AWS
Microsoft Azure
Google Cloud
Key takeaways
1
Classic compute: Pay for DBUs plus Azure VMs.
2
Serverless compute: Infrastructure is included in DBU pricing.
3
Monthly cost: Depends on workload, size, runtime, and region.
4
Savings: Reduce idle compute before purchasing commitments.
How Azure Databricks Pricing Works
Azure Databricks uses consumption-based pricing. The amount you pay depends on the resources running your workloads and the billing model they use.The first distinction is between classic compute and serverless compute.
| Cost component | Classic compute | Serverless compute |
|---|---|---|
| Databricks Units (DBUs) | Charged separately | Included in serverless DBU rate |
| Azure virtual machines | Billed separately | Included in serverless compute pricing |
| Storage | Additional charges may apply | Additional charges may apply |
| Networking | Additional charges may apply | Additional charges may apply |
| Main pricing driver | Cluster size and running time | Metered workload consumption |
Microsoft publishes the current workload rates and eligible configurations on its official Azure Databricks pricing page.
Although prices are quoted per DBU-hour, usage is generally metered at per-second granularity.
For classic compute:
What Is a Databricks Unit (DBU)?
A Databricks Unit (DBU) measures processing capability consumed by Databricks workloads. DBU consumption depends on the compute configuration and workload type.Although prices are quoted per DBU-hour, usage is generally metered at per-second granularity.
For classic compute:
DBU cost = DBUs per hour × runtime in hours × DBU rate
For example, a workload consuming 20 DBUs per hour for 10 hours at an illustrative rate of $0.30 per DBU-hour costs:
20 × 10 × $0.30 = $60 in DBU charges
This excludes the separately billed Azure VMs and other resources.
For more accurate estimates, use the relevant SKU and region from Microsoft’s pricing tables. For example, Microsoft lists Premium Jobs Compute at $0.30/DBU-hour and All-Purpose Compute at $0.55/DBU-hour, before separate classic VM costs.
For more accurate estimates, use the relevant SKU and region from Microsoft’s pricing tables. For example, Microsoft lists Premium Jobs Compute at $0.30/DBU-hour and All-Purpose Compute at $0.55/DBU-hour, before separate classic VM costs.
Azure Databricks Pricing by Workload
Different Databricks workloads use different pricing rates and compute configurations. The following are published US-dollar pay-as-you-go DBU rates shown on the official Microsoft Azure Databricks pricing page (checked October 20, 2026). They are not all-in hourly cluster prices. Verify region, contract, and current rates before purchasing.| Workload | Published rate | Typical use / pricing consideration |
|---|---|---|
| Jobs Light Compute | $0.22/DBU-hour | Supported lighter-weight classic jobs; VMs extra |
| Jobs Compute | $0.30/DBU-hour | Scheduled classic jobs and ETL; VMs extra |
| All-Purpose Compute | $0.55/DBU-hour | Interactive classic clusters; VMs extra |
| Automated Serverless Compute | $0.47/DBU-hour | Serverless jobs and supported automated workloads |
| Interactive Serverless Compute | $1.00/DBU-hour | Serverless interactive workloads |
| SQL Compute | $0.22/DBU-hour | Classic SQL compute; VMs extra |
| SQL Pro Compute | $0.55/DBU-hour | Classic SQL Pro; VMs extra |
| Serverless SQL | $0.70/DBU-hour | Managed SQL warehouses |
| Serverless Real-Time Inference | $0.082/DBU-hour | Specialized inference, not a general cluster starting price |
Rates shown are workload-specific DBU rates. For classic compute, add VM and applicable ancillary charges. Some serverless features use DBU multipliers or separate charges.
These workloads should not be compared using DBU rates alone. Their runtime, infrastructure requirements, and billing meters can differ.
All-Purpose Compute is generally used for interactive development, while Jobs Compute is designed for automated processing.
For either workload, calculate the Databricks platform charges and underlying Azure infrastructure charges separately.
