Estimate your total monthly Databricks spend across DBU (Databricks Units) and underlying cloud VM cost. Choose SKU, cluster size, runtime hours, and cloud - get an instant monthly total with a side-by-side Snowflake comparison.
How is Databricks billed?
Databricks charges you in two parts: DBUs (Databricks Unit consumption based on compute type and tier) and the underlying cloud VM cost (EC2 / Azure VM / GCE). Serverless SKUs bundle both into a single DBU rate so the VM line disappears, but the effective per-hour cost is higher because Databricks is now managing the VMs for you.
SQL Serverless: $0.70/DBU - instant-start, auto-scaling, no VM management
Model Serving Serverless: $0.07-0.30/DBU - ML inference endpoints
Databricks vs Snowflake: how do the costs compare?
Direct comparison is tricky because the units differ. Snowflake charges per-credit (a Medium warehouse = 4 credits/hour = ~$16/hour on Enterprise), while Databricks charges DBU + VM. For a rough equivalent: a Databricks M5.xlarge Jobs Photon cluster with 4 workers runs around $7-9/hour all-in, vs a Snowflake Medium at $16/hour - but Databricks needs you to manage cluster startup/shutdown, while Snowflake auto-suspends in seconds. Factor in dev-time savings when comparing.
Pro tips to reduce Databricks spend
Use Jobs clusters for production - never All-Purpose, which is 3x more expensive per DBU.
Enable Photon on CPU-heavy SQL and ETL - the 1.5x DBU premium typically cuts runtime by 2-5x.
Use spot instances on driver + workers for non-SLA jobs - 70-90% VM cost reduction.
Set auto-termination to 10-20 min on All-Purpose clusters - idle clusters are the #1 cost leak.
Right-size with cluster event logs - if CPU rarely exceeds 40%, downsize one tier.