February started with a software massacre.

Adobe is down 69% from its pandemic peak, with a market cap slashed from $350 billion to $107 billion.

ServiceNow crashed 54% from its all-time high of $239 to $109.

Microsoft shed $360 billion in market value in a single session.

Salesforce, Atlassian, HubSpot, Intuit - all of these names are getting obliterated.

The iShares Expanded Tech-Software Sector ETF (IGV) plunged 27% from its highs. Over $300 billion in market value evaporated from the software sector in one trading day.

Wall Street dubbed it the "SaaSpocalypse." And unlike your typical tech correction, in which the primary culprit is usually rates or recession fears, this one's a bit different. The true threat to these businesses isn't macro. It's the disruption that comes with agentic AI.

The business model that powered 15 years of double-digit software growth is being fundamentally dismantled by AI agents. It’s happening right now.

Microsoft CEO Satya Nadella telegraphed this a year ago.

December 2024, BG2 podcast - go to the 46-minute mark.

He said business applications - CRMs, ERPs, project management tools - would "collapse" in the agent era because they're essentially "CRUD databases with a bunch of business logic."

His warning implied that business logic is moving to the AI tier. This would mean that AI agents would be the ones orchestrating workflows across multiple databases without needing the application layer. The software becomes invisible and the GUI becomes irrelevant.

At the time, this sounded like hyperbole and I thought Nadella was just being provocative or trying to make the headlines.

Then Anthropic dropped Cowork plugins on January 31, 2026. Legal workflow automation. Contract review. NDA triage. Compliance checking. All as markdown skill files sitting on top of Claude, their AI agent.

Within 48 hours, LegalZoom crashed 20%. Thomson Reuters dropped 16%. RELX (parent of LexisNexis) fell 14%. The selloff cascaded into Salesforce, ServiceNow, and across the entire software universe.

The "AI agents replace SaaS" narrative is both right and wrong. The nuance matters here and we need to be careful to make the correct distinction, because this isn’t a clear death knell for all SaaS businesses. It just means the model is changing.

The Rise of Agentic AI

Traditional SaaS is human-operated software. You log in. You click buttons. You navigate dashboards. You manually update records. It's fundamentally a tool that requires a human in the loop.

Agentic AI flips this.

Instead of you using Salesforce to update a CRM, you tell an AI agent "update the client meeting notes and flag three follow-up items." The agent accesses Salesforce's backend directly, executes the workflow, and moves on to the next task. No clicks. No dashboard. No human intervention needed.

Anthropic's Model Context Protocol (MCP) and OpenAI's Frontier platform - both officially launched in the past 60 days - enable exactly this. These are AI agents that autonomously orchestrate tasks across multiple business systems without touching a GUI.

This is the very beginning of the shift of software from "tool you operate" to "service that operates for you."

The Death of Per-Seat Pricing

SaaS companies built billions in market value on one pricing model that most businesses are familiar with. This pricing model was built on per-seat, per-month licensing. You would pay based on how many employees use the software. For SaaS companies, more users = more revenue. It was the best business model going. Simple. Predictable. Lucrative.

But AI agents don't occupy seats. An AI agent replacing 10 customer service reps doesn't require 10 licenses, just compute credits. The entire revenue model collapses.

Salesforce is scrambling to adapt with their Agentic Enterprise License Agreement (AELA) which is a flat-fee, all-you-can-eat AI pricing structure. It's a desperate pivot away from seat-based economics. And they're taking a loss on early contracts just to lock customers into the platform before they defect entirely.

Gartner estimates 40% of enterprise SaaS will include outcome-based pricing elements by 2026 (up from 15% in 2024). That means you're paying for results delivered instead of just access. You will start seeing pricing that looks like "$1.50 per AI-resolved support ticket" instead of "$150 per support agent seat."

From Wall Street's perspective, this is a nightmare for valuing software businesses.

Predictable recurring revenue will become unpredictable consumption revenue. Gross margins will likely compress significantly. Most importantly, this breaks the calculation on customer lifetime value - which assumed recurring cash flow for access with a certain amount of

Budget Reallocation to AI Infrastructure

The silent killer isn't that AI replaces SaaS. It's that AI starves SaaS of budget.

