Linda Park

The Market Is Punishing SaaS Stocks Because Nobody Trusts the AI Story Anymore

March 21, 2026

I want to say something that might be unpopular: the market is not wrong this time.

I know. Everyone on CNBC is calling it panic. Bank of America called it an "indiscriminate selloff." The CEO of Arm Holdings used the phrase "micro-hysteria." Wedbush published a note saying the crash reflected an "Armageddon scenario for the sector that is far from reality." Smart people with expensive suits are lining up to tell you that this is all overblown.

But here's what I keep coming back to: the market has been watching SaaS companies promise AI transformation for three years, and at some point, the check comes due. I think that point was January 2026.

What Actually Happened

Between mid-January and mid-February 2026 alone, approximately one trillion dollars was wiped from the collective value of software stocks, with the S&P North American Software Index posting its worst monthly decline since the 2008 financial crisis. That is not a small number. That is not a rotation. That is the market deciding, collectively and loudly, that it has been waiting long enough.

Since the iShares Expanded Tech-Software Sector ETF peaked on September 19, 2025, it has fallen roughly 30%. For context, the broad technology indexes like XLK and QQQ are essentially flat over the same period, and the semiconductor ETF is up 30%. Software didn't fall because the economy cratered. It fell because something specific changed in how investors think about what these companies are worth.

The catalyst was a series of AI product launches, most notably Anthropic's Claude Cowork tool and OpenAI's enterprise agent, Frontier, demonstrating that AI agents can now handle complex knowledge work autonomously. The market's interpretation was simple: if AI agents can replicate what enterprise software does, then enterprise software is finished.

That's the story everyone ran with. And it's only half right.

The Per-Seat Problem Is Real and It's Been Real for a While

The piece of this I think people are underselling isn't the AI agent threat - it's what AI agents reveal about the underlying business model. If 10 AI agents can do the work of 100 sales reps, you don't need 100 Salesforce seats anymore. You need 10. That's a 90% reduction in seat revenue for the same work output.

The per-seat pricing model - the engine of SaaS revenue growth - depends on a simple equation: more employees equals more seats equals more revenue. AI agents break this equation. When one agent can handle the work of five CRM operators, the customer doesn't need five seats anymore. The revenue-per-customer metric that investors have relied on to value SaaS companies is being compressed from both sides: fewer new seats and existing seat reduction.

I've been noticing this from the buyer's side for two years. Our renewal quotes keep going up. I wrote about that earlier this year - there's a genuinely bizarre gap between what SaaS stock prices are doing and what the companies are charging their existing customers. The vendors figured out that if they couldn't grow headcount-based revenue, they'd grow it through price increases on the people already locked in. The "growth" is mostly vendors raising prices on captive customers. That works until it doesn't. And it doesn't work when AI gives those captive customers alternatives.

Gerald asked me at Costco - we were waiting for samples, the usual - whether the software companies we use were going to be around in five years. He doesn't follow any of this closely, but he'd read something on his phone. I said probably, but they'll look different. He nodded like that answered it. I've been thinking about that conversation since January.

The AI Story Broke Because It Was Always a Story

Here's my actual opinion: the SaaS sector got caught doing something it shouldn't have been doing, which is promising AI transformation as a growth narrative before it had AI transformation as a product reality.

Customers say they would prefer to buy AI-enabled solutions from their incumbent vendors. They trust them. They know they are secure. They believe they will be around for the long term. But most incumbents have yet to deliver compelling offerings or prove they can win this new spending.

That is the crux of it. The customers wanted the AI products from the vendors they already use. The vendors promised those products. The vendors did not deliver those products in any meaningful way. And then Anthropic and OpenAI showed up with products that actually work. So now investors are asking: wait, who's actually building AI, and who's been doing investor relations?

Even when stalwarts such as ServiceNow, SAP, Salesforce, Adobe, and others post solid quarterly results, investors focus on the massive investment these companies are making in their AI capabilities and expect to see revenue and growth that can be directly tied to these expenditures. That's a fair ask. If you've been telling me for two years that AI is going to transform your product, I want to see that in the revenue line. Not in the press releases. In the numbers.

On January 29, 2026, US software stocks suffered a steep selloff after SAP's cloud revenue forecast and ServiceNow's earnings disappointed investors, intensifying fears that AI-driven automation is eroding the value proposition of traditional SaaS offerings. SAP shares plunged more than 16% after its cloud backlog and 2026 revenue forecast missed expectations, while ServiceNow fell 11% despite projecting strong subscription growth. The S&P 500 Software and Services Index dropped 8.7% to a nine-month low, with other major players like Salesforce, Adobe, Datadog, Microsoft, Atlassian, Zscaler, Intuit, and HubSpot also posting significant declines.

ServiceNow fell 11% while projecting strong subscription growth. Read that again. The market looked at a company growing subscriptions and still sold the stock. That's not panic. That's a structural repricing of what subscriptions are worth when you're not sure you need them in three years.

What the Numbers Actually Say About the Future

Now here's where I'll push back on myself a little, because intellectual honesty matters.

Gartner's February 2026 forecast projects worldwide software spending will grow 14.7% in 2026 to more than $1.4 trillion, accelerating from 11.5% growth in 2025. That represents roughly $180 billion in net new software spending in a single year. Global SaaS spending specifically is projected to rise from $318 billion in 2025 to $576 billion by 2029, according to Forrester.

