Azure Databricks pricing: Serverless SQL at $0.70 per DBU and warehouse sizes
Azure Databricks charges for compute in Databricks Units (DBUs), and the rate per DBU depends on the type of workload. As listed in Microsoft's official Azure retail price feed for the East US region on November 11, 2026, a Premium Serverless SQL warehouse costs $0.70 per DBU-hour, a Premium SQL Pro warehouse costs $0.55 per DBU-hour and Premium Jobs Compute costs $0.30 per DBU-hour, all in USD at pay-as-you-go list prices.
The DBU rate is only half of the calculation. The other half is how many DBUs a warehouse burns per hour, and Microsoft documents that figure for serverless SQL warehouses by size: a 2X-Small consumes 4 DBUs per hour, a Small consumes 12 and a 4X-Large consumes 528. Multiplying the two gives an hourly cost that runs from $2.80 for the smallest serverless SQL warehouse to $369.60 for a 4X-Large in East US.
For BI teams that point Power BI, Tableau or SQL clients at Databricks, the warehouse size and the number of hours it runs decide most of the bill. The Standard tier, which had lower DBU rates, is no longer an option: Microsoft's pricing page states that the Azure Databricks Standard tier will be retired on October 1, 2026, so new planning starts from Premium rates.
··········
AZURE DATABRICKS DBU RATES BY WORKLOAD IN EAST US.
Microsoft defines a DBU on its Azure Databricks pricing page as "a unit of processing capability per hour, billed on per-second usage." Each workload type has its own meter in the Azure price feed, and the meters below are the Premium pay-as-you-go list prices returned for East US when checked on November 11, 2026. Prices are published per region, so a workspace in another Azure region can show different figures.
........
Workload (Premium tier) | Price per DBU-hour (USD) | Typical use |
Serverless SQL | $0.70 | BI dashboards and SQL queries |
SQL Pro Compute | $0.55 | Pro SQL warehouses |
SQL Compute (classic) | $0.22 | Classic SQL warehouses |
All-Purpose Compute | $0.55 | Interactive notebooks on clusters |
Jobs Compute | $0.30 | Scheduled jobs on clusters |
Jobs Light Compute | $0.22 | Lightweight scheduled jobs |
Interactive Serverless Compute | $0.95 | Serverless notebooks |
Automated Serverless Compute | $0.45 | Serverless jobs and pipelines |
Model Training | $0.65 | Fine-tuning and training |
Database Serverless Compute | $0.26 | Lakebase compute |
........
The classic SQL rate appears in the Azure price feed under the meter name "Premium SQL Analytics DBU." Photon-enabled All-Purpose and Jobs compute carry the same per-DBU rate as the non-Photon meters ($0.55 and $0.30), but Microsoft's page warns that "Enabling Photon will increase DBU count," so the hourly cost of a Photon cluster is higher even though the unit price is unchanged.
Storage that Databricks manages for you is billed separately in Databricks Storage Units (DSUs), listed at $0.026 per DSU in East US. Microsoft's DBU consumption page applies multipliers to the DSU for different storage features, such as 1X per GB of stored data and 10X for AI Search storage.
··········
SERVERLESS SQL WAREHOUSE SIZES AND HOURLY COST.
Microsoft's "Serverless DBU consumption by SKU" page, last updated in September 2026, lists the DBUs per hour for each serverless SQL warehouse size. The hourly cost column below multiplies those DBUs by the $0.70 East US list rate and is calculated, not published by Microsoft.
........
Serverless SQL size | DBUs per hour | Hourly cost at $0.70 (USD) |
2X-Small | 4 | $2.80 |
X-Small | 6 | $4.20 |
Small | 12 | $8.40 |
Medium | 24 | $16.80 |
Large | 40 | $28.00 |
X-Large | 80 | $56.00 |
2X-Large | 144 | $100.80 |
3X-Large | 272 | $190.40 |
4X-Large | 528 | $369.60 |
........
The DBUs per hour climb steeply with each size, from 4 at 2X-Small to 528 at 4X-Large, and the hourly cost rises in the same proportion. On the Azure pricing page, the Serverless SQL footnote adds that "In addition to DBU, data transfer, NAT gateway, and Private Endpoint per-hour usage will also be billed," so a serverless warehouse with private networking or heavy cross-region traffic will cost more than the DBU line alone.
Databricks recommends serverless SQL warehouses for most workloads in its sizing documentation, updated October 8, 2026. Serverless warehouses use what Databricks calls Intelligent Workload Management, which queues queries when capacity is short, adds capacity when queue time grows and scales down when demand drops. The same page advises starting with a single larger warehouse and sizing down if needed, and it suggests increasing the cluster size when the query profile shows queries spilling to disk.
··········
SQL PRO AND CLASSIC WAREHOUSES ON AZURE VIRTUAL MACHINES.
Pro and classic SQL warehouses have lower DBU rates ($0.55 and $0.22 in East US), and they run on Azure virtual machines that need vCPU quota in your subscription. Microsoft's sizing page maps each size to instance types: a Small warehouse uses a Standard_E16ds_v4 driver and four Standard_E8ds_v4 workers, and a Medium uses a Standard_E32ds_v4 driver with eight Standard_E8ds_v4 workers. The Azure pricing page presents classic compute as a virtual machine price plus a DBU price, so the per-DBU comparison with serverless leaves out the VM charge that pro and classic warehouses also incur.