Its DBU rate includes the underlying infrastructure rather than requiring an additional classic-cluster VM charge.
However, serverless is not automatically cheaper.
Compare the total expense of completing the same workload, including execution time, DBU consumption, storage, and relevant additional charges.
Microsoft recommends benchmarking representative workloads and reviewing their actual billing records rather than relying on assumed utilization.
For supported services and billing categories, see Microsoft’s serverless DBU consumption documentation.
These workloads should not be compared using DBU rates alone. Their runtime, infrastructure requirements, and billing meters can differ.
Classic All-Purpose and Jobs Compute
Classic clusters run on Azure virtual machines selected for the workload.All-Purpose Compute is generally used for interactive development, while Jobs Compute is designed for automated processing.
For either workload, calculate the Databricks platform charges and underlying Azure infrastructure charges separately.
Azure Databricks Serverless Pricing
Serverless compute simplifies infrastructure management because Databricks manages the compute resources.Its DBU rate includes the underlying infrastructure rather than requiring an additional classic-cluster VM charge.
However, serverless is not automatically cheaper.
Compare the total expense of completing the same workload, including execution time, DBU consumption, storage, and relevant additional charges.
Microsoft recommends benchmarking representative workloads and reviewing their actual billing records rather than relying on assumed utilization.
For supported services and billing categories, see Microsoft’s serverless DBU consumption documentation.
Azure Databricks Pricing Per Hour and Month
The easiest way to estimate Azure Databricks cost is to calculate the DBU charges, add any separately billed infrastructure, and multiply by the expected runtime.Azure Databricks Monthly Cost Formula
For classic compute:
Monthly cost = DBU charges + VM charges + storage + networking + other applicable charges
Where:
DBU charges = DBU/hour × active DBU-billed hours × DBU price
VM charges = Number of VMs × billed VM hours × VM hourly price
For serverless compute, replace the separate DBU and VM components with the applicable serverless consumption charges.
Example: Azure Databricks Classic Cluster
Consider a hypothetical Jobs Compute workload with the following configuration.| Pricing input | Example value |
|---|---|
| Number of VMs | 4 |
| DBUs per VM-hour | 2 |
| DBU price | $0.30 |
| VM hourly price | $0.50 |
| Runtime | 8 hours/day |
| Operating days | 22 days/month |
The DBU rate of $0.30/DBU-hour matches Microsoft’s published Premium Jobs Compute rate, but the $0.50 VM-hour assumption and this four-VM configuration are illustrative, not a current Azure VM quote. The calculation assumes 2 DBUs per VM-hour and excludes additional charges.
Step 1: Calculate monthly runtime
8 × 22 = 176 hours
Step 2: Calculate DBU charges
4 × 2 × 176 × $0.30 = $422.40
Step 3: Calculate VM charges
4 × 176 × $0.50 = $352.00
Step 4: Calculate total compute cost
$422.40 + $352.00 = $774.40 per month
Step 1: Calculate monthly runtime
8 × 22 = 176 hours
Step 2: Calculate DBU charges
4 × 2 × 176 × $0.30 = $422.40
Step 3: Calculate VM charges
4 × 176 × $0.50 = $352.00
Step 4: Calculate total compute cost
$422.40 + $352.00 = $774.40 per month
| Component | Monthly estimate |
|---|---|
| Databricks DBUs | $422.40 |
| Azure VMs | $352.00 |
| Total compute | $774.40 |
Storage, networking, managed disks, taxes, discounts, and additional services are excluded.
The example assumes all four VMs remain provisioned for the stated runtime. Actual driver and worker configurations, autoscaling, and startup behavior can produce different costs.
The example assumes all four VMs remain provisioned for the stated runtime. Actual driver and worker configurations, autoscaling, and startup behavior can produce different costs.
What Happens If the Cluster Runs 24/7?
