Snowpipe and Snowpipe Streaming are serverless ingestion services with separate Snowflake credit meters. Their costs depend on the ingestion method and the work performed.
Snowpipe loads files from cloud storage as they arrive. Snowflake charges credits for the serverless compute used to load those files. File count, file size, notification frequency, and transformation work affect the amount of work required.
Many small files can create more ingestion overhead than fewer larger files with the same total volume. Batching files where latency permits can reduce the number of load operations.
Snowpipe Streaming writes rows directly through a streaming client instead of waiting for files in cloud storage. Snowflake charges based on the service consumption defined in the current Service Consumption Table. The service has different operational characteristics from file-based Snowpipe, so ingestion latency and data volume should be measured together.
The lower-cost option depends on the workload. Compare credits, data volume, file or row rate, latency, and operational overhead over a representative period. A small test can understate production cost when row volume or event frequency grows.
METERING_DAILY_HISTORY reports daily usage by service type. Snowflake's Snowpipe and Snowpipe Streaming documentation describe the services and their billing references.