Snowflake bills virtual warehouse compute by the second after a warehouse starts, with a 60-second minimum for each start.
A warehouse that runs a five-second query is billed for 60 seconds. A warehouse that runs for 90 seconds is billed for 90 seconds. Repeated short runs can therefore produce more billed time than query execution time suggests.
The minimum applies each time a warehouse resumes. It does not apply once per day or once per query. Warehouse size determines the credit rate during the billed interval, so a larger warehouse increases the cost of every minimum charge.
WAREHOUSE_METERING_HISTORY reports credits consumed by warehouse and hour. QUERY_HISTORY reports query start and end times. Comparing the two shows whether a workload spends most of its credits processing queries or starting and waiting between them.
For example, a warehouse that resumes 120 times for five-second queries receives at least 120 minutes of billed time, even though the queries execute for 10 minutes in total. The exact credit cost depends on the warehouse size and Snowflake edition.
Per-second billing makes short workloads less expensive than an hourly model, but the 60-second minimum still matters for intermittent jobs. Grouping work into fewer runs can reduce minimum charges. A longer idle-time limit can preserve a warm warehouse, while a shorter limit can reduce time spent running without queries. The right choice depends on query frequency and latency requirements.
Snowflake describes the minimum billing period in its warehouse documentation.