Databricks Cluster Cost Calculator
What will this Databricks cluster actually cost per month? Enter DBUs/hour, your price per DBU and node count to get the DBU cost, the cloud VM cost kept separate, and the combined total, with every step of the arithmetic shown.
- Free with a free account
- Runs on our server in milliseconds
- Your input is not stored or logged
Create a free account to use the calculator
Members use every tool on AIDataLegends for free. It takes a name, an email and a password, and you come straight back here.
Not ready yet? The worked example below shows exactly what the tool produces.
Worked example
An 8-worker cluster where each worker consumes 2 DBUs/hour and the driver 1 DBU/hour, at $0.55 per DBU and $0.68/hour per instance for the underlying cloud VMs, running 10 hours a day on business days, with autoscaling between 4 and 12 workers. The calculator's answer:
What this means
- The range assumes the cluster sits at one end the whole time. Real autoscaling workloads usually land somewhere in between — use the range as bounds, not a midpoint estimate.
- Serverless SQL warehouses and serverless jobs bill DBUs per query/run, not per cluster-hour, and often include compute in the DBU price. This calculator models a classic cluster that runs for a block of time; it won't match a serverless bill.
The breakdown
DBUs/hour = (worker nodes × DBUs/hour/node) + driver DBUs/hour
= 17
DBU cost/hour = DBUs/hour × price per DBU = 9.35
VM cost/hour = (nodes + 1 driver) × instance price/hour = 6.12
Total/hour = 15.47
Total/month = Total/hour × hours/day × days/month = 3,352.35
Starting points, not an invoice: list prices vary by cloud, region and contract. Use your own negotiated or published rates for an exact number.
Why DBU cost and cloud VM cost are separate
A Databricks invoice has two distinct line items, and people new to the platform often see one total and assume it is the whole bill:
- DBUs (Databricks Units) are Databricks' own usage metric — a measure of compute capacity used per hour, billed at a price per DBU that depends on your cloud, your tier (Standard, Premium or Enterprise) and the workload type (All-Purpose Compute, Jobs Compute, SQL Compute). This is what shows up on the Databricks invoice itself.
- The underlying cloud VM (an EC2 instance, an Azure VM, a GCE instance) is billed separately, by your cloud provider, at its normal on-demand or committed-use rate. Databricks doesn't mark this up or include it in the DBU price — it's a completely different bill, sometimes on a completely different invoice.
The calculator keeps the two apart for exactly that reason: the cloud VM price per hour is optional, and leaving it blank gives you the DBU cost alone (the number Databricks itself bills you for). Fill it in for the all-in total you actually pay across both bills.
Where to find your numbers
- DBUs/hour per node comes from the instance type's listed DBU consumption rate on the Databricks pricing page for your cloud, or from the "DBU" column in your workspace's usage/billing dashboard for an instance type you already run.
- Price per DBU depends on cloud, tier and workload type, and on any committed-use discount your org has negotiated — use your contract's rate when you have one, the public list price otherwise.
- Cloud VM price per hour is your cloud's on-demand, reserved or committed-use rate for that instance type and region — from AWS/Azure/GCP pricing pages, not from Databricks.
Nothing here is looked up automatically and nothing is cached: every number comes from what you type in, so the answer stays correct even after prices change.
The levers that actually move this number
- Workload type. Jobs Compute and SQL Compute are typically priced lower per DBU than All-Purpose Compute for the same instance — running steady batch work on interactive (All-Purpose) clusters is one of the most common ways teams overspend.
- Auto-termination. An interactive cluster left running idle bills the same as a busy one; set a short auto-termination window on clusters people forget to shut down.
- Autoscaling bounds. A wide min–max range covers bursty load but can also mask a floor that's higher than the workload needs on quiet days — the range in this calculator shows both ends so you can see the spread.
- Photon. Photon-enabled compute costs more per DBU but can finish jobs faster; it's worth it when the speedup outweighs the higher rate, and a cost regression when it doesn't.
- Committed-use discounts on the cloud VM side (reserved instances, savings plans) reduce the infrastructure half of the bill without touching the DBU price at all.
Limitations and assumptions
- This models a classic (All-Purpose or Jobs) cluster running for a block of hours. Serverless SQL warehouses and serverless jobs bill DBUs per query or per run, often bundling compute into the DBU price — a different model this calculator doesn't represent.
- The autoscaling range assumes the cluster sits at one bound the whole time; real workloads usually land somewhere between the two, so treat the range as bounds, not an estimate.
- Storage, networking egress, and any other cloud services the workload touches (a warehouse, a catalog, a message queue) are not included — this is compute cost only.
- Prices change by cloud, region, tier and contract, and this tool has no pricing data of its own — it is pure arithmetic on the numbers you provide.
Frequently asked questions
What is a DBU?
Is the DBU cost the whole Databricks bill?
Why is my driver node a separate input?
Does this work for SQL Serverless or serverless jobs?
Why do I need an account?
Keep going
- ArticleDatabricks SQL Serverless Cost Governance: Budgets, Alerts, and Query-Level Attribution
- ArticleDatabricks Lakehouse Platform for Real-Time, Agentic Data and AI Workloads
- ToolSpark File Size & Partition Planner
- ToolJSON to PostgreSQL DDL Generator
- ToolSQL Data Quality Check Generator
- ToolCron & Airflow Schedule Builder
- All toolsBrowse the free tools for data engineers