Software Lost $400B Last Week. Should You Care?

Four hundred billion dollars evaporated from software stocks last week. The iShares software ETF dropped 8.7%, compounding a 7.4% loss from the week before. By Friday, the sector was dramatically underperforming the S&P 500, and the headlines were declaring the "death of software."

I've seen this movie before. Different soundtrack, same plot.

In 2000, it was "brick-and-mortar is dead." In 2008, it was "finance is dead." In 2016, it was "retail is dead." The pattern is always the same: a real change happens, markets overreact, everyone scrambles to explain why this time is actually different, and eventually reality settles somewhere between the hype and the panic.

But here's the thing—this time, there's an actual mechanism worth understanding, even if the panic is premature.

The CEOs Say Everything Is Fine

Box CEO Aaron Levy told CNBC the narrative "somewhat misunderstands this idea of where companies tend to spend their resources and their time and energy." His argument: companies will keep their SaaS subscriptions and build AI agents on top of them. In the 20-year history of cloud software, Levy said, "This is the most exciting moment we've ever had."

Salesforce's Marc Benioff was singing the same tune. AgentForce is apparently the fastest growing product in the company's history. Plus, "we've got all the customer data"—that moat should hold.

Look, I get it. When your stock is tanking, you don't go on CNBC and say "yeah, we're probably screwed." But these aren't just defensive statements. There's a plausible scenario where they're right. The incumbents have distribution, integrations, compliance frameworks, and most importantly, they already have access to enterprise data. That's not nothing.

What's interesting is what customers are saying. The Wall Street Journal talked to Authentic Brands' chief digital officer about the newly released Claude plugin for legal work—one of the catalysts for last week's selloff. He said the company already has software for contract review. But he added: "Everyone feels very empowered to raise their hand and say, 'Hey, how can we fold this in?'"

That sentence is the whole story. The friction for experimentation has dropped meaningfully. Companies aren't ripping out their existing software—yet. But they're a lot more willing to try alternatives than they were six months ago.

The Valuation Problem Nobody Wants to Talk About

Brad Gerstner of Altimeter Capital offered the most honest explanation I've seen for the dramatic drop. Even if you don't believe SaaS is doomed, software has been the best performing sector for two decades based on extremely strong, predictable growth. Software stocks were trading at 30 to 35x profit, which assumes decades of continued outperformance.

Gerstner's point: "AI disruption lowers predictability of future cash flows." If AI contracts that horizon from multiple decades to a single decade, the math changes dramatically. You don't have to believe software is dying to understand that uncertainty about future growth rates should compress valuations.

That's the unsexy truth behind the selloff. It's not that investors think Salesforce will be gone in five years. It's that they're less confident about what growth looks like in years six through twenty.

The Seat Model Is Genuinely Weird Now

Here's where things get interesting. The traditional SaaS model charges per seat—per user. That made perfect sense when you were selling tools for humans to use. But as Eric Goldhar pointed out: "We used to buy software for humans to use. Now we buy agents to do the work. If your product charges by the user, you're selling a tax on productivity."

That's not a hypothetical problem. If companies are laying off workers and replacing them with AI agents, SaaS companies literally have fewer seats to sell. Even if the incumbent platforms survive, the per-seat pricing model looks increasingly ridiculous.

When Sam Altman was asked "Is software dead?" he gave a characteristically oblique answer: "Every company is now an API company, whether they want to be or not." Translation: the business model of software has fundamentally changed. Software isn't dying—it's becoming infrastructure.

Thompson Reuters: Good Numbers, Bad Vibes

Thompson Reuters is a perfect case study in how none of this is straightforward. Their stock dropped 20% last week after that Claude legal plugin launched. But their actual financials? Seven percent revenue growth, 9% profit growth, margins up to 44.3%. CEO Steve Hasker told investors: "We are seeing tangible benefits from our continued investments in AI."

Morning Star maintained their fair value assessment, suggesting the stock is now massively oversold. Morgan Stanley slashed their price target by a third. Same company, same earnings, wildly different interpretations.

The shift isn't from "will this AI stuff work?" to "will pricing power and renewals hold up as lookalike features spread?" That's a very different question, and it applies pressure even to companies executing well.

Security: The Actual Opportunity

Microsoft lost $218 billion in market cap last week, but their internal response is instructive. According to The Information, commercial CEO Judson Althoff circulated a memo telling sales staff to emphasize Microsoft's edge in agent security and compliance when competing against OpenAI's new platform.

Meanwhile, security researchers found that around 400 skills in OpenAI's Claw Hub contained malicious code—including some of the most popular ones. OpenAI partnered with VirusTotal to scan uploads, but their own blog post admits it "won't catch everything."

This is where I actually agree with the optimists: if you're a security engineer, this is going to be a very good year. The agent ecosystem is moving faster than security can keep up, and enterprises are rightfully terrified of that gap. The companies that can credibly solve agent security will have real pricing power.

What's Actually Happening

The SaaS apocalypse narrative is overblown, but the underlying pressure is real. Software isn't dying—it's repricing. The certainty that made software stocks trade at premium multiples is gone. The per-seat model that worked for two decades looks increasingly awkward. And the friction for customers to experiment with alternatives has dropped dramatically.

That doesn't mean Salesforce and Microsoft are going bankrupt. It means the next decade won't look like the last two. Which, if you've been in this industry long enough, is just Tuesday.

The real question isn't whether software survives—of course it does. The question is what the business model looks like on the other side of this transition, and which companies figure that out before their competitors do.

—Mike Sullivan, Technology Correspondent