The market has priced the destruction. It hasn’t priced the expansion. And the asymmetry between the two is the most important structural insight in enterprise software today.
The Destruction Thesis Is Linear
When an AI agent replaces a human worker, the destruction follows a simple path: replace humans → lose seats → revenue declines. One input, one output. Jasper loses a customer. That revenue evaporates. End of story.
This is why the SaaS sell-off was so violent — linear destruction is easy to model. Wall Street compressed 5 years of disruption into a 14-day sell-off because the math is straightforward: fewer seats = less revenue.
The Expansion Thesis Is Multiplicative
But here’s what the market is only now beginning to price: every agent deployed creates demand across every infrastructure layer simultaneously.
Deploy one agent:
- Data queries increase → Snowflake revenue up
- Security checks increase → CrowdStrike revenue up
- Logs and monitoring increase → Datadog revenue up
- Identity validations increase → Okta revenue up
- Compliance checks increase → regulatory systems revenue up
Deploy ten agents and multiply every line.
Application-layer destruction is one-to-one. Infrastructure-layer expansion is one-to-many.
The Pricing Model Tells You Everything
The expansion categories already survived the pricing model transition that’s killing Tier 2 SaaS:
- Snowflake charges per query
- Datadog charges per host and per metric
- CrowdStrike charges per endpoint and per module
- Okta charges per identity
Their revenue models are architecturally aligned with an agent-driven world. Revenue tied to machine usage, not human seats. The pricing transition that threatens per-seat SaaS was already completed by infrastructure companies years ago.
The Palantir Proof Point
Palantir is the most dramatic validation of the expansion thesis:
- Q4 2025: revenue grew 70% YoY to $1.41B
- U.S. commercial revenue grew 137% YoY
- Rule of 40 score hit 127%
- FY2026 revenue guidance of 61% YoY growth
CEO Alex Karp called it “indisputably the best results that I’m aware of in tech in the last decade.” Palantir’s ontology — its structured approach to making enterprise data legible to AI systems — is exactly what the agentic era demands.
The Structural Takeaway
The application layer shrinks. The infrastructure layer thickens.
The new commanding heights are not dashboards or workflows. They are:
- Data access and governance (Snowflake, Databricks, Oracle)
- Security and zero-trust enforcement (CrowdStrike, Palo Alto, Zscaler)
- Observability and monitoring (Datadog, Dynatrace)
- Identity and authorization (Okta, CyberArk)
- Deterministic systems of record (SAP, Oracle, Palantir)
Agents compress the visible software layer. They expand the invisible one.
The market that AI destroys is worth $2 trillion. The market it opens is worth three times more.
Explore the full interactive SaaS Expansion Map →
FourWeekMBA · The Business Engineer · February 2026







