The "Death of Software" Fallacy: Why I’m Buying the Panic (Again)
As a growth investor, 2022 was my trial by fire. I’m open about it: I rode the downturn all the way, and at its worst, my portfolio was down over 40%.
That experience taught me how to spot a "Market Panic" versus a "Fundamental Shift."
Right now, in January 2026, we are watching a massive, indiscriminate sell-off in software. The narrative driving it is simple, seductive, and almost certainly wrong.
The narrative is: "AI writes code for free. Therefore, SaaS companies have no moat. Therefore, margins will collapse to zero."
This is the 2026 version of the "Interest Rates will kill Tech" argument from 2022.
It has a grain of truth, but it misses the entire forest.
The "CIO Dilemma" (Why SaaS Isn't Dead)
The bear case assumes that because a company can build its own CRM with AI, it will.
This ignores the most boring, critical part of enterprise software: Liability.
A CIO does not buy Salesforce because they love the code. They buy it for the SLA (Service Level Agreement). They buy it so that when the system breaks on Black Friday, they have someone to sue. They buy it for SOC2 compliance, data governance, and security.
If an internal, AI-generated tool breaks and leaks customer data, the CIO gets fired. If Salesforce breaks, the CIO gets to yell at a vendor. That "insurance" is the moat. Companies buy SaaS to outsource liability, not just to rent code.
The Jevons Paradox of Code
The second flaw in the "Death of Software" argument is economic.
In the 19th century, economists thought that as steam engines became more efficient (using less coal), coal consumption would drop. The opposite happened. Efficiency made steam power cheaper, so we used it everywhere. Coal consumption skyrocketed.
This is Jevons Paradox. And it is happening to software right now.
AI is crashing the cost of producing code. The market thinks this means the value of software shrinks. In reality, it means the volume of software will explode.
Yesterday: Only Fortune 5000 companies could afford custom ERPs.
Tomorrow: The local bakery, the freelancer, and the HOA will have custom software stacks, because vendors can now afford to build them.
We are moving from a world of 100 million software seats to 5 billion software endpoints.
The Great Bifurcation: Seats vs. Outcomes
However, not all SaaS is safe.
The sell-off is correctly identifying a risk, but applying it to the wrong victims.
The real risk is to companies that charge Per Seat. If AI agents replace human workers, "Seat-Based" revenue collapses.
But there is a second category of software: Outcome-Based Engines. These are companies that don't charge for the human logging in; they charge for the transaction that happens.
If AI generates more emails, they process more volume.
If AI builds more apps, they serve more ads.
If AI runs the marketing, they sell the data.
For these companies, the "AI Revolution" is not a deflationary threat; it is a massive inflationary tailwind. Yet, the market is dumping them in the same bucket as the "Seat-Based" dinosaurs.
My Move
I am not selling my software stocks.
I am short "Empty Boxes", like $CRM, tools that rely on humans to input data. Quote from @satyanadella
"Business applications exist, that's probably where they'll all collapse, right in the agent era... they are essentially CRUD databases with a bunch of business logic. The business logic is all going to these agents... and once the AI tier becomes the place where all the logic is, then people will start replacing the back ends."
I am long the "platforms" that own the data and tax the transaction.
Think $APP or $ZETA: they are currently trading at a discount because of this panic, but their fundamentals are "Anti-Fragile" to AI.