Using the same configuration for 730 hours:DBU charges: $1,752
VM charges: $1,460
Total estimated compute: $3,212/month
Compared with the scheduled workload, continuously running the same cluster would cost approximately $2,437.60 more per month.
This illustrates why scheduling and automatic termination are important for non-production workloads.
Actual billing may differ where clusters scale dynamically, enter idle pools, or receive pricing discounts. To compare with another published configuration, Microsoft provides a 10 × DS13v2 Jobs Compute, 100-hour example totaling $1,198 ($598 in VMs and $600 in DBUs) on its official pricing page. This is Microsoft’s example, not a guaranteed current quote for your subscription.
This illustrates why scheduling and automatic termination are important for non-production workloads.
Actual billing may differ where clusters scale dynamically, enter idle pools, or receive pricing discounts. To compare with another published configuration, Microsoft provides a 10 × DS13v2 Jobs Compute, 100-hour example totaling $1,198 ($598 in VMs and $600 in DBUs) on its official pricing page. This is Microsoft’s example, not a guaranteed current quote for your subscription.
Classic vs. Serverless Databricks: Which Costs Less?
Neither option is universally cheaper.| Consideration | Classic compute | Serverless compute |
|---|---|---|
| Infrastructure control | More configuration control | Managed by Databricks |
| VM charges | Separate | Included in serverless DBU rate |
| Capacity management | Customer manages clusters | Databricks manages capacity |
| Best comparison metric | Total cost per completed workload | Total cost per completed workload |
| Common risk | Idle or oversized VMs | Higher-than-expected DBU consumption |
For frequent workloads with predictable capacity requirements, carefully optimized classic compute can be cost-effective.
For variable workloads, serverless may reduce operational overhead and unused provisioned capacity.
The correct decision requires testing the same workload on both configurations. For the separate infrastructure-cost question, see our Azure VM cost optimization guide.
Important: A lower DBU price does not necessarily mean a lower total bill, especially when one option includes infrastructure and the other does not.
However, the Standard tier was retired on October 1, 2026. Remaining Standard workspaces were scheduled to transition automatically to Premium.
For new pricing estimates, use current Premium and applicable workload rates rather than outdated Standard-tier comparisons.
Premium provides enterprise-oriented capabilities, including additional security, governance, and access-control functionality.
Optional offerings, including Enhanced Security and Compliance and Mission Critical, may introduce additional costs.
Verify the relevant features and charges on the official pricing page.
Use the Azure Pricing Calculator or the Databricks Azure pricing page.
Follow these steps:
For variable workloads, serverless may reduce operational overhead and unused provisioned capacity.
The correct decision requires testing the same workload on both configurations. For the separate infrastructure-cost question, see our Azure VM cost optimization guide.
Important: A lower DBU price does not necessarily mean a lower total bill, especially when one option includes infrastructure and the other does not.
Azure Databricks Pricing Tiers in 2026
Azure Databricks historically offered Standard and Premium workspace tiers.However, the Standard tier was retired on October 1, 2026. Remaining Standard workspaces were scheduled to transition automatically to Premium.
For new pricing estimates, use current Premium and applicable workload rates rather than outdated Standard-tier comparisons.
Premium provides enterprise-oriented capabilities, including additional security, governance, and access-control functionality.
Optional offerings, including Enhanced Security and Compliance and Mission Critical, may introduce additional costs.
Verify the relevant features and charges on the official pricing page.
How to Estimate Azure Databricks Pricing With a Calculator
Microsoft and Databricks provide pricing tools for estimating workloads before deployment.Use the Azure Pricing Calculator or the Databricks Azure pricing page.
Follow these steps:
1
Choose your Azure region. Pricing and available configurations vary by region.
2
Select the workload. Identify Jobs Compute, All-Purpose, SQL, or the applicable serverless SKU.
3
Enter the compute configuration. For classic clusters, account for driver and worker resources, node types, and expected scaling.
4
Estimate runtime. Include daily hours, monthly execution frequency, and expected workload duration.
