A star investor who just crushed 99% of his rivals is ringing the loudest alarm yet on the software industry. Nick Evans, who runs Polar Capital’s $12 billion global technology fund, has dumped almost every application-software stock in his portfolio — including Salesforce, Adobe, ServiceNow, SAP, and HubSpot. His verdict: AI isn’t just disrupting the sector. It’s threatening to wipe most of it off the map.
Nick Evans’ Warning on Software Stocks
- Existential AI threat: Application software faces wipeout as advanced AI agents replicate workflows, replacing rigid apps with natural-language interfaces and universal brains.
- Massive market fallout: $1T+ erased from software valuations in 2026; stocks like Salesforce, Adobe, ServiceNow down >20%, with heavy short-selling.
- Survival outlook grim: Most firms won’t adapt fast enough—only a handful (data pipes or AI leaders) survive; Evans dumped holdings, advising underweight positions urgently.
The numbers are brutal. U.S. software stocks have lost more than $1 trillion in market value since late January. An ETF tracking the sector is down over 21% in 2026 alone. Hedge funds are piling into shorts — Asana has 25% of its float sold short, TeraWulf more than 35%. Even Microsoft, usually a safe haven, is now seeing shorts pile in on weakness. Oracle, Salesforce, Adobe, and ServiceNow have each dropped more than 20% this year.
So what exactly is happening? Here’s the why and the how — explained without the hype.
Table of Contents
The “Why”: AI Is Eating the Entire Application Layer
Traditional enterprise software solves specific problems with rigid interfaces: log into Salesforce for CRM, Workday for HR, Adobe for creative tools. These apps are expensive, complex, and require training.
AI agents flip the script. You no longer need ten different logins and dashboards. You describe what you want in plain English — “analyze last quarter’s sales, flag risks, draft the board report, and book follow-up meetings” — and a single AI system does it all by pulling data from everywhere.
The moats that made software giants rich are crumbling:
- User interfaces → replaced by natural-language conversation
- Workflow automation → handled by AI agents that chain tools together
- Data silos → bypassed when models can access raw data directly
Snowflake CEO Sridhar Ramaswamy put it bluntly on the Big Technology Podcast: the big AI labs want “all of the data for all of the enterprises easily available to them.” Once that happens, “everything else… is just a dumb data pipe that feeds into that big brain.”
In other words, your $200-per-user-per-month CRM license becomes unnecessary when Claude, GPT, or whatever comes next can act as a universal enterprise brain.
Evans draws the newspaper parallel: the internet didn’t kill journalism — it killed the old distribution and monetization model. Most newspapers never recovered. He believes the same fate awaits most application-software vendors.
The “How”: The Mechanics of the Takeover (It’s Already Started)
- Agentic AI + Plugins
Anthropic’s recent Claude updates introduced industry-specific “Cowork” agents that can read documents, run compliance checks, analyze financials, and manage CRM workflows — tasks companies currently pay billions for in SaaS licenses. One plugin launch reportedly wiped $285 billion from software and data-provider stocks in a single day. - Retrieval-Augmented Generation (RAG) + Enterprise Data Access
AI models no longer hallucinate when they can pull live data from your Snowflake, Databricks, or database via secure APIs. The software vendor’s role shrinks from “application” to “data provider.” - All-in-One AI Platforms
OpenAI’s Frontier, Anthropic’s agent teams, and similar offerings from Google and others are racing to become the single interface for business. Customers may simply cancel their SaaS subscriptions and route everything through one AI “brain.” - Rapid Model Improvement
Evans’ key warning: “react swiftly, as the models improve and disruption accelerates.” Each new leap (better reasoning, longer context, cheaper inference) makes yesterday’s specialized software less necessary.
But Will It Really Kill 99%?
This is where the debate gets interesting. Some argue the fear is overdone:
- Incumbents are frantically adding AI features (Salesforce’s Agentforce, Adobe’s Firefly, ServiceNow’s AI agents).
- New software will still be needed to manage, secure, and govern the AI systems themselves.
- Data quality, compliance, and integration complexity could keep human-built layers alive.
Evans isn’t buying it. He’s not waiting to see who adapts. He’s already out — except for a small Microsoft position and some call options — and says most software shares remain “toxic.”
The Bottom Line
The 2026 software rout isn’t just another tech correction. It’s the market pricing in a once-in-a-generation platform shift: from dozens of specialized applications to a handful of ultra-powerful AI brains that consume data directly.
Whether 99% of today’s software companies actually disappear is still an open question. But the direction is unmistakable. The companies that survive will be the ones that either:
- Own the data pipes (Snowflake, Databricks), or
- Become the AI brains themselves (OpenAI, Anthropic, Google, Microsoft), or
- Build irreplaceable governance, security, and orchestration layers around the new AI stack.
Everyone else? They risk becoming the next Blockbuster — profitable until the day they suddenly weren’t.
The models are only getting better. The window to react is closing. As Nick Evans just proved with his own portfolio, sometimes the smartest move is to get out before the crowd even realizes the game has changed.