I've spent over a decade building, selling, and scaling SaaS and infrastructure products — from early-stage startups to enterprise platforms. I've watched this industry survive the "cloud is a fad" era, the ZIRP hangover, and at least two rounds of "software is dead" narratives. It's never been dead. But it has always been evolving, and the companies that refuse to see the shifts clearly are the ones that don't make it. What's happening right now is real, it's significant, and it deserves a more honest conversation than either the doomsayers or the cheerleaders are offering. No, the sky is not falling — but if you're building or investing in software and you're not paying close attention, you're going to get caught off guard.
For those of you who might not pay close attention to the private credit or stock markets, you might not have noticed that software stocks are getting crushed. Hedge funds are dumping SaaS positions at a pace we haven't seen since 2008. Private credit firms with $100 billion in software exposure are watching their balance sheets deteriorate in real time. Traders on Wall Street are calling it a "SaaSpocalypse."
Meanwhile, on the other side of the hype cycle, AI evangelists promise that every industry will be transformed within 18 months and that trillion-dollar markets are being created overnight.
Both camps sound confident. I think both are partially right and meaningfully wrong.
As someone who has spent years building in the data infrastructure space, I've been watching these narratives collide — and honestly, the lack of nuance in either direction is doing real damage to how companies are being evaluated, funded, and built.
Here's my honest take.
Confidence in the SaaS model has shattered
Let's start with the numbers. As Jamin Ball recently noted in Clouded Judgement, the median next-twelve-months revenue multiple for cloud software has dropped to 4.1x — the lowest in a decade. The median free cash flow multiple sits at 18.9x, roughly 30% below the previous 10-year low. These aren't normal fluctuations. This is a structural repricing.
The reason goes deeper than a bad earnings season. SaaS businesses have long been valued as predictable cash flow machines — spend aggressively early, flip to profitability, then compound. The math behind those valuations rests on two foundational assumptions: that retention rates remain high and stable, and that the business has meaningful terminal value. AI is now calling both assumptions into question simultaneously.
If customers leave legacy SaaS vendors for AI-native alternatives, retention craters and the cash flow model breaks. If entire software categories get commoditized, the terminal value for some of these companies may genuinely be zero. Even if you disagree with the most bearish case, the probability of those outcomes is higher today than it was a year ago — and that alone justifies lower multiples.
The disruption is real, but it's not uniform
The core anxiety driving the selloff is straightforward: AI can now do things that used to require purpose-built software. Legal review tools, customer service platforms, content management systems, basic analytics — the list of categories where AI is a credible alternative grows every week. That's not hype. That's happening.
But the leap from "AI can replace some software categories" to "sell everything with a SaaS business model" is exactly the kind of overcorrection markets are prone to. "Software" is not a single thing. Treating all software companies as equally exposed to AI disruption is like saying every business that uses electricity is equally vulnerable to a grid failure. The exposure varies enormously depending on where you sit in the stack.
Three very different realities under one label
Software that AI can replace. These are application-layer products whose core value proposition is a workflow that AI can now perform directly. Document review, templated content generation, basic data entry automation, simple customer routing. If your product is essentially a codified process wrapped in a UI, and that process can now be handled by a foundation model with a good prompt, the threat is real and immediate. This isn't a valuation problem — it's an existential one.
Software that needs to evolve. This is the largest and most interesting group. Most horizontal SaaS platforms aren't going to disappear overnight, but they face intensifying pricing pressure and feature commoditization. As Ball points out, the deeper issue isn't that someone will "vibe code" a replacement for Salesforce — it's that the marginal cost of creating software has collapsed, which will flood every category with competition and commoditize markets faster than incumbents can respond.
The stock market is already sorting this group in real time. HubSpot, a strong company by any traditional SaaS metric, saw its stock drop roughly 50% in 2025 as investors questioned whether SMB CRM and marketing automation can defend its pricing against AI-native alternatives. Adobe fell around 35% despite genuinely impressive AI capabilities in Firefly — the market's concern isn't that Adobe isn't innovating, it's that standalone AI tools can now deliver "good enough" creative output for the majority of use cases at a fraction of the cost. Atlassian and Monday.com saw similar declines as investors recalibrated what project management and collaboration software is worth in a world where AI agents can coordinate work autonomously.
These are not failing companies. They are strong businesses facing a fundamental question: can they integrate AI deeply enough to become more valuable, not less? The market is right to ask hard questions here. It's wrong to assume the answers are universally negative.
Software that AI depends on. Infrastructure — the systems that move data, manage compute, handle security, orchestrate distributed workloads — doesn't get replaced by AI. It gets consumed by it. Every AI workload needs to ingest data at scale, process events in real time, route outputs to downstream systems, and do all of this reliably across global environments. The rise of AI is arguably the single biggest demand driver infrastructure software has ever seen.
The companies that have figured this out are thriving while the rest of the sector burns. Cloudflare's stock rose over 80% in 2025 — not because it added AI features, but because its edge computing infrastructure is where AI inference actually runs. As one analyst noted, Cloudflare isn't selling AI features; it's selling the pipes that AI runs on. Datadog saw its AI-native customer revenue grow from 4% to 11% of total revenue in a single year, with over a dozen AI-native companies each spending more than $1 million annually on its observability platform. More AI workloads means more complexity to monitor, more logs to analyze, more security threats to detect. Snowflake's growth re-accelerated to nearly 30% as enterprises recognized that data infrastructure is the foundation AI needs before it can do anything useful. CrowdStrike climbed over 50% because AI doesn't reduce cybersecurity threats — it creates entirely new attack surfaces that need defending.