Meta announced $135 billion in AI capex for 2026. Microsoft: $75 billion annually. The four major hyperscalers combined to spend over $470 billion on AI infrastructure this year alone, according to company guidance.

That money comes from somewhere. And one of the places it’s coming from is enterprise software budgets. Gartner reports that enterprise IT spending on AI is up 110% year-over-year, while overall IT budgets grew just 8%. AI is cannibalizing SaaS seat licenses, expansion deals, and new application purchases.

CFOs are consolidating vendors. The average number of SaaS applications per company declined 18% in 2025, with 82% of companies actively reducing supplier counts, per BetterCloud's 2025 SaaS Management Index. Enterprises want fewer tools, not more. They want platforms, not point solutions. And increasingly, they're questioning whether they need SaaS applications at all when AI agents can orchestrate the same workflows across commodity databases.

Public SaaS companies that traded at 8-10x ARR multiples in 2019 peaked at 15x in 2021. By August 2025, the median settled at 6.1x revenue. But that was before the SaaS meltdown. As of today the median public SaaS revenue multiple sits at 4.0x, the lowest since 2016.

Forward P/E multiples compressed from 39x to 21x for the median software company, per analysis of 157 public SaaS firms. Adobe now trades at 12x forward earnings versus its five-year average of 30x.

Private SaaS valuations are holding up slightly better for now - with a median pricing of 6-7x ARR for companies with strong growth and Rule of 40 compliance. But deal activity is bifurcating (just like the economy). Top-tier performers with proprietary AI capabilities commanding 10x+ multiples, while generic SaaS companies struggle to get 4x.

The market's message is clear. Being a SaaS company is no longer a valuation premium. You need defensible differentiation in an AI-native world, or you're getting rerated.

Jensen Huang, Nvidia's CEO, went on record February 3, 2026 saying "The notion that AI is somehow going to replace software companies is the most illogical thing in the world."

He's right. SaaS isn't dying, just changing. The monolithic, UI-heavy, per-seat subscription model is dead. Now we as investors have to begin to understand what is going to replace it.

Salesforce spent $8 billion acquiring Informatica. ServiceNow dropped $7.75 billion on Armis. These represent a strategic repositioning for an AI-agent-orchestrated world where owning the data layer matters more than owning the interface.

The companies that survive won't be the ones with the best dashboards. They'll be the ones with:

If you're holding traditional SaaS stocks, here's how I'm thinking about positioning:

Avoid Pure-Play SaaS Without AI Defensibility

Trim or remove positions in companies that are "SaaS wrappers around databases" with weak AI strategies. That's HubSpot (down 51% YTD), Monday.com (down 44%), Atlassian (down 30+%). These are getting commoditized in real-time.

Overweight AI Infrastructure Winners

The hyperscalers spending $470B on AI aren't getting disrupted, they're the ones creating the disruption. Microsoft, Amazon, Google are all in this position as well as the shovel sellers like Nvidia, data center REITs like Digital Realty (DLR), power infrastructure plays like Constellation Energy (CEG), and many more in the infrastructure chain.

Selective SaaS Survivors With Data Moats

Not all SaaS is dead. Palantir (government/defense contracts), Snowflake (data warehousing), MongoDB (developer infrastructure), these are businesses that have structural advantages that agents can't easily replicate. But position sizing matters and it’s important as an investor to bear in mind the risks associated with the valuations many of these businesses are running at today.

The breakdown in SaaS isn't about technology destroying an industry. It's about business model obsolescence meeting margin compression meeting budget reallocation all at once.

Traditional SaaS ran on 15 years of predictable growth, expanding multiples, and compounding revenue. But the per-seat subscription model was always going to break when software stopped requiring humans to click buttons. The companies that adapt, embedding agentic AI, pivoting to outcome pricing, and most crucially building platform ecosystems, will survive. The rest are in trouble.

The Earnout Investor provides analysis and research but DOES NOT provide individual financial advice. Jamie Dejter may have a position in some of the stocks mentioned. All content is for informational purposes only. The Earnout Investor is not a registered investment, legal, or tax advisor, or a broker/dealer. Trading any asset involves risk and could result in significant capital losses. Please, do your own research before acquiring stocks.