August 5, 2026
Margins? More like drama-gins
Something is changing in the unit economics of software
AI software’s easy-money era may be ending, and the comments are already fighting about it
TLDR: AI features are making software companies pay more every time users do something, threatening the fat profits that made the industry famous. In the comments, people are split between “this is a real business crisis” and “relax, costs will fall like they always do.”
For years, software companies had what looked like a magic trick: build an app once, sell it to tons of people, and keep most of the money. That was the dream behind the modern subscription software boom. But now that users expect chatty, smart, AI-powered features, every click can come with a real bill attached. Suddenly, software starts looking less like a money-printing machine and more like a business where serving each customer actually costs something.
And the community? Oh, they are not quietly nodding along. One camp basically said, “Calm down, software always cost money to run,” arguing this is just the next version of server bills and that costs will probably shrink over time anyway. Another camp is already gaming out survival tactics, like pushing more of the work onto your phone or laptop so companies don’t bleed cash every time you ask a question. Then came the spicy existential take: if AI makes it easier for everyone to build custom tools, why pay for software subscriptions at all? That’s the kind of comment that makes founders spill their coffee.
The drama boils down to this: is AI ruining the old software business model, or is this just a temporary panic before costs drop again? The funniest part is that beneath all the money talk, the comments read like a reality show reunion: optimists yelling “prices always fall,” skeptics muttering “margins were never magic,” and tinkerers whispering, “fine, we’ll just run it ourselves.”
Key Points
- •The article says traditional software achieved gross margins of about 75% to 85% because additional users were inexpensive to serve.
- •AI products often require LLM inference for user interactions, introducing direct per-use compute costs that scale with usage.
- •The article argues that AI creates a tradeoff between margin and product quality because model choice affects both cost and user experience.
- •Many AI companies manage costs by using multiple models, routing simple tasks to smaller or fine-tuned models and harder tasks to frontier models.
- •The article cites ICONIQ survey data stating that average gross margins for AI products are around 52% in 2026, below historical SaaS levels.