Snowflake and Databricks use consumption-based pricing for compute. Snowflake measures consumption in credits, while Databricks uses Databricks Units (DBUs). The units include different costs, so their listed rates are not directly comparable.
Neither a credit nor a DBU represents a fixed dollar amount. A Snowflake credit costs about $2 on Standard, $3 on Enterprise, and $4 on Business Critical, before regional differences. A Databricks DBU ranges from about $0.15 for Jobs compute to about $0.70 for SQL Serverless on AWS. In both systems, compute cost equals unit consumption multiplied by the rate for the applicable tier and workload.
A Snowflake credit includes the infrastructure that runs the workload. Snowflake operates the compute in its cloud account and bills the customer for credits and storage.
Databricks classic compute separates these costs. Databricks bills for DBUs, while the cloud provider bills for the virtual machines and storage used by the workload. Databricks serverless includes the cloud compute in a higher DBU rate. A comparison that excludes the cloud infrastructure charge therefore understates the total cost of Databricks classic compute.
Snowflake scales compute by warehouse size, and each size increase doubles the credit rate. Credit consumption is 1 per hour for an XS, 4 for a Medium, and 32 for a 2XL. A warehouse larger than its workload requires consumes additional credits without necessarily reducing runtime.
Databricks costs vary by cluster configuration and compute type. Running a scheduled pipeline on All-Purpose compute rather than Jobs compute costs three to four times more per DBU on AWS for identical work. Databricks configurations therefore require decisions about both cluster size and compute type.
On Snowflake, excess cost commonly results from warehouses that remain active after work completes and the per-second minimum applied to small queries. On Databricks, common causes include clusters without auto-termination, oversized clusters, and scheduled jobs running on All-Purpose compute. Snowflake Cortex and Databricks Genie also have separate usage meters, and automated use does not receive a free tier by default.
Neither platform is consistently less expensive for every workload. Comparative cost depends on workload characteristics, configuration, and idle time. A complete comparison includes the cloud infrastructure invoice for Databricks classic compute and measures both platforms with representative workloads. On either platform, reducing idle time and matching compute capacity to the workload lowers consumption.