Highlights
A single day in February 2026 wiped about $285 billion off software stocks, as Wall Street started pricing in what AI does to software.
Build versus buy just flipped: a custom build went from months and a dev team to hours and a prompt.
Gartner expects 35 percent of point-product SaaS to be replaced or absorbed into agent ecosystems by 2030. The generic, horizontal tools are the most exposed.
Two kinds of SaaS stay safe: the giants with scale and data gravity, and the deep specialists whose complexity is the moat. The generic middle and the thin point apps are what fade.
The software market is still growing, so this is a reshaping, not a collapse.
In February 2026, $285 billion in software stock value vanished in a single day. Around the same time, founders started describing apps to a chatbot and getting working software back by lunch, and “vibe coding” landed in the dictionary. So it’s a fair question: is SaaS dying?
The market acted like the answer was yes. Jeremy Kahn argued the opposite in Fortune, that history and economics still favor the incumbents. They’re both a little bit right.
Build versus buy is a decades-old debate, and the time and cost of custom software is what usually pushed people to buy off the shelf. AI flipped that math. When a founder can describe what they need and get a working app back in an afternoon, paying $200 a month for a tool that covers 10 percent of their workflow stops making sense.
Four forces are driving the panic. Call them the horsemen.
1. The great unbundling
Gartner expects 35 percent of point-product SaaS to be replaced or absorbed into larger agent ecosystems by 2030. The exposed ones are horizontal point solutions that do one narrow thing at scale: survey builders, schedulers, basic CRMs, simple project trackers, lightweight dashboards.
Built for everyone, they fit no one in particular. Netlify’s own team rebuilt the SaaS they used to pay for, swapping an employee-survey platform and parts of their customer-success stack for AI-built internal apps shaped to how Netlify actually works. Skyp’s Alex Shartsis built his own CRM with Claude in about 20 minutes, for $20 a month. When building custom is trivial, “good enough for the average user” loses to “built for how I work.”
2. The vibe coding wave
The people doing the replacing aren’t all engineers. In March 2025, a quarter of Y Combinator’s Winter 2025 batch had codebases that were 95 percent or more AI-generated, the number Garry Tan summed up as “the age of vibe coding is here.”
The term was coined by Andrej Karpathy in February 2025 and went mainstream fast, capped by Collins naming it Word of the Year that November. And it isn’t just toys: BCG found AI-native firms scaling to as much as $100 million in revenue with teams of 15 to 30 people. Headcount is no longer a prerequisite to scale.
The old build-versus-buy math tilted toward buy because building was slow and expensive. Now building takes hours. The SaaS tool that covers 10 percent of your needs isn’t competing with a dev team anymore. It’s competing with your lunch break.
3. The margin meltdown
SaaS got priced for a world where software was expensive to build and maintain. Per-seat licensing made sense when customers stayed put, because switching was costly and the alternative was a dev team.
Both assumptions broke at once. The selloff that opens this piece started with a wave of AI-agent launches from the big cloud and software vendors, and the market did the math in public: if agents do the work seats used to, per-seat pricing is a melting asset. The marketing agency that ran on ten people and a SaaS stack now runs on two and a set of agents producing the same output. Enterprise buyers are already questioning per-seat pricing when agents handle what used to need human seats.
The companies that haven’t repriced for the new cost structure are the ones bleeding customers.
4. The rise of bespoke
Vertical SaaS hit roughly $130 billion in 2025 and keeps growing at about 16 percent a year, faster than software overall. The harder a workflow is to generalize, the safer it is. Enterprises are shifting from off-the-shelf SaaS to AI-built custom apps for the workflows where generic tools create friction.
Deloitte’s 2026 outlook expects AI agents to mostly augment incumbent SaaS next year rather than replace it, because the established players’ deep footprints across complex workflows are hard to supplant. The new building happens around them: custom CRMs, bespoke tenant portals, dashboards that track what the operator cares about instead of what a product team decided everyone should.
What survives, and what doesn’t
The panic priced in an extinction, but the money tells a different story. Even after the selloff, Gartner still expects worldwide IT spending to grow 13.5 percent in 2026, to $6.31 trillion, with software leading. The market isn’t shrinking. It’s changing shape.
What changes shape is the middle. Two kinds of software come through fine, and they sit at opposite ends. The giants, the Salesforce-scale platforms with years of data, deep integrations, and switching costs measured in retraining whole orgs, stay put. So do the deep specialists, the tools built for workflows too complex and regulated to generalize, where the complexity itself is the moat.
What thins out is everything soft in the middle and everything too shallow to defend: the broad, generic, “good enough for everyone” tools, and the thin point apps that do one small thing. AI builds custom replacements for those over a lunch break.
The line isn’t company size. It’s switching cost and integration depth, the moat. A giant and a deep specialist both have one, scale on the one end and complexity on the other. The generic middle and the thin edge have neither.
So when build versus buy comes back around, AI has changed the default. For most of what a team runs, the thin tools, the generic ones, the 10-percent-fit subscriptions, the answer is now build. Buy is for the few categories that still hold a moat: the giants you can’t replace, the specialists you can’t out-complex.
Software has spent 40 years getting cheaper, faster, and closer to the person who uses it. AI coding finally delivers the “democratized software development” that no-code and low-code promised for years and never quite reached.
So, is SaaS dying? In a way, yes. Per-seat pricing by the month, and settling for bog-standard industry software: those days are over. But SaaS itself is evolving. It’s the age of the builder. The frontier is wide open.
