AI Threatens Traditional SaaS Gross Margins
Industry analysts note a fundamental shift where AI is undermining the high gross margins and low incremental costs that have defined traditional SaaS businesses. The next generation of software is expected to be larger and more complex. This could lead to lower-margin business models, but with greater leverage for platforms that own the core workflow and data.
- While traditional SaaS companies aim for 80-90% gross margins, AI-native companies often operate with structurally lower gross margins of 50-65% due to the variable costs of AI model inference. Early-stage AI startups can see margins as low as 25% before optimizing infrastructure and pricing. - The core economic shift is from near-zero marginal costs for new users to a significant "cost-per-action" for AI-powered features. Every time a user interacts with an AI model, it incurs a direct, variable cost related to compute power, which can surpass cloud infrastructure expenses. - In response to high inference costs, 92% of AI software companies now use mixed pricing models that combine subscriptions with consumption fees. Purely seat-based pricing is in sharp decline as it fails to align with the value AI delivers, which is often independent of the number of human users. - Companies are moving towards usage-based pricing tied to metrics like tokens processed, API calls, or workflows executed. Some are even experimenting with outcome-based pricing, such as Intercom's fixed price per resolved support ticket. - Venture capitalists are adapting their evaluation metrics for AI companies, focusing more on data moats and model performance rather than traditional SaaS metrics like short-term ARR growth. The concern for investors is that AI agents could relegate traditional software to being passive data stores, eroding their pricing power. - To improve margins from the initial 25% range to a more mature 60%+, founders focus on shifting from third-party APIs to custom models, implementing intelligent routing to manage compute costs, and refining hybrid pricing models. - The rise of "agentic AI" that handles complex workflows independently threatens to erode the value of seat-based licenses. If AI makes individual employees more productive, companies may require fewer licenses, pressuring a foundational revenue model for established SaaS vendors. - AI enables software to be sold against a company's labor budget, a much larger total addressable market than software budgets. This is because AI can replace or augment human tasks, justifying transactional or outcome-based pricing that scales with the value delivered.
