Earlier this year, Salesforce shares lost 30% of their value in a matter of weeks…
The entire move was based on a single narrative - that AI was about to make enterprise software obsolete.
The logic sounded reasonable…
New AI coding tools from Anthropic and OpenAI had gotten good enough to write real software… so why keep paying Salesforce or Workday thousands of dollars per employee… when you could just describe what you want to Claude and let it build the thing?
The SaaSpocalypse was on…
But since then something interesting has occured…
Last month Salesforce reported better sales and profits than expected, raised its outlook, and the stock jumped more than 22% the next day.
Other SaaS stalwarts like Workday, ServiceNow and Snowflake all posted strong numbers too…
Apocalypse? What apocalypse?
Because what the doommongers missed is that writing code is the easy part…
But most enterprise software work isn't writing code.
It's maintaining it… patching it… and stitching it into the tangle of other systems a company has bolted together over a decade or more.
You cannot hand an AI that decade of wiring in a sentence and expect it to rebuild the whole thing overnight. And there is a gigantic leap between building something and running it for a 5,000 person enterprise.
2 stocks in particular exemplify this…
CrowdStrike (one of my largest personal holdings) and Palo Alto Networks both ripped higher through August.
CrowdStrike jumped around 20% in a single day after earnings and PANW is up roughly 78% this year and will be joining the S&P 100 later this month.
Why those two specifically?
Because cybersecurity is mission-critical. That means software a company cannot switch off without something breaking. And no competent CFO is going to fire their security provider to save a few dollars and bet the entire company on AI-written code catching the next breach.
So here is the line I draw…
I'm still bearish on the software names that are nice-to-have rather than can't-live-without.
Those are the ones AI actually threatens, and I think the market is still too relaxed about them.
But the notion that you can vibecode away a decade of enterprise entrenchment… meaning have an AI casually rewrite it all from a plain-English prompt… was always the fantasy of people who never had to run the software they were betting against.
AI is no doubt making the software business harder… but that is a very different thing from killing it.
And a handful of names in are setting up very nicely right now… I’ll be sharing more about how I’m positioning myself later this week
Oliver
P.S.
Sam Altman himself admitted this disruption is taking far longer than he expected, for a simple reason: people just keep buying from the companies they already buy from.