Snowflake automatically collects and stores metadata for every column in each micro-partition, including MIN/MAX values, distinct counts, NULL counts, and other properties. This metadata is maintained in the Cloud Services Layer and is used to optimize query performance and enable efficient pruning.