5
Add additional costs. Account for storage, networking, disks, add-ons, and separately billed services.
6
Apply eligible discounts. Evaluate commitments only after establishing the optimized consumption baseline.
For an existing deployment, reconcile your forecast with actual Azure charges and Databricks billing records.
Why Azure Databricks Bills Can Be Higher Than Expected
A Databricks estimate may understate costs when it overlooks resources outside the DBU rate.| Cost driver | Why it matters |
|---|---|
| Idle classic clusters | Provisioned VMs continue generating charges |
| Oversized instances | More capacity increases infrastructure and DBU costs |
| Autoscaling | Worker counts can increase during execution |
| Managed disks | Storage resources may incur additional charges |
| Data transfer | Applicable networking and egress charges increase costs |
| Additional features | Serving, pipelines, security, or other services may have separate meters |
One important distinction involves instance pools.
Microsoft states that idle instances in a pool do not generate DBU charges, but their underlying virtual machines can continue generating infrastructure charges.
Therefore, zero DBU consumption does not always mean zero Azure compute cost.
Use job-specific compute for scheduled workloads where appropriate, rather than keeping general-purpose clusters continuously available.
For non-production environments, coordinate schedules with actual working hours.
Oversized instances increase compute spending without necessarily improving execution performance.
Autoscaling can adjust available workers, but review the resulting runtime and consumption to ensure it reduces total cost.
A lower workload-specific DBU rate can help, but the final comparison must include execution duration and infrastructure.
Eligible organizations can obtain up to 37% savings on DBU charges, depending on their purchase tier and agreement.
These purchases use Databricks Commit Units (DBCUs), which are consumed by eligible DBU usage.
However, DBCU purchases do not eliminate separately billed infrastructure, storage, or networking charges.
Review Microsoft’s Databricks pre-purchase documentation before buying.
For a detailed explanation of purchasing risk, read our Azure Databricks cost commitments guide.
These discounts apply to qualifying infrastructure charges, not the Databricks software DBU meter.
Serverless compute does not expose a separate customer-billed classic VM layer to discount with those commitments.
Read our Azure Savings Plans guide and Azure VM cost optimization guide to understand eligible coverage.
The system.billing.usage table can help attribute serverless usage to jobs, notebooks, and other supported workload metadata.
Microsoft documents these fields in its serverless cost monitoring guide.
For broader financial governance, see our Azure cost management best practices.
At Usage.ai, we focus on identifying and managing eligible cloud commitment opportunities.
For classic Azure Databricks deployments, eligible underlying VM usage may support Azure compute commitment savings. This is separate from Databricks DBU pre purchase pricing.
Usage.ai can help teams evaluate eligible Azure commitment coverage and potential optimization opportunities across supported infrastructure.
This does not mean Usage.ai directly reduces Databricks DBU rates, changes cluster settings, or guarantees savings for every Databricks workload.
The important step is to determine which charges qualify before evaluating commitment-based optimization. For broader coverage and sizing, see our Azure Savings Plan guide.
Microsoft states that idle instances in a pool do not generate DBU charges, but their underlying virtual machines can continue generating infrastructure charges.
Therefore, zero DBU consumption does not always mean zero Azure compute cost.
How to Reduce Azure Databricks Costs
Cost optimization should start with workload efficiency before introducing long-term commitments.1. Schedule Clusters and Enable Auto-Termination
Terminate interactive clusters when they are no longer needed.Use job-specific compute for scheduled workloads where appropriate, rather than keeping general-purpose clusters continuously available.
For non-production environments, coordinate schedules with actual working hours.
2. Rightsize Compute and Use Autoscaling
Choose instances based on observed CPU, memory, execution time, and workload requirements.Oversized instances increase compute spending without necessarily improving execution performance.
Autoscaling can adjust available workers, but review the resulting runtime and consumption to ensure it reduces total cost.
