Snowflake Query Acceleration Service uses serverless compute to process eligible portions of queries in parallel with warehouse compute. Its credits are billed separately from the warehouse.
The service is designed for outlier queries with large scans and selective filters, and for some large data modification operations. Eligibility depends on the query plan, warehouse size, and available service capacity. Not every query benefits from it.
Query Acceleration Service is available on Enterprise Edition and higher. It can be enabled on a warehouse with ENABLE_QUERY_ACCELERATION = TRUE.
The service is billed by the second while its compute resources are in use. QUERY_ACCELERATION_MAX_SCALE_FACTOR limits the amount of extra capacity a warehouse can lease. A Medium warehouse uses 4 credits per hour, so a scale factor of 5 permits up to 20 additional credits per hour when the service uses the full allowance.
The scale factor is an upper bound, not a guaranteed charge. The service uses only the resources required by eligible queries and available at execution time.
QUERY_ACCELERATION_HISTORY reports credits used by the service. Compare those credits with warehouse credits, execution time, and the number of eligible queries before and after enabling it. A faster query is not a cost reduction unless the additional service credits are offset by lower warehouse usage or a meaningful latency requirement.
Snowflake documents Query Acceleration Service and its billing history view.