The headline number is always revenue. The number that decides whether it's a good business is what it keeps after the cost of serving customers. For AI companies that gap can be huge, because every answer costs money to generate.
What you'll have by the end: a clear sense of gross margin, why it's different for AI, and a simple model you can fill in for any company.
Two words, in plain English
- Cost to serve: what it costs to deliver the product to one customer. For an AI app, that includes the "inference" cost, the bill it pays the model provider every time you hit send.
- Gross margin: revenue minus cost to serve, as a percentage. It's what's left before salaries, marketing, and everything else. High is good; negative means it loses money on every sale.
Why AI is different
Classic software costs almost nothing to serve one more user, so margins are high. AI is different: every response burns compute the company pays for. If a company charges $20 a month but heavy users cost $30 in compute, it's losing money as it grows, unless prices rise or costs fall. That's the question to always ask.
The simple model
Fill this in for any company you're curious about. Estimates are fine, just label them:
Company: [name]
Revenue (per month or year): $[amount] [source + date]
Cost to serve that revenue: $[amount] [estimate?]
- biggest piece for AI: model / compute cost
Gross margin = (Revenue - Cost to serve) / Revenue
= $[..] / $[..] = [..]%
Read it:
- 70%+ : healthy software-like margins
- 30-60%: real costs to watch, common for AI apps
- under 0%: losing money per customer, needs a fixWhat it doesn't tell you
Gross margin is a starting point, not the bottom line. Salaries, marketing, and research still come out of what's left. But if the gross margin is thin or negative, nothing downstream fixes it, so it's the first thing worth understanding.
Want to put this to work on a real company? The research a business guide walks the whole thing through on Suno.