3. Compare Jobs Compute With All-Purpose Compute
Scheduled workflows should be evaluated using job-appropriate compute configurations.A lower workload-specific DBU rate can help, but the final comparison must include execution duration and infrastructure.
4. Evaluate Databricks Pre-Purchase Discounts
Microsoft offers one-year and three-year Databricks Unit pre-purchase plans.Eligible organizations can obtain up to 37% savings on DBU charges, depending on their purchase tier and agreement.
These purchases use Databricks Commit Units (DBCUs), which are consumed by eligible DBU usage.
However, DBCU purchases do not eliminate separately billed infrastructure, storage, or networking charges.
Review Microsoft’s Databricks pre-purchase documentation before buying.
For a detailed explanation of purchasing risk, read our Azure Databricks cost commitments guide.
5. Optimize Eligible Azure VM Commitments
Classic Databricks clusters may generate Azure VM usage eligible for separate VM reservation or Azure savings plan discounts.These discounts apply to qualifying infrastructure charges, not the Databricks software DBU meter.
Serverless compute does not expose a separate customer-billed classic VM layer to discount with those commitments.
Read our Azure Savings Plans guide and Azure VM cost optimization guide to understand eligible coverage.
6. Monitor Actual Usage and Costs
Use Azure Cost Management alongside Databricks billing system tables to identify costly workloads.The system.billing.usage table can help attribute serverless usage to jobs, notebooks, and other supported workload metadata.
Microsoft documents these fields in its serverless cost monitoring guide.
For broader financial governance, see our Azure cost management best practices.
How Usage.ai Helps Optimize Eligible Azure Commitments
Optimizing Databricks runtime and managing Azure pricing commitments are separate financial decisions.At Usage.ai, we focus on identifying and managing eligible cloud commitment opportunities.
For classic Azure Databricks deployments, eligible underlying VM usage may support Azure compute commitment savings. This is separate from Databricks DBU pre purchase pricing.
Usage.ai can help teams evaluate eligible Azure commitment coverage and potential optimization opportunities across supported infrastructure.
This does not mean Usage.ai directly reduces Databricks DBU rates, changes cluster settings, or guarantees savings for every Databricks workload.
The important step is to determine which charges qualify before evaluating commitment-based optimization. For broader coverage and sizing, see our Azure Savings Plan guide.
Frequently asked questions
How much does Azure Databricks cost per month?
Azure Databricks monthly costs depend on workload type, compute configuration, runtime, and region. Classic clusters incur DBU and VM charges, while serverless compute uses bundled DBU pricing. For example, the four-VM classic cluster calculated above costs $774.40 for 176 hours using Microsoft's published $0.30 Jobs Compute DBU rate and an illustrative $0.50 VM-hour rate, before additional charges.
How much does an Azure Databricks DBU cost?
There is no single universal DBU price. Microsoft lists Jobs Light Compute at $0.22, Jobs Compute at $0.30, and All-Purpose Compute at $0.55 per DBU-hour. Specialized serverless inference is listed at $0.082 per DBU-hour. Check the official pricing page for your configuration and any separate charges.
Is Azure Databricks free?
Azure Databricks is generally a paid service. Eligible new customers may receive promotional credits or trial benefits, subject to Microsoft's and Databricks' applicable terms. Compute and other charges can apply once those benefits are exhausted.
Is Azure Databricks serverless cheaper than classic compute?
Not necessarily. Serverless includes the underlying infrastructure in its DBU charges, while classic compute bills VMs separately. Compare the complete cost of running the same workload, including processing time and additional charges.
How can I reduce Azure Databricks pricing?
Start by terminating idle clusters, rightsizing compute, using appropriate job configurations, and monitoring DBU consumption. Then evaluate DBCU pre-purchases for predictable platform usage and separate Azure VM commitments for eligible classic infrastructure. Purchase commitments only after confirming expected utilization.
Share
Facebook
X
LinkedIn
Reddit
