Databricks jobs compute pricing: $/DBU on AWS, Azure and GCP, plus the VM bill

Jobs Compute is the rate Databricks charges for scheduled, non-interactive work on classic clusters. The $/DBU number is easy to find. What most estimates miss is how many DBUs a cluster burns per hour, and the separate cloud VM bill that usually comes out bigger than the Databricks line.

The short answer

• AWS and Google Cloud: $0.15 per DBU on Premium and $0.20 on Enterprise, the same in every region.
• Azure: $0.30 per DBU on Premium in 43 commercial regions. Databricks notes that Azure Premium corresponds to Enterprise on AWS and GCP, so the like-for-like comparison is $0.30 vs $0.20, not $0.30 vs $0.15.
• Photon doesn't change the rate. Jobs Compute Photon lists at the same $/DBU as Jobs Compute on all three clouds. It changes how many DBUs an instance uses per hour (see Is Photon worth it?).
• The VM bill is extra and often larger. In the 5-node example below, the cloud VMs were 57% to 65% of the hourly total on AWS and GCP, and 39% on Azure.

Jobs Compute $/DBU by cloud and tier

Jobs Compute with Photon is billed at the same $/DBU as Jobs Compute on all three clouds.

AWS and Google Cloud list one classic jobs rate for all regions. On Azure, the $0.30 Premium rate applies in 43 commercial regions; the exceptions in the data are the US Gov regions at $0.375 and the two Jio India regions (Central and West) at $0.975. Azure's Standard tier ($0.15) still appears in the retail price list, but Microsoft retired it on October 1, 2026; see Azure Databricks Standard to Premium for what that move does to a bill.

Price history for each SKU: AWS Jobs Compute, Azure Jobs Compute, GCP Jobs Compute.

How many DBUs a jobs cluster uses

Every classic instance type has a DBU rating per hour. A few common 4- and 8-vCPU types on Jobs Compute:

On Azure, the same 8-vCPU, 32 GB shape is rated 1.50 DBUs/hour on a D8s v3 and 2.00 on a D8s v5, 33% more. That's the same jump as going from Premium ($0.15) to Enterprise ($0.20) on AWS or GCP, so the instance pick moves the Databricks line as much as the tier does.

What a 5-node jobs cluster costs per hour

Example, not a quote. Assumptions: 1 driver + 4 workers of the same 8-vCPU, 32 GB instance, Jobs Compute, no Photon, on-demand Linux VM list prices in AWS US East (N. Virginia), GCP Iowa (us-central1) and Azure East US. A month here means 8 hours a day for 22 days (176 hours).

Two things stand out. On AWS and GCP, a DBU-only estimate misses more than half the bill. And on Azure, with the same VM cost as AWS, the cluster costs about 50% more per hour than AWS Enterprise (and 67% more than AWS Premium), because the D8s v5 is rated at more DBUs and Azure Premium's $/DBU is higher. Run your own numbers in the AWS, Azure and GCP total cost calculators.

Classic jobs vs serverless jobs

Serverless jobs compute has a much higher $/DBU, but it includes the VMs, so there's no separate cloud bill:

Serverless rates vary by region more than classic ones do: AWS Premium runs from $0.35 in the US to $0.46 in Brazil, and Azure from $0.322 in Chile Central to $1.081 in the two Jio India regions ($0.45 in East US, $0.65 in Switzerland West). Whether serverless comes out cheaper depends mostly on how much of a classic cluster's billed time is startup and idle; the serverless vs classic guide works through it, with a calculator.

Has Jobs Compute gotten more expensive?

Not since we started tracking. DBU Watch has checked Jobs Compute list prices daily since October 15, 2026, and none has changed since. On Azure, the current Premium Jobs Compute price in East US and West Europe carries an effective date of March 1, 2019. We don't have verified AWS or GCP price history from before October 15. Any future change shows up on the price history page, or download the full history.

Keeping jobs compute costs down

• Run scheduled work on Jobs Compute, not All-Purpose. Databricks' own guidance is that non-interactive workloads cost significantly less on job compute.
• Compare tiers like for like. Azure Premium lines up with AWS and GCP Enterprise, not Premium.
• Price the VMs, not just the DBUs. Spot instances, reservations and savings plans cut the larger half of a classic bill; they don't apply to serverless.
• Check DBU ratings before picking an instance. Two VMs with the same vCPU and RAM can differ by almost 50% in DBUs per hour.
• New to DBUs? Start with Databricks pricing explained.

Prices are public list prices (AWS and Google Cloud from Databricks' pricing data; Azure from the Azure Retail Prices API; VM prices from AWS and Google Cloud public on-demand price lists and the Azure Databricks pricing calculator data) as of October 27, 2026. Discounts and committed-use contracts aren't included.