Microsoft's serverless DBU page does not list DBUs per hour for pro and classic warehouse sizes, so a like-for-like cost comparison needs the figures from your own billing data or the Azure pricing calculator. The scaling rules for pro and classic are documented, though: these warehouses have "a fixed limit of one cluster per 10 concurrent queries," you set a minimum and maximum number of clusters, a query waiting five minutes in the queue triggers a scale-up, and the warehouse scales down after 15 consecutive minutes of low load. Each extra cluster adds its own DBUs and VM hours for as long as it runs.
For a team that already uses serverless SQL, a related point of comparison is the BigQuery on-demand vs editions pricing model, where Google charges per TiB scanned or per slot-hour instead of per warehouse hour.
··········
WORKED EXAMPLE: MONTHLY SERVERLESS SQL COST FOR A BI TEAM.
The following example is illustrative and uses only the East US list prices and the documented DBU consumption above. A BI team runs one Small serverless SQL warehouse for production dashboards ten hours a day on 22 business days a month, and a 2X-Small warehouse for development six hours a day on the same days.
........
Warehouse | DBUs per hour | Hours per month | DBUs per month | Monthly cost (USD) |
Small (production) | 12 | 220 | 2,640 | $1,848.00 |
2X-Small (development) | 4 | 132 | 528 | $369.60 |
Total | 3,168 | $2,217.60 |
........
At that pace the two warehouses come to $26,611.20 a year at list price. Moving the production warehouse up to Medium for the same 220 hours would double its DBUs to 5,280 and its monthly cost to $3,696.00, so the warehouse size, checked against the query profile and the Peak Queued Queries metric, has a direct effect on the monthly total.
The same 3,168 DBUs a month would cost $1,397.09 if the team secured the maximum discount of 37% that Microsoft advertises for Databricks Commit Units (DBCU), a saving of $820.51 a month. Microsoft describes DBCU as one- or three-year pre-purchases that offer "up to 37% savings over pay-as-you-go DBU prices," so the actual discount depends on the size and term of the commitment. The example excludes data transfer, NAT gateway and Private Endpoint charges, storage and any Genie usage beyond the free allowance.
A similar sizing exercise for Snowflake is in the Snowflake pricing guide to credits, editions and warehouse sizes, where warehouse size also sets the credits consumed per hour.
··········
GENIE, ADD-ONS AND OTHER AZURE DATABRICKS CHARGES.
Several items on the Azure pricing page change the final bill for analytics teams. Genie, the natural-language analytics feature that Databricks recently brought to Slack and Microsoft Teams in general availability, "provides 150 DBUs of free usage every month," and Microsoft's page states that Genie usage is subject to a limited-time promotional price of 25%, with the offer expiring on January 31, 2027.
The Enhanced Security and Compliance add-on is priced at 10% of list price on top of the Azure Databricks spend in the selected workspace, and the price feed lists it at $0.10 per DBU-hour in East US. The page also mentions a Mission Critical add-on at 30% of list, with a 50% promotional discount until June 30, 2027. Some serverless features consume DBUs with a multiplier: Microsoft's consumption page lists Data Quality Monitoring at 2X and Data Classification at 3X on the Automated Serverless SKU, and it states that the DBUs shown on the bill and in system tables already include the multiplier.
The table below summarizes what is published for these items.
........
Item | Published price or rule | Source page |
Genie | 150 free DBUs per month, 25% promo price to Jan 31, 2027 | Azure Databricks pricing |
Enhanced Security and Compliance | 10% of list price | Azure Databricks pricing |
Mission Critical add-on | 30% of list, 50% promo to Jun 30, 2027 | Azure Databricks pricing |
Databricks Commit Units | Up to 37% off pay-as-you-go DBUs | Azure Databricks pricing |
Standard tier | Retired October 1, 2026 | Azure Databricks pricing |
........
··········
AZURE DATABRICKS COSTS TO CHECK IN YOUR OWN TENANT.
The Azure price feed and Microsoft's documentation confirm the East US Premium rates of $0.70 per DBU for Serverless SQL, $0.55 for SQL Pro and $0.22 for classic SQL, along with the documented 4 to 528 DBUs per hour for serverless SQL warehouse sizes from 2X-Small to 4X-Large. They also confirm that serverless networking charges are billed on top of DBUs and that the Standard tier was retired on October 1, 2026.
Three things depend on your setup. Rates differ by Azure region, the DBCU discount depends on the commitment you negotiate, and pro or classic warehouses add Azure virtual machine charges that the DBU rate does not show. Microsoft's serverless consumption page does not publish DBUs per hour for pro and classic warehouse sizes, so those comparisons need your own usage data from system tables or the Azure pricing calculator.
Before budgeting, check the region of each workspace, the size and running hours of every SQL warehouse, whether queries spill to disk, and whether Photon or private networking is enabled. Teams comparing Databricks with Microsoft's own analytics stack can set these figures against Power BI Pro, Premium Per User and Fabric capacity costs for the same user base.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org