So people are still buying software. A lot of software. The demand isn't collapsing. But here's the split that I think the overall numbers obscure: investors are fleeing because of fears that SaaS companies won't be the provider of choice for AI agents; their traditional per-seat model will become obsolete, leading to revenue declines; vibe coding will allow startups to replicate the features of complex SaaS platforms, eroding their moat; and SaaS products are fundamentally too complex, and users struggle to manage the SaaS sprawl of hundreds of applications that don't talk to each other.

That list is not entirely wrong. I manage our team's SaaS stack and I can tell you - the sprawl problem is real. We have tools that don't talk to each other, tools we pay for that three people use, tools that got added because someone was excited in 2021. When I look at something like what we actually spend on project management software, the question of whether any of this is being rationalized by AI is absolutely on my mind.

Tory mentioned last week - in his relentlessly optimistic way, despite the fact that his car is apparently still impounded - that he'd been using an AI agent to do stuff he used to do in three different apps. He didn't cancel those apps yet. But he will. That's exactly the dynamic the market is pricing in.

The Companies That Are Actually Fine

Not every SaaS company is in the same position, and I think the indiscriminate selloff framing has a point here even if I disagree with the overall "this is just panic" conclusion.

"There's likely to be cannibalization of SaaS by AI-driven workflows and that will impact the multiple the sector trades on," Rolf Bulk, tech equities analyst at Futurum Group, told CNBC. That said, Bulk argued that a subset of software providers, especially those running mission-critical enterprise workloads such as Oracle and ServiceNow, still have a sustained "right to earn." The depth of their data and entrenched role in customer workflows make them more likely to coexist with AI rather than be replaced outright.

Popular SaaS categories for investors now include startups building AI-native infrastructure, vertical SaaS with proprietary data, systems of action, and platforms deeply embedded in mission-critical workflows. But investors are quite bored by startups building thin workflow layers, generic horizontal tools, light product management, and surface-level analytics - basically, anything an AI agent can now do.

That distinction matters for people buying software, not just people investing in it. The tools that have your data, that sit inside your core operations, that would take a year to migrate off - those are different from the tools that do something an AI chat interface can now handle. The latter category is in real trouble. The former has time to adapt, if they move.

The SaaS companies struggling to raise right now are the ones that can easily be replicated. "Generic productivity tools, project management software, basic CRM clones, and thin AI wrappers built on top of existing APIs fall into this category," one investor said. "If the product is mostly an interface layer without deep integration, proprietary data, or embedded process knowledge, strong AI-native teams can rebuild it quickly."

Derek asked me this week if Disney would ever build their own CRM. He was using it as a metaphor for something about the Star Wars franchise, I stopped listening around the second minute. But the point isn't entirely wrong - big companies with proprietary data and the resources to build are starting to ask whether they need to buy everything they used to buy. That's new.

What This Means If You're Running a Business

The stock market story and the business-owner story aren't the same story, but they're related.

If you're a business buying SaaS tools right now, you have more leverage than you did two years ago. Vendors are aware that the "just add AI to the roadmap" story isn't working anymore. In many cases, the vendors themselves have not sorted out how they would get paid for those AI solutions. Which means they're negotiating differently than they were. They need to prove they're not in the dying category.

I renewed a contract last month and got a discount I didn't expect. I almost didn't ask for it. The account manager offered it before I finished my sentence about evaluating alternatives. That's a different conversation than I was having in 2023.

The sector's forward P/E has collapsed from roughly 35x at the end of 2025 to around 20x today, back to levels the market hasn't really seen since 2014. That compression is real money that vendors no longer have available for growth initiatives, sales teams, and customer acquisition. The companies with weaker unit economics are going to tighten. Some will consolidate. Some will get acquired. A few will fail. None of that is good for customers who've built workflows around those tools.

I brought casseroles to the office Tuesday and Chris asked, completely sincerely, whether he should be worried that the tools we use might shut down. He'd read something. I told him: the big names are fine, the niche ones are worth watching. He looked relieved. He doesn't fully understand what a vendor EOL means but he has good instincts about when something sounds serious.

If you use a CRM that felt a little precarious before this, it's worth doing an honest assessment of what alternatives look like now while you can make the switch without a gun to your head. Same goes for any tool in the "thin workflow layer" category. Sales intelligence is an area where the AI disruption is particularly acute - these tools are only as good as their data, and AI-native alternatives are getting more capable every quarter.

My Actual Take

The market is not having a hysterical episode. It is doing what markets do - it is pricing in a structural shift that slow-moving quarterly earnings reports can't yet capture. Public SaaS growth rates have declined every single quarter since the 2021 peak. The AI crash narrative just gave the market permission to finally re-rate what the numbers have been screaming for three years.

The companies calling it panic are, with a few exceptions, the ones whose story looks the shakiest right now. That's not a coincidence.

At the beginning of 2025, the median revenue multiple for software firms still stood above 7; today, it has dropped below 5. That's a real repricing of real expectations. It reflects the growing belief that a lot of SaaS companies built durable-looking businesses on a pricing model that AI is now quietly corroding.

Some of them will adapt. The ones with deep data, genuine workflow ownership, and the organizational speed to build real AI features - not chatbot overlays, not "AI-powered" dashboard labels - will come out of this fine. Maybe better than fine, because their weaker competitors won't survive the repricing.

The ones that were mostly selling interfaces and subscription lock-in? The market figured that out. And it's not wrong.

Gerald is going to ask me at some point whether we should change any of the tools we use because of all this. I'm going to tell him: not yet, but I'm watching. Because that's the honest answer. The story isn't over. But the AI promise chapter definitely is.