List-price math only. DBU rates exclude the underlying VM compute you separately pay to AWS / Azure / GCP. Real Databricks bills typically split 40-60 percent DBU and 40-60 percent infrastructure, depending on instance family. The deep audit models your real usage_table data and ranks reduction wins (Jobs migration, tier downgrade, autoscale ceiling, photon enablement) by dollar impact.
| Line item | Math | Monthly |
|---|---|---|
| — | ||
| Compute type | Standard | Premium | Enterprise |
|---|---|---|---|
| Jobs Compute | $0.07 | $0.10 | $0.13 |
| All-Purpose Compute | $0.40 | $0.55 | $0.65 |
| SQL Serverless | — | $0.55 | $0.70 |
| DLT Pro | $0.25 | $0.36 | $0.50 |
One-page checklist of the biggest Databricks DBU levers — Jobs-vs-All-Purpose migration audit, the forgotten-interactive-cluster pattern, when Premium tier actually pays for itself, SQL Serverless vs dedicated SQL Warehouse decision tree, Photon enable for the 2-3x speedup, instance pool warm-pool strategy, spot instance ratio for Jobs Compute, autoscale ceiling traps, and the cluster termination policy mistake that costs $500 overnight. PDF sent to your inbox.
monthly_cost = dbu_per_hour × hours_per_day × days_per_month × parallel_clusters
Example: All-Purpose Standard on AWS at $0.40/DBU × 8 hours/day × 22 work days × 1 cluster = $70.40/month. The same workload on Jobs Standard at $0.07/DBU drops to $12.32/month — an 82 percent reduction with zero change in actual work performed. Switch from Standard to Premium and the All-Purpose rate jumps to $0.55/DBU = $96.80/month (37 percent increase). The forgotten-cluster pattern: a developer leaves an All-Purpose Premium cluster running 24/7 instead of 8 hours work-day only. 24 × 30 × $0.55 = $396/month, $4,752/year — for a single forgotten cluster. Multiply that across a 15-person data team and the surprise number reaches five figures.
Two biggest reductions on most teams: migrate scheduled production workloads from All-Purpose to Jobs Compute (4-5x cheaper at the DBU layer with zero code changes), and downgrade non-regulated workspaces from Premium to Standard (saves the +$0.10-0.20/DBU governance markup teams pay without using Unity Catalog policies).
A DBU (Databricks Unit) is Databricks' unit of processing capability per hour. It's a normalized measure that lets Databricks price compute across clouds and instance families with a single number. The confusion comes from three layered multipliers: compute TYPE (Jobs $0.07-0.13 vs All-Purpose $0.40-0.65 vs SQL Serverless $0.55-0.70), TIER (Standard, Premium, Enterprise step up $0.10-0.20 each), and CLOUD provider (AWS, Azure, GCP each publish slightly different rates). On top of the DBU bill you also pay the cloud for VM compute. Most teams quote a POC on $0.07/DBU Jobs and discover production runs on $0.55/DBU All-Purpose Premium — an 8x surprise at identical workload.
Jobs Compute is for non-interactive scheduled work — ETL pipelines, batch training, Airflow DAGs, scheduled notebooks. It spins up for the run, executes, tears down. The DBU rate is 4-5x cheaper than All-Purpose because there's no attached user, no kept-warm idle, no autorestart on driver death. All-Purpose is for interactive notebook development and shared analyst sandboxes where humans attach and iterate. Killer rule: any scheduled production workload running on All-Purpose is overpaying 4-5x. The fix is one config change in the workflow JSON — switch new_cluster to job_cluster and the DBU rate drops from $0.55 to $0.10 instantly.
Premium adds roughly 40-60 percent over Standard (All-Purpose Standard $0.40 -> Premium $0.55). What it unlocks is governance and security: Unity Catalog with table-level ACLs, row filters, column masking, audit logs, IP allow-lists, table access control, SSO, credential passthrough. Multiple teams sharing a workspace, regulated data (PII, PHI, PCI), or any compliance requirement → Premium pays for itself in the first audit. Single team running internal analytics on non-regulated data → Standard is fine, Premium is dead weight. Enterprise adds another +$0.05-0.10/DBU for HIPAA, customer-managed keys, private link. Don't pay for Enterprise without a compliance gate forcing it. Common waste pattern: workspaces on Premium because that was the signup default, with zero Unity Catalog policies configured.
SQL Serverless looks more expensive per DBU ($0.55-0.70) because the rate bundles the underlying compute Databricks runs for you on their account. With Jobs or All-Purpose you separately pay AWS / Azure / GCP for the VMs plus the Databricks DBU on top. With SQL Serverless, Databricks pre-warms a shared warehouse pool, owns the infrastructure, and charges a single all-in DBU rate. Apples-to-apples (DBU + VM cost), SQL Serverless is usually cheaper for bursty SQL workloads because warm-pool removes 1-3 minutes of cluster-start per query. SQL Serverless is more expensive when your workload is steady-state high-throughput — at that point a dedicated SQL Warehouse on All-Purpose right-sizes better. Rule of thumb: serverless wins for unpredictable BI with idle gaps; dedicated wins for always-on 9-to-5 dashboard fleets.
No. Compute types, tiers, cloud-provider picks, DBU/hr rates, hours per day, days per month, and cluster counts all run locally in your browser. The page fires an anonymous pageview beacon and CTA-click events so we can measure whether the calculator is useful — no inputs, no email (unless you submit one to the cheat-sheet form), no IP stored raw.