Wall Street's SaaS Panic: Is the Software Boom Actually Over?

Wall Street is having a moment. And by moment, I mean a full-blown existential crisis about whether the entire software industry is about to get vibe-coded out of existence.

Salesforce down 21%. Snowflake down 23%. HubSpot down 36%. These aren't minor corrections—this is what traders are calling "get me out" style selling. Jeffrey Favuza, who works an equity trading desk at Jefferies, put it plainly: "We call it the SaaS apocalypse."

The trigger? A legal plugin for Claude and some impressive demos of people building their own versions of enterprise software in an afternoon. Suddenly, investors are asking whether companies need to keep paying thousands per seat for software when AI can just... make it.

The Vibe Coding Revolution (Maybe?)

Here's where things get interesting. CNBC anchor Deirdre Bosa tweeted that she woke up one morning and decided to recreate Monday.com with Claude for fun. An hour later, she had a working prototype plugged into her calendar and Gmail that surfaced a kid's birthday she'd forgotten about. She added a caveat: "To be clear, me being able to vibe code a personal tool is not going to disrupt the software trade, but someone who does know what they're doing very well might."

Except... what if it is? Chris Pasarski from Y Combinator shared that one of their portfolio companies just got off a demo with a prospect who'd built internal sales workflows with Replit to replace a SaaS tool they were currently paying for. The prospect was non-technical. And Pasarski is seeing this "more and more over the last 2 months."

The math here makes investors nervous. Casey Smith outlined what he calls "the great SaaS meltdown" in three parts: the market has stopped rewarding growth-at-all-costs, inference costs are squeezing traditional high margins, and there's what he calls "the seat crisis"—why pay for 100 seats when AI lets 10 people do the same work?

Enterprise Reality Check

But hold on. Because there's another side to this, and it's coming from people who actually work inside large companies.

Prompt engineer Claude (yes, really) writes: "'SaaS is dead' says someone who's never stepped foot in a company with more than seven people." James Blunt expanded the thought: "Large enterprises don't run on apps. They run on decades of layered systems, ERP, mainframes, custom services, data warehouses, compliance controls, and fragile integrations nobody dares touch without a 12-month change plan."

Nvidia CEO Jensen Huang went even harder, calling the market "just plain wrong." His thought experiment: if you were AGI (the ultimate AI), would you reinvent every tool from scratch, or just use the screwdriver that already exists? He argues it's obviously the latter. HubSpot founder Dharmesh Shah was so enthusiastic about this framing that he wished he could give it "100 likes."

And there's real-world evidence for this skepticism. Clara CEO Sebastian Simikowski famously tried to replace Salesforce with custom-built alternatives. A year later, he wrote: "I don't think it's the end of Salesforce. It might be the opposite." His company's specific needs made the experiment worthwhile, but he doubts most companies will follow suit.

The Nuance Nobody Wants to Hear

The truth is probably somewhere in the deeply unsatisfying middle. Investor Chia Wang has a framework I keep coming back to: "AI makes strong software companies stronger and weak software companies weaker. This is because the moat of strong software companies was never software but rather distribution, proprietary data, workflow integration, enterprise lock-in, network effects, trust and compliance, etc. Whereas the moat of weak software companies was just software."

In other words, if your only value prop was "we wrote some code that does a thing," you might be cooked. But if your value is in your relationships, your data, your integrations, your compliance certifications—you're probably fine. Maybe even better positioned to adopt AI agents into your existing product.

Ben Thompson points out another angle: "While they can write infinite software thanks to AI, so can every other software company." The real issue isn't that companies will stop using software—it's that they might not want to buy more of it. They need to save budget for their own AI experiments, for tokens, for whatever comes next.

Wall Street Journal's Dan Gallagher wrote a piece with a perfect title: "AI Won't Kill the Software Business, Just Its Growth Story." Even if SaaS survives, the days of automatic revenue expansion might be over. That changes valuations a lot, even if it doesn't kill companies.

What's Actually at Stake

Investor Gokul Rajaram raised something that made me pause. Even if Jensen Huang is right that AI will just use existing tools, there's a catch: "AI chooses the tool to use. An AI will select the optimal tool based on several criteria which change over time. This doesn't bode well for a specific piece of software having a long-term relationship with a customer."

In other words, vendor lock-in dissolves when AI is making the purchasing decisions. If your AI assistant can switch you to a competitor every quarter based on price or features, that's commoditization by another name.

John Lober offered the most optimistic take I've seen: maybe we'll finally get good software. Most B2B SaaS, he notes, "was never good. Most software is hideously broken. Terrible UX, bugs, and obvious functionality that's missing." If AI forces companies to compete on quality instead of just extracting value from market position, we might end up with a similarly structured SaaS market—just with software that's actually pleasant to use.

That's... honestly kind of compelling?

Markets vs. Reality

One thing to remember: we've been here before. There have been a dozen AI-related selloffs since ChatGPT launched in late 2022. They ping-pong between "AI is overhyped" and "AI is going to disrupt everything." This is the first time we've seen markets try to price in the death of an entire sector, but that doesn't mean they're right.

Markets move on expectations and fear. Enterprise architecture moves on 12-month change approval processes. Those are very different timelines.

Still, something is shifting. The specific question of whether software is dead matters less than what happens next: which companies adapt, which pricing models survive, and whether this forces the industry to actually build better products. The disruption is real. The timeline and scope? That's what everyone's trying to figure out—usually while panic-selling their positions.

Yuki Okonkwo