Guide · Unit economics
AI startup gross margin: a practical finance guide
AI startups often look like software companies from the customer side and infrastructure businesses from the cost side. That is why gross margin deserves more attention than a generic SaaS template gives it.
By Serge Mochtchenkov, CFA · Fractional CFO for AI startups
Gross margin starts with classification
Gross margin equals revenue less cost of revenue, but the analytical challenge is deciding which costs are directly required to deliver the product. For many AI products, model/API usage and usage-driven infrastructure are obvious candidates. Product R&D for future capability is different. Apply a consistent policy and explain it.
Build the margin from operational drivers
Do not forecast inference cost as a permanent percentage of revenue if the underlying cost is driven by requests, tokens, model mix, cache rate, or another technical variable. Model the driver. That creates a bridge between the product roadmap and the financial plan.
Segment where economics differ
A large enterprise account may have better retention but far heavier model usage. A free user may have zero revenue and non-zero inference cost. Different product tiers may route traffic to different models. Segment margin where those differences are material.
Model the improvement path
Gross margin can improve through routing, smaller models, caching, context optimization, batching, provider negotiation, reserved capacity, product constraints, or pricing. Put the expected timing and effect of those initiatives into the forecast. That makes the margin story testable.
Use benchmarks carefully
External benchmarks can help frame a conversation, but AI products differ dramatically. Compare like with like and focus on the company's own trajectory, contribution economics, and ability to explain the drivers.
Investor-ready explanation
A strong answer is: this is our current margin; these are the three drivers; this is the sensitivity to usage and provider price; these technical changes are scheduled; this is the expected financial impact; and this is what happens in the downside case.
Work with me
Need the model behind this decision? I build the financial system for AI-native startups preparing to raise, improve runway, or defend their unit economics. Book a free 30-minute intro call.
Related reading