Snowflake serverless features use managed compute and separate credit meters rather than a customer-managed virtual warehouse.
Serverless usage can include Snowpipe, serverless tasks, automatic clustering, search optimization, materialized view maintenance, and Cortex AI functions. Each feature has its own meter and rate. A warehouse-only report therefore does not show the full account cost.
The service runs the underlying compute for the feature. The account is billed for the credits recorded by that service, while SQL that surrounds the feature can also consume warehouse credits.
METERING_DAILY_HISTORY provides daily totals by service type. Feature-specific usage views provide more detail for services that expose them. Compare those records with WAREHOUSE_METERING_HISTORY when reconciling account usage with an invoice.
An AI function called inside a query can create two charges: a serverless AI charge and a warehouse charge for the query that invokes it. Similar separation applies to maintenance services that run alongside ordinary warehouse work.
Measure serverless usage by feature, estimate the volume generated by each workload, and include refresh or maintenance work in the cost comparison. Capacity commitments should be evaluated after both warehouse and serverless consumption are stable.
Snowflake documents service meters in the overall cost and account usage references.