Stop calling it a collapse. It isn't one.
Software stocks shed roughly $1 trillion in market cap between late 2025 and mid-2026. The headlines were predictable. "AI kills SaaS." "The subscription model is over." "Cursor just ended the enterprise software industry."
None of that is what the numbers say.
What the numbers say is simpler and less dramatic and therefore completely ignored by everyone who makes money on drama: SaaS was overpriced relative to defensibility, AI exposed that overpricing in about eighteen months, and the market is correcting. That's it. That's the whole story.
Here's what actually happened. For roughly a decade, SaaS companies traded at 10x, 15x, sometimes 20x forward revenue on the assumption that switching costs were permanent and net revenue retention above 110% would last forever. The pitch was always the same: get a customer in, expand the contract, never lose them. It worked. Until the marginal cost of building a competing workflow dropped to approximately zero.
That's the AI effect on SaaS. Not "AI replaces software." It's that AI demolished the moat-building timeline. What used to take a funded startup three years to replicate now takes a well-prompted team three months. So the premium the market was paying for "defensibility" was suddenly mispriced. Violently so.
The selloff wasn't irrational. It was late.
Ignore the noise about which specific companies are "AI-proof." That's cope dressed as analysis. What actually matters right now is the compression happening underneath the headlines. Gross margins in public SaaS are contracting by 4 to 8 points across the median cohort as AI infrastructure costs eat into what used to be nearly-free delivery. Meanwhile customer acquisition costs are climbing because every category now has three new entrants who raised on an AI narrative and will burn cash to acquire logos. That combination... margin down, CAC up... is what a maturing industry looks like. It's not a crisis. It's arithmetic.
The companies that are going to survive this aren't the ones with the best AI story. They're the ones with genuine workflow lock-in... think Veeva in life sciences, ServiceNow in IT operations, Procore in construction... where the switching cost isn't about the software, it's about the data, the compliance trail, the regulatory dependency. Those companies will compress too. But they won't evaporate.
The ones that will evaporate are the ones that were always just slightly-better spreadsheets charging SaaS multiples. You know exactly which category I mean. $40/month project management tools with no proprietary data model and no reason to exist beyond a slick onboarding flow. Those are getting repriced to zero or consolidated into platforms that have actual distribution. That's healthy. Brutally healthy, but healthy.
What irritates me about the coverage isn't the pessimism. Pessimism is fine. I'm a pessimist. What irritates me is the lack of precision. "SaaS is dying" tells you nothing. Which SaaS? At what margin profile? With what NRR? Against which competitive set? The $1 trillion number is real but it's an aggregate, and aggregates in a bifurcating market are useless for making decisions.
The bifurcation looks like this right now. Tier one: mission-critical, compliance-adjacent, data-moated software. Multiples compress but companies survive... call it 8 to 12x forward revenue long-term. Tier two: horizontal productivity software with no proprietary angle. Gets commoditized or acquired for distribution at 3 to 5x. Tier three: point solutions solving a problem AI now solves natively. Goes to zero or pivots. That's not a collapse. That's a market doing its job.
Chanos spent years being right about things the market refused to believe until it had to. The SaaS believers are in that same position now. Not wrong that the category survives. Wrong that everything inside it deserves the same treatment.
The $1 trillion wasn't panic. It was the market finally reading the fine print.