiples with enormous debt loads, operating in a higher-rate environment with compressed exit multiples.

And, to reiterate:

Their problem isn’t AI disruption. Their problem is financial optimism that assumed permanently low interest rates and permanently high valuation multiples.

Conflating the two groups makes for a better apocalypse narrative. But it’s intellectually lazy.

A public SaaS company growing 25% with positive free cash flow is not the same as a PE-owned (or backed) software company leveraged at 10x EBITDA struggling to service its debt. The former is thriving. The latter is in a self-inflicted liquidity crisis.

And that distinction matters enormously. Because the private credit stress is real and getting worse.

If you want to worry about something in the software world, worry about that. Not about whether ChatGPT is going to replace Salesforce.

Final Thoughts

The SaaSpocalypse makes for compelling headlines. It fits neatly into the broader narrative of AI disrupting everything. And it contains enough truth to sound plausible.

But when I look at the actual data, what I see is not an industry in crisis. I see an industry in transition. The business models are changing. The pricing is changing. The capabilities are expanding.

But the fundamental need for high-quality software has not diminished.

My advice? Ignore the macro panic. Look at the micro fundamentals. If a SaaS company is growing 25% annually, expanding its customer base, improving margins, and integrating AI into its product in ways that create genuine value, it’s not dying. It’s evolving.

And evolution, last I checked, is the opposite of extinction.

Thank you for reading today’s note, and see you again next week.

(Yes, I’m changing the cadence back to weekly again.)

Research Disclaimers and Limitations

GrowthWaves and its author are not sponsored by or compensated by any company mentioned in this note. This is independent editorial analysis and does not constitute investment, financial, or legal advice. The author may have relationships with, work with, or hold equity in companies referenced; however, no content in this piece was influenced, commissioned, or incentivized by any such relationship. AI tools were used as a research assistant in the preparation of this piece. All claims are sourced and linked throughout.

Sources

Intro

Argument #1: Public SaaS Earnings

Argument #2: AI Companies Using Enterprise Software

Argument #3: Vibe Coding Reliability

Argument #4: Market Segmentation

Counter-Argument

The Real Danger: Private Credit

GrowthWaves by George Chasiotis

Discussion about this post

Ready for more?