If you’ve looked at your stock portfolio this week, you’ve seen the carnage. Last week alone, the market rout wiped out approximately $2 trillion from software, data, and exchange stocks. The narrative driving this sell-off is simple, terrifying, and pervasive: Generative AI, specifically the new wave of “long-running” agents from Anthropic, is about to obliterate the SaaS business model.
The fear is that AI agents will not only replace the human “seats” that SaaS companies charge for, but will also be able to “vibe code” bespoke software solutions on the fly, turning massive enterprise platforms into commoditised databases. To put this in context, the public sentiment about software stocks hasn’t been this low since the 2000’s bubble burst.
But before we write the obituary for the entire software asset class, we need to look at the data. While the disruption is real, the market’s reaction looks increasingly like a historic overcorrection driven by fear rather than fundamentals.
The Trigger: The “Long-Running” Agent
The catalyst for this week’s bloodbath was Anthropic’s release of upgraded capabilities for Claude, specifically “Cowork” and “Claude Code” (but also the open source OpenClaw personal agent). Unlike previous chatbots, these are “long-running agents” capable of executing complex, multistep tasks, like organising files, synthesising research, or writing code—over extended periods without constant human hand-holding.
Investors looked at tools that can autonomously automate legal drafting or coding and immediately dumped shares of LegalZoom (-20%), Intuit (-11%), and EPAM Systems (-13%). The logic is linear: if an agent can do the work of a junior lawyer or a developer, companies will buy fewer software licenses, and the “seat-based” pricing model that has powered the last decade of venture returns will collapse.
The Private Equity Contagion
Adding fuel to the fire is the massive exposure of private equity. Firms like Thoma Bravo and Vista Equity Partners spent the last decade loading up on software companies, betting on their “sticky” recurring revenue. This culminated in a deal frenzy where valuations doubled between 2012 and 2022.
Now, the bill is coming due. Lenders are getting nervous about “disruption risk,” with major players like Blackstone and Apollo seeing their own shares slide as the market questions the value of the software debt they hold. We are already seeing cracks; Thoma Bravo, for instance, had to cut the acquisition price of Verint by a third last year as AI fears spooked credit markets. If PE firms can’t exit these positions or refinance their debt because the terminal value of software is in question, we could see a liquidity crunch that spills over into the broader tech ecosystem.
Why This Is Likely an Overcorrection
Despite the grim headlines, there is a strong case to be made that the pendulum has swung too far. According to a new analysis by Jefferies, software sentiment has hit lows not seen since the 2008 Global Financial Crisis or the Dot-Com crash. When sentiment hits these extremes, it often signals a dislocation between price and value.
Here is why the “Software is Dead” thesis might be wrong:
1. The “System of Record” Moat The bear case assumes that because AI can write code, it can easily replace enterprise platforms. This ignores the reality of the “System of Record.” Incumbents like Salesforce, ServiceNow, and Workday don’t just provide a UI; they own the proprietary data and the complex, entrenched workflows of the Global 2000. As Orlando Bravo recently noted, the franchise value of these companies is their deep domain expertise and integration, which is not easily replicated by a generic LLM wrapper. It is incredibly difficult to displace a trusted solution that holds a company’s financial or customer data with a “vibe-coded” app, regardless of how good the code is.
2. Pricing Models Will Evolve, Not Vanish The fear of “seat compression” (fewer humans = fewer licenses) is valid, but it assumes software companies will stand still. We are likely moving toward a hybrid model of seats plus consumption. If an AI agent makes a customer service rep 10x more productive, the software vendor captures that value through outcome-based pricing (e.g., charging per resolved case) rather than just per user. Jefferies argues that AI could actually be accretive to margins in the long run by reducing development costs and increasing the value delivered per employee.
3. Incumbents Are “Agentic Factories” Far from being sitting ducks, incumbents are aggressively embedding AI. Vista Equity’s Robert Smith describes his portfolio as becoming an “agentic factory,” integrating agents to make their specific software critical for the AI era. We are seeing a divergence between Platforms (Systems of Record) and Point Solutions. Point solutions that are merely “thin wrappers” (like basic legal drafting tools or simple schedulers) are indeed in the “kill zone”. But platforms that integrate these agents into a secure, compliant ecosystem will likely capture more spend, not less.
The VC Takeaway
Sam Altman recently warned that we are entering the “fast fashion era of SaaS,” implying shorter life cycles for software products. But he also noted that companies owning the “system of record” will remain “incredibly valuable”.
For VCs, the lesson here isn’t to abandon software, but to change the filter. The era of funding “forms over databases” is over. The alpha now lies in:
- Infrastructure & Security: As agents proliferate, we need the “guardrails” to manage them. Cybersecurity firms like Palo Alto Networks and CrowdStrike are becoming even more critical as the threat surface expands.
- Vertical AI with Data Moats: Startups that don’t just generate text/code, but possess proprietary datasets that generalist models like Claude cannot access.
- The Oversold Incumbents: With 42% of software stocks trading near historical trough valuations, there are likely bargains in the public markets for companies that have been punished indiscriminately despite having sticky retention and real cash flow.
The market is currently pricing in a scenario where AI replaces all software. History suggests technology usually expands the TAM rather than destroying it. The winners will be those who can transition from selling tools to humans, to selling platforms for agents.
I'm a former general partner at google ventures, head of Google for Entrepreneurs in Europe and founding head of Campus London, Google's first physical hub for startups.
I'm also the founder of Techbikers, a non-profit bringing together the startup ecosystem on cycling challenges in support of Room to Read. Since inception in 2012 we've built 11 schools and 50 libraries in the developing world.
- Weekly Firgun Newsletter – February 27, 2026 - February 27, 2026
- The Rise of the Personal AI Agent - February 26, 2026
- Pre-Seed Is Stronger than ever, but less startups are getting funded - February 25, 2026
