How Pricing Works on Databricks
I highly recommend checking out this blog post on how databricks pricing works from my colleague @MENDELSOHN CHAN
Databricks has a consumption based pricing model, so you pay only for the compute you use.
For interactive development, scheduling ETL and ML jobs, and our Databricks SQL classic compute, there is a DBU (Databricks Unit) cost charged by Databricks as well as associated AWS EC2 / Azure VM / GCP GKE, storage, and egress costs that are charged by your cloud provider. For Databricks SQL Serverless (AWS | Azure), you are charged one low rate that includes both the Databricks and cloud provider costs. A DBU is a unit of processing power per hour, billed on a per second basis.
Unlike some other cloud data platforms, Databricks separates out Databricks and cloud provider costs to Databricks to pass on savings from your existing negotiated cloud provider discounts, reserved instance pricing (AWS | Azure | GCP), as well as savings from leveraging EC2 Spot Instances, Azure Spot VMs, or GCP Preemptible VM Instances. Spot alone can save you up to 80% on cloud provider compute costs.
Databricks tags automatically propagate to the cloud compute instances to easily associate and monitor costs for both Databricks and your cloud provider (AWS | Azure | GCP).
Refer to our pricing page for additional pricing details. For further information and a deep dive on this topic, we recommend getting in touch with your Databricks account team.
Let us know if you have any pricing questions in the comments!