Even ServiceNow, which straddles the line between application and platform, generated over $600 million from its AI assistant products alone and grew subscription revenue 21% by positioning itself as an "AI Control Tower" for enterprise workflows — not competing with AI, but becoming the orchestration layer that AI agents operate within. Notably, ServiceNow's retention rates haven't taken a hit yet, which may be an early signal that well-positioned platforms can weather this storm.
The pattern is clear: the companies winning aren't the ones bolting AI features onto existing products. They're the ones whose core infrastructure becomes more essential as AI adoption scales.
Yet many of these infrastructure companies are still being sold off alongside the ones AI is actually displacing, simply because they carry the "software" label.
The financial feedback loop is making things worse
What makes this moment especially treacherous isn't just the technology thesis — it's the credit cycle layered on top of it. Business development companies have roughly $100 billion in exposure to software companies. As software valuations decline, BDC balance sheets deteriorate. As BDCs tighten credit, software companies lose access to growth capital. As growth slows, valuations fall further.
This dynamic doesn't discriminate. A company with strong fundamentals and growing revenue can get caught in the same credit squeeze as one that's genuinely being disrupted, simply because both carry the "software" label. Default rates in private credit could reach 13% if AI disruption plays out aggressively, according to UBS — a projection that makes lenders cautious across the board, not just with the most exposed borrowers.
What the bears are getting right
The fundamental insight — that the marginal cost of creating software has collapsed — is correct and profound. This is not a temporary dislocation. When anyone can build a functional application in hours instead of months, the structural economics of the industry change permanently.
The value shifts away from the application itself and toward the underlying data, the integrations, the operational complexity, and the reliability of the systems that power it. Software that survives long-term will be software that's hard to replicate — not because of its UI, but because of the engineering depth and infrastructure moats it embodies.
That's a healthy and overdue reckoning for parts of the industry, even if the process of getting there is painful.
What the bears are getting wrong
The timeline is being compressed unrealistically. Yes, AI can generate a basic application from a prompt. No, that does not mean enterprise software disappears next quarter. Adoption curves, procurement cycles, compliance requirements, integration complexity, and organizational inertia all mean that even genuinely disrupted categories will take years to fully turn over. Markets are pricing in a revolution that will actually unfold as an evolution.
The all-or-nothing framing is also creating mispricing in both directions. Some companies will see their growth accelerate because of AI adoption. Others will see specific product lines threatened while their core platform becomes more essential. Painting every software company with the same brush guarantees you'll be wrong about most of them.
What the AI optimists are getting wrong
On the flip side, the unbounded enthusiasm deserves its own reality check. Not every AI demo translates to an enterprise deployment. Not every proof of concept survives contact with production data, regulatory requirements, and organizational change management. The gap between "this is technically possible" and "this is deployed at scale in a Fortune 500" is still measured in years for most use cases.
We've seen this pattern before. Cloud computing was genuinely transformative, but the timeline from early hype to mainstream enterprise adoption was roughly a decade. Mobile, same story. AI will be faster because the infrastructure is better, but "faster than previous platform shifts" is not the same as "instantaneous."
Companies making long-term bets based on the assumption that every AI promise will be fulfilled on schedule are just as exposed as the ones ignoring the threat entirely.
Cooler heads will prevail
I expect the next 12-18 months to be painful but ultimately clarifying. Ball makes a smart observation: what will change the market's mind is several quarters of stable retention rates from established software companies in the face of AI challengers. ServiceNow's early Q4 results suggest that's possible for well-positioned platforms. If more companies demonstrate that retention is holding, the panic-driven repricing will start to correct.
The market will develop more precision in how it evaluates software companies, distinguishing between those that are genuinely in AI's path and those that are being caught up in category-level panic. The early data is already here: the 2025 stock performance gap between infrastructure winners and application-layer losers was stark, and that divergence will only sharpen as earnings continue to separate reality from narrative.
Infrastructure companies will eventually get re-rated as the market recognizes that AI workloads don't reduce demand for data movement, real-time processing, and distributed systems — they dramatically increase it. Application-layer companies will bifurcate sharply between those that integrate AI successfully and those that don't.
And the credit cycle will unwind on its own timeline, unfortunately causing collateral damage to strong companies that happen to carry the wrong label.
My advice to anyone building or investing in this space: resist the urge to react to the loudest narrative, whether that's doom or unbounded optimism. Focus on fundamentals — retention, efficiency, genuine technical differentiation, and whether your product becomes more or less essential as AI adoption grows.
The companies that panic-rebrand as "AI-native" overnight will look desperate in hindsight. The ones that quietly build indispensable technology will look prescient. And the investors who maintain discipline while others oscillate between euphoria and panic will be the ones who capture the real value being created right now.
The sky isn't falling. But it is changing shape. The winners will be the ones who study the new landscape carefully rather than running for cover or chasing mirages.
Top comments (0)