Something Is Changing in the Unit Economics of Software
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
AI is fundamentally eroding the high-margin economics that defined SaaS for 15 years. Traditional software had near-zero marginal cost per user, enabling 75-85% gross margins. AI-native products require LLM inference calls per user interaction, introducing real per-unit costs that scale with usage. Survey data from ICONIQ shows average gross margins on AI products at ~52% in 2026, well below SaaS norms. This forces a rethink of the classic SaaS playbook: aggressive customer acquisition justified by improving unit economics no longer holds when each customer carries ongoing compute costs. Usage-based pricing is emerging as the rational response to variable cost-to-serve. While inference costs are falling, Jevons paradox suggests cheaper calls lead to more calls per user rather than margin improvement. The new playbook resembles capital-efficient business building: unit economics from day one, pricing that reflects real costs, and growth funded by the business itself.
