The Databricks Unit (DBU) is a measure of compute consumption used for Databricks billing. DBUs are expressed as hourly rates and metered by the second. As of 2026, a single Databricks Unit on AWS lists near $0.15 for Jobs compute and climbs past $0.70 for SQL Serverless. The DBU charge equals consumption multiplied by the rate for the workload, while classic compute also produces a separate cloud infrastructure charge.
Classic compute generates two charges. Databricks charges for DBUs, while the cloud provider charges for the virtual machines and storage supporting the workload. A DBU-only estimate excludes the AWS or Azure infrastructure charge. Serverless compute includes cloud infrastructure in a higher DBU rate.
DBU rates vary primarily by workload type. Jobs compute has the lowest rate. All-Purpose compute supports interactive notebooks and runs three to four times higher per DBU on AWS, with a smaller multiple on Azure. SQL Warehouses are offered as Classic, Pro, and Serverless products with different rates. The Standard, Premium, and Enterprise plan tiers also affect pricing, as do cloud and region.
Representative US on-demand list rates on AWS in 2026 span from roughly $0.15 per DBU for Jobs compute to roughly $0.70 for SQL Serverless, with All-Purpose sitting near $0.55. These figures are approximate because Databricks revises its rates and charges different prices across clouds. The current pricing page provides the applicable rates.
An All-Purpose cluster that burns 11 DBUs an hour across a 7-hour daily run over 23 working days consumes 1,771 DBUs, near $974 at $0.55 before cloud infrastructure costs. The same work on Jobs compute at $0.15 reduces the DBU portion to about $266. In this example, changing the compute type reduces the DBU charge by more than two thirds without changing the job.
Batch jobs running on All-Purpose compute can cost three times the Jobs compute rate. Clusters without auto-termination continue billing while idle, and clusters with more workers than required consume excess DBUs during every run. AI features such as Genie have separate DBU meters, and automated requests sometimes receive no free allowance.
Databricks system tables report DBU consumption and dollar cost by workload.
-- databricks dbu cost by workload type, current list prices
select u.sku_name,
sum(u.usage_quantity * p.pricing.default) as list_cost
from system.billing.usage u
join system.billing.list_prices p
on u.cloud = p.cloud and u.sku_name = p.sku_name
where p.price_end_time is null
group by u.sku_name
order by list_cost desc
The resulting data supports cluster right-sizing, auto-termination settings, and migration of scheduled work from All-Purpose to Jobs compute.
What does a DBU represent?A DBU, or Databricks Unit, is the measure Databricks uses to bill compute consumption by the hour, charged per second. Your cost equals the DBUs you consume times the workload's dollar rate, and on classic compute the cloud infrastructure bill sits on top.
What sets the price of a single DBU?The workload type, plan tier, cloud, and region all set it. In 2026, representative US on-demand rates run from roughly $0.15 per DBU for Jobs compute to roughly $0.70 for SQL Serverless. The Databricks pricing page lists the figures in force.
Is the cloud VM cost baked into the DBU rate?On classic compute it is not, so Databricks bills the DBUs while your cloud provider bills the VMs and storage on a separate invoice. Serverless merges the two into one higher DBU rate.
My Databricks bill beat my DBU estimate, so what happened?The estimate probably left out the cloud infrastructure, assumed a cheaper workload rate than the one that ran, or overlooked idle time and All-Purpose usage.