Calculate your AI and third-party API costs per user and at scale. Model usage per user, cost per call and monthly volumes to see total API spend and gross margin impact at 100, 1,000 and 10,000 users.
API calls per user per month:
Third-party API costs — particularly AI/LLM APIs — have become one of the most significant and fastest-growing cost lines for modern SaaS companies. Unlike infrastructure which has significant fixed components, API costs often scale almost linearly with users or usage. Modelling per-user API cost and its impact on gross margin is essential before pricing your product.
AI API costs can be reduced significantly through: (1) Model selection — using smaller, faster models (GPT-4o Mini, Claude Haiku) where full capability is not needed can reduce cost by 10–50×; (2) Caching — caching identical or similar prompts to avoid redundant API calls; (3) Prompt optimisation — shorter, more focused prompts cost less; (4) Fine-tuning on smaller models to replace expensive large model calls.
If your product's costs scale with usage, your pricing should too. Usage-based pricing (charging per API call, per document processed, per user action) shifts cost scaling risk to customers and makes unit economics more predictable. Pure seat-based pricing with high per-user API costs creates a situation where your most active (and valuable) customers are also your most expensive to serve.
LLM API costs vary by model and usage intensity. GPT-4o: approximately $0.005 per 1K tokens input; Claude 3.5 Sonnet: $0.003 per 1K input tokens. For a typical AI feature with 50 LLM calls per user per month at 500 tokens each: ~25,000 tokens = approximately $0.12-0.25 per user per month. At 1,000 users: $120-250/month. Heavier usage (document analysis, long context) can cost $2-10+ per user per month — which must be factored into pricing above the standard £10-50/month SaaS price point.