Snowflake pricing consists primarily of compute, storage, and data transfer charges.
Compute accounts for 80 to 90 percent of a typical Snowflake bill. These rates are representative US on-demand prices and vary by cloud, region, and contract. The proportions matter most for accounts spending $200K a year or more.
Snowflake bills for compute, storage, and data transfer, with an additional Cloud Services category that usually has no charge. Compute is measured in credits, while storage is billed per terabyte of compressed data per month. Data transfer charges apply when data moves out of a region or across clouds. Compute is generally the largest component because active warehouses consume credits at rates determined by their size.
The dollar value of a credit depends on the Snowflake edition. On AWS US East, on-demand rates run about $2.00 per credit on Standard, $3.00 on Enterprise, $4.00 on Business Critical, and $6.00 on VPS. Higher editions add security and governance features, while equivalent workloads consume the same number of credits on each edition.
Virtual warehouses consume credits at rates set by warehouse size. An extra-small warehouse uses 1 credit per hour, and each size up doubles it: small is 2, medium is 4, large is 8, extra-large is 16, 2XL is 32, 3XL is 64, and 4XL is 128 credits per hour. Snowflake bills per second with a 60-second minimum on every resume, so a warehouse that runs a five-second query is billed for a full minute.
Storage runs roughly $40 per terabyte per month on-demand in AWS US regions, and a capacity contract can bring that toward $23 per terabyte per month. Billing is based on compressed volume. Snowflake compresses data during loading, and columnar micro-partitions typically reduce its size three to five times, so 1 TB of raw CSV might bill as 200 to 300 GB. Time Travel and Fail-safe retention can increase the stored volume.
Cloud Services handles query planning, metadata, and access control. Snowflake includes usage up to 10 percent of daily compute credits without an additional charge and bills the excess. Accounts that cross the threshold often run thousands of small queries or perform heavy information schema operations.
Snowflake runs serverless features on managed compute with credit meters separate from warehouse usage. These features include Snowpipe for continuous ingestion, serverless tasks, automatic clustering, materialized view maintenance, search optimization, and Cortex AI functions. Automatic clustering cost increases with table size and the frequency of data changes.
A large Enterprise warehouse at $3.00 per credit consumes 8 credits an hour. Running it for approximately five and a half hours a day across 21 working days produces near 924 credits for the month, or a little under $2,800. An additional 14 TB of compressed storage at $23 per TB costs about $322. In this example, the warehouse accounts for roughly 90 percent of the combined total.
This query groups metered credits by warehouse:
-- credits by warehouse, last 30 days
select warehouse_name, sum(credits_used) as credits
from snowflake.account_usage.warehouse_metering_history
where start_time >= dateadd('day', -30, current_timestamp())
group by 1
order by credits desc;
Edition and contract terms determine the credit rate. A committed-spend capacity contract lowers the per-credit price when usage is predictable. Warehouse controls include right-sizing so a 2XL does not run work suited to a medium. Consolidating bursty small queries accounts for the 60-second minimum. Serverless and clustering usage require separate cost comparisons.
They cost more only through the credit price. A query consumes the same number of credits on Standard as on Business Critical, and what changes is the dollar value of each credit, from about $2.00 on Standard to $4.00 on Business Critical on AWS US East. The premium buys security and compliance features, not faster compute.
Cloud Services is free up to 10 percent of your daily compute credits, so a charge means you crossed that threshold. The trigger is usually a flood of small, metadata-heavy queries rather than a pricing change. Look for jobs firing thousands of tiny statements or repeatedly hitting the information schema.
Cut compute, in almost every case. Warehouse credits are 80 to 90 percent of a typical bill and storage a few percent, so tuning warehouse sizing and idle time returns far more than anything you do with storage retention.
It turns on your committed volume and your negotiation, so no fixed public discount exists. On-demand is the list rate, and a capacity commitment trades an upfront spend promise for a lower price per credit. It pays off once usage is steady enough to forecast.