There’s no margin for error in softwareThere’s no margin for error in software
Why do investors like software stocks? Because they have high recurring revenues and extremely high margins.
Why are investors worried about the impact of AI on software stocks? At the most basic level, AI tools reduce the barriers to entry and the cost of creating software.
Nothing shows traders’ willingness to shoot first and ask questions later (or not bother to ask questions at all!) when the crux of the case for owning software seemingly shows cracks more than**** the reaction to ServiceNow’s Q1 results and updated outlook.
- ServiceNow cratered over 17% after the software company’s Q1 margins came in shy of estimates. Full-year guidance for ServiceNow’s gross and operating margins was revised lower, while subscription revenues got a big bump.
- There are some extenuating circumstances that cut both ways: integrating recently acquired businesses is the proximate cause of the expected sales bump and operating margin pressure, according to management.
- But given how important margins have been to the investment case for software stocks — and the significant profitability premium they’ve enjoyed relative to the S&P 500 as a whole — details don’t seem to matter.
In early February, Nvidia CEO Jensen Huang called the idea that the software industry would be replaced by AI the “most illogical thing in the world,” arguing that AI agents will leverage existing software tools rather than reinvent them.
The Takeaway
The bear case for software is that AI tools render many established giants obsolete. But going the way of the woolly mammoth isn’t something that happens overnight. You won’t be able to find any of them to ask, obviously, but I’m told it was a 10,000- to 16,000-year process.
Well before obsolescence comes the threat of incremental substitution. And margin pressure would be one way you’d expect competitive pressures to be absorbed. At the surface level, ServiceNow is affirming a base case for software stocks that traders have spent months fearing, which still apparently hasn’t taken the industry to levels where it’s viewed as attractively valued.
I used to need software from vendors. Now, with (for the moment) the exception of primitives, I don't. I can have everything I need (and I can code the primitives manually). The only consideration is how good your product is vs how many tokens it costs me to make it better for my bespoke use case. Remember my rant about this.
In most cases, what this means is disruption, right now. Is Salesforce abstract enough (its one of the more abstract giant softwares) or can I replace it with postgres and some vibed up interfaces and reports? Depends a bit on your skill as a software architect, but generally... postgres wins.
It won when I had teams of great coders implementing my vision and it wins now when I have just Claude and the occasional low level library I code manually. The most annoying thing to me nowadays is when someone fucks up their product by botting too much and then I have to replace their shit (for me, not for you, because I don't care about your use cases. Learn to use Claude?)
However, without fat profit margins, who is going to spend the money to maintain postgres? End users? I think it unlikely. So that's stage 2 and it's ugly.
The real doomsday planning in software right now is to prep for the need to fall back to just a compiler and still get what you need. This is why I'm more and more often writing c++, std only, for my own automation projects. I'm betting that hardware-first companies like Apple will need clang++ to survive.