Usage-Based Infrastructure Pricing
Infrastructure SaaS (Snowflake, MongoDB, Confluent, Elastic, Databricks) prices based on resource consumption rather than user count. A Snowflake customer pays for compute credits used during query execution and storage consumed. A MongoDB Atlas customer pays for compute instances and data transfer. As customer workloads grow, spend grows automatically.
This model has a fundamental implication: revenue forecasting is harder. A seat-based SaaS company knows next quarter's revenue within ±5% (seats × price). An infrastructure SaaS company's revenue depends on how much its customers' workloads grow — correlated with their own business growth but not perfectly predictable. Snowflake's "product revenue" forecasting challenge was a major public market concern post-IPO.
Read how these companies disclose revenue visibility through RPO and Remaining Performance Obligations.
Net Expansion and Cohort Economics
Infrastructure SaaS NRR is structurally higher than most SaaS sectors because revenue compounds with customer data gravity. Once a company's critical data lives in Snowflake or Databricks, migrating it is costly and risky — so retention is high (GRR 90–95%+), and expansion happens naturally as data volumes grow.
Cohort analysis is the right lens: a 2019 Snowflake cohort that was spending $100K/year might be spending $2M+/year by 2024, with zero upsell required. This compounding cohort expansion is the core value proposition of infrastructure SaaS — and why investors will pay 15–20x EV/Revenue for the best operators in the category.
Compare this to EV/Revenue benchmarks across other sectors to see the premium infrastructure commands.
Live Infrastructure SaaS Benchmarks
Gross Margin Compression from Cloud COGS
Pure software SaaS runs 75–85% gross margins because COGS is primarily hosting costs plus support. Infrastructure SaaS has additional COGS: the compute, storage, and networking costs that back the customer's actual workloads. Snowflake purchases cloud capacity from AWS, Azure, and GCP and re-sells it with a margin. This intermediary margin compresses gross margins to 60–70%.
The path to gross margin expansion is: (1) more efficient data processing software (same query on less compute), (2) reserved capacity purchases at lower cost, and (3) proprietary hardware (Snowpark containers, Databricks Photon) that reduces dependence on third-party cloud. Track this trajectory in the income statement COGS breakdown.