Stories from the SaaS Wars: 2026-05-23
Some stories I found interesting this week, mostly focused on the SaaS Wars.
We’ve got a few stories from the SaaS wars—and takes, of course. We have an infinite supply of takes, far exceeding demand.
Dispatch from the SaaS data access wars:
Every week, there’s a new story about so-and-so blocking data access from such-and-such. This week, Microsoft is supposedly blocking a certain kind of access from external tools to Power BI. The article contains this haunting quote: “The friction centers on Power BI, a Microsoft product nearly all Fortune 500 firms use to analyze data about their operations in charts and other visual formats.” Really? Nearly all?
I’d like to take this chance to share a conspiracy theory. Power BI isn’t real. I hear about it all the time. It’s allegedly everywhere. And yet… I’ve never used it or seen it be used. I’m tech-old. I’ve been around lots of data environments. Sounds fake to me.Did you hear SpaceX is going to be IPO?
No? I didn’t either. I couldn’t find a single article about it to share. See if you can find one somewhere on the internet.Customers demand shorter contracts over uncertainty of “what comes next”
There’s a story on AI dollars crowding out traditional business software purchases. The pool of dollars isn’t fixed, and if there is a large return to these AI products, why couldn’t these firms borrow to cover token purchases and repay the loans with excess profits? So, something is wrong here. Either credit markets (unlikely), the returns from AI spend are currently negative (plausible until companies learn the right way to deploy it at their firm), or there’s a principal-agent problem (highly likely). The person doing the buying has a fixed budget, so even though the company could just expand borrowing, there’s no cheap way for the agent to pass the information up the chain.
The second half of the story is more plausible, but speculative. Companies are worried that a wave of new, better products will emerge over the next few years and are reluctant to be locked into a long-term contract. That makes sense, but I really haven’t seen any evidence of useful AI-native business software, aside from the models and coding stuff coming from the primary labs. The AI components added to existing SaaS have been useful, but AI features are not really useful on their own. You need something that works well for all the more mundane features of the app first, and that turns out to be a hard problem.
It reminds me of every time I’ve seen what I think is an opportunity to exploit in an industry. I realize that, while I might be right about the opportunity, to actually profit from it, you’d have to launch a competitor that does all the other aspects of the industry right and then just adds the new insight. That’s very hard to do.Intuit moves to consumption-pricing.
Everyone is doing it because it’s the only thing you can do, now that the Variable Cost Revolution is well underway. What is the Variable Cost Revolution, you ask? For a long time, one reason software companies were so profitable was that their costs were almost entirely fixed. Costs did not increase very much as they added more customers. So, with scale, they made oodles of money. But AI has changed that. If you run an AI feature in your software, you have a variable cost product. As demand for the product increases, your costs increase meaningfully. LLM-based features have changed the cost structure of the entire industry, and the pricing structure follows the cost structure.
Thanks for reading!
Zach
Connect at: https://linkedin.com/in/zlflynn