ARR had a good run. But it’s over.
Annual Recurring Revenue became the dominant metric in SaaS for a simple reason. It worked. When marginal costs were near zero, revenue was a reasonable proxy for value creation. If subscriptions grew and churn stayed low, profits would eventually follow.
Not anymore.
AI did not just add features to SaaS. It reintroduced real marginal costs. Tokens cost money. Inference costs money. Fine-tuning costs money. Agents that actually do work consume compute continuously, not occasionally.
ARR ignores all of that.
Two customers can generate the same ARR and have radically different cost profiles. One uses light prompts occasionally. Another runs agents all day. One benefits from reuse and caching. Another requires fresh inference every time.
ARR treats them as identical.Your contribution margin does not.
The issue is not that ARR ignores fixed costs. The issue is that ARR assumes marginal costs are negligible. In an AI-driven product, that assumption is false.
When marginal costs return, revenue stops being the right headline number.
Enter ARM: Annual Recurring Margin
ARM does not fix this by suddenly caring about fixed costs. Like ARR, it largely ignores them.
The difference is what ARM does care about.
Annual Recurring Margin is about contribution margin. It asks a harder and more honest question: After paying the variable costs required to serve my recurring customers, how much money do I actually keep?
This includes model usage, inference, compute tied directly to customer behavior, and any other costs that scale with use. Fixed costs still matter, but ARM forces clarity on whether growth actually contributes to covering them.
That distinction matters now in a way it never did before.
ARM Changes How You Think
Once you care about ARM, several uncomfortable truths appear quickly.
Not all ARR is good ARR. Some customers are quietly unprofitable. Some features attract usage you cannot afford. Some pricing metrics reward behavior that destroys contribution margin.
ARR celebrates growth.ARM demands selectivity.
It pushes better packaging, better pricing metrics, and clearer decisions about which customers and use cases you actually want to scale.
The Quiet Casualty: LTV
Here’s what most companies miss: Lifetime Value calculations that ignore variable costs are broken.
ARR-based LTV models assumed contribution would eventually emerge because marginal costs were near zero. In an AI world, usage behavior determines contribution. That means LTV must be rebuilt on contribution margin, not revenue.
ARM is the recurring expression of that reality.
This Is Not Anti-Growth
This is not an argument against ARR. Revenue still matters. Growth still matters.
But ARR belongs to a world where ignoring marginal costs was mostly safe.
AI companies are software plus compute businesses. In that world, contribution margin is not a finance detail. It is part of the product.
The companies that win will not be the ones with the biggest ARR charts. They will be the ones who understand, design for, and defend their ARM.
Share your comments on the LinkedIn post.
Now, go make an impact!
Tags: AI, AI Companies, AI SaaS, Growth Decisions, Marginal costs, profitability, SaaS, SaaS Growth
