Two trillion dollars.

That’s roughly how much market cap evaporated from traditional SaaS companies in a roughly 30-day window spanning February and March 2026. Atlassian dropped 35%. Salesforce dropped 28%. Marketing automation platforms that were charging $3,000 a month discovered that AI agents could do the same work for $200.

The tech press called it the SaaSpocalypse.

I’ve been thinking about it differently: it’s the biggest opportunity redistribution in software in a decade.

What Actually Happened

Per-seat SaaS was built on a specific assumption: that human workers needed workflow software to do their jobs. One seat per human, billed monthly. The entire pricing model required humans in the loop.

AI agents don’t have seats. They don’t need a license to access a CRM. They execute workflows autonomously, across multiple systems, without a human clicking through a UI.

The moment enterprise buyers realized they could replace a $3,000/month platform with an agent doing the same job for $200, the math changed instantly. CFOs started canceling contracts. The $2 trillion evaporated in weeks.

What Survived

Not everything collapsed. A few categories came through cleanly:

Vertical SaaS — tools so deeply embedded in a specific industry’s workflow that they become infrastructure. Healthcare software that talks to specific EMR systems. Legal tools that understand jurisdiction-specific document formats. These aren’t generic workflow managers; they’re domain specialists with years of industry-specific integrations.

Consumption-based pricing — companies that had already moved off per-seat models and onto credit/usage billing. HubSpot, for example, actually reported revenue growth during the SaaSpocalypse period by pivoting to an agent-credit model. They weren’t selling seats; they were selling outcomes.

Agent-native startups — companies that were built from the beginning for AI-first workflows. They didn’t have legacy per-seat revenue to protect, so they never had to choose between defending old pricing and embracing the new reality.

The Opportunity

Here’s the part no one is writing about: $2 trillion in enterprise buyer budget just became available.

Those buyers didn’t stop needing the capabilities those SaaS tools provided. They stopped paying for the legacy delivery mechanism. They’re actively looking for replacements that fit the new model — agentic, consumption-based, vertically specific.

For a solo builder, this is genuinely good news. The legacy SaaS companies that are struggling have two problems they can’t solve quickly: they have hundreds of employees dependent on existing revenue streams, and they have codebases built for human-in-the-loop workflows that don’t translate cleanly to agent architectures.

A solo builder has neither problem. You can ship a focused, agent-native replacement for a specific niche that a legacy platform served, price it at $200/month instead of $3,000, capture the buyers fleeing the incumbents, and build on modern pricing from day one.

The formula isn’t complicated:

  1. Find a legacy SaaS vertical where buyers are visibly unhappy right now
  2. Identify the core workflow they actually use (not all 400 features, the one job)
  3. Build an agent that does that job
  4. Price on consumption, not seats
  5. Reach those displaced buyers through their community

The Trap to Avoid

The temptation is to build something horizontal — an “AI agent that does anything.” That’s how you end up competing with OpenAI and losing.

The SaaSpocalypse winners are vertical specialists. Not “AI for sales.” AI for insurance underwriting. Not “AI for HR.” AI for onboarding at companies with specific compliance requirements.

The more specific the niche, the less competition, the higher the willingness to pay, and the easier the distribution. The buyers in a niche are concentrated — they hang out in the same forums, attend the same conferences, follow the same thought leaders.

$2 trillion just changed hands. The question is just which specific pocket of it you’re going to go after.

The research data behind this post came from tracking TechCrunch, Outlook India, and the SaaStr AI predictions thread published in early March 2026.