Databricks licensing is the set of terms and price mechanisms under which Databricks, Inc. sells its Data + AI Platform: data engineering (Lakeflow Jobs and Pipelines), SQL warehousing, interactive data science compute, model serving, the Lakebase Postgres database, storage and related managed services. Databricks does not sell software licences in the classic sense. Customers buy consumption of a cloud service, metered mainly in Databricks Units (DBUs) and charged per SKU at a published list price.[1] The contract is the Master Cloud Services Agreement (MCSA), which governs “access to and use of the Databricks Services” and was last updated on 2026-02-20.[2]
The platform runs on three public clouds, but the commercial route differs by cloud. Databricks sells its Platform Services directly on Amazon Web Services and Google Cloud Platform. Azure Databricks is a “Databricks Powered Service” provided by Microsoft Corporation, and a customer must contract with Microsoft to use it.[3] The pricing page says that charges for Azure Databricks “are billed directly by Microsoft and are governed by your Azure subscription terms”.[1] The MCSA expressly does not govern Databricks Powered Services.[2] Catalog proof: Azure Databricks is priced and billed by Microsoft.
Editions
Databricks sells platform tiers, not product editions. On AWS there are two tiers, Premium and Enterprise, and new accounts start on Premium.[5] The former Standard tier has reached end of life. Remaining Standard workspaces on AWS and Google Cloud were upgraded automatically to Premium on 2025-10-01.[5][15] The Lakeflow Jobs pricing page gives 2026-10-01 as the upgrade date for Azure.[4] The tier names do not line up across clouds: Databricks notes that the Premium tier on Azure Databricks corresponds to the Enterprise tier on AWS and Google Cloud.[4] Catalog proof: Platform tiers are Premium and Enterprise; Standard tier retired.
Two platform add-ons sit on top of the tiers. Enhanced Security and Compliance is charged at 15% of Product Spend. Mission Critical, which adds Managed Disaster Recovery and includes Enhanced Security and Compliance, is listed at 30% of Product Spend, with a promotional rate of 15% until 2027-06-30.[7]
Inside a tier, the price depends on the workload type. Each one is a separate SKU family with its own list price per DBU.
| Product line | Main SKUs | AWS list price, Premium (USD per DBU) | Deeper article |
|---|---|---|---|
| Lakeflow Jobs | Jobs Compute, Jobs Serverless Compute | 0.15 classic; 0.35 serverless[4] | DBU pricing and platform tiers |
| Databricks SQL | SQL Compute, SQL Pro Compute, Serverless SQL Compute | 0.22; 0.55; 0.70[12] | same |
| Interactive compute | All-Purpose Compute, All-Purpose Serverless Compute | 0.55; 0.75[17] | same |
| Lakeflow Pipelines | DLT Core, Pro and Advanced Compute; serverless | 0.20 to 0.36 classic; 0.35 serverless[18] | same |
| Storage | Databricks Storage | 0.023 per DSU[8] | same |
| Commitments | AWS, GCP, Azure, SAP and Flexible Commits | Negotiated | Committed-use contracts and billing |
Metrics
The Databricks Unit (DBU) is the main unit. Databricks defines it as “a normalized unit of processing power on the Databricks Lakehouse Platform used for measurement and pricing purposes”. The number of DBUs a workload consumes depends on processing metrics such as the compute resources used and the amount of data processed.[1] Other units cover other resources.
- The Databricks Storage Unit (DSU) measures storage. Stored data counts as 1 DSU per GB-month on every cloud, and operations add DSUs per 1,000 requests.[8]
- A Gigabyte for networking charges is 2^30 bytes. A private connectivity endpoint hour is rounded up to the nearest whole hour.[1]
- Lakebase compute is charged per Capacity Unit hour.
- Add-ons are charged as a percentage of Product Spend, which is spend at list price before discounts.[7]
Every usage record carries a SKU name and a usage unit in the system.billing.usage system table. The billing_origin_product column shows which product produced the usage.[9]
Counting / floors
Pay-as-you-go use is metered “at per second granularity”.[1] Serverless compute has a one-minute minimum and does not charge for virtual-machine start-up time.[4] On classic compute, the DBU price excludes the virtual machines, which the cloud provider bills separately, “including idle time when leveraging instance pools”. Enabling Photon on a classic Jobs cluster raises the DBU emission rate by 2.9 times compared with non-Photon.[4] Catalog proof: Serverless compute has a one-minute minimum and excludes VM start-up time; Classic compute VMs are billed separately by the cloud provider; Photon raises the DBU emission rate on classic compute.
There are no user or core minimums. The floor in most deals is the commitment itself. Under a committed-use contract, any part of the Universal Usage Commitment that is still unpaid at the end of the Term is owed as a True-up Payment.[6] Catalog proof: Unconsumed commitment is payable as a True-up Payment.
Virtualization & partitioning
Databricks does not license per processor, core or host, so partitioning rules of the kind described in virtualization and partitioning do not apply. On classic compute, the instance type affects how many DBUs a cluster emits per hour. Databricks publishes the available instance types per cloud through its Cloud Provider Directory.[3] Serverless products remove instance choice: Databricks picks and scales the resources and charges only DBUs.[4]
Cloud / BYOL
There is no bring-your-own-licence model. The customer brings its own cloud account for classic compute and pays the cloud provider for instances, while Databricks charges DBUs. Serverless runs in a Databricks-managed compute plane and its price includes the instances.[4] Customers can buy Databricks on AWS through AWS Marketplace, and Databricks on Google Cloud is tied to a Google billing account and order through Google Cloud Marketplace.[5][15] Under a commitment, the Additional Billing and Commitment Terms name a “Billing Source” for each platform: Databricks or AWS Marketplace for AWS, Databricks or GCP Marketplace for Google Cloud, Azure for Azure Databricks, and SAP for SAP Databricks.[14] Catalog proof: Billing Source determines who invoices each platform.
Programs
- Pay-as-you-go. There are no up-front costs. Credit-card accounts are billed monthly at Price List rates.[1]
- Universal Usage Commitment. A committed-use contract can be split into Specified Commitments per platform plus a Flexible Commit. It gives discounts that grow with the commitment size.[1][6]
- SKU Groups. These are cross-service and service-specific groupings that contracts use to apply pricing. Some SKUs are excluded.[13]
- Promotional Discount. These are temporary, generally available discounts flagged on the Price List.[1]
- Free trial and Free Edition. The trial gives usage credits valid for 14 days and then converts to pay-as-you-go.[11] Free Edition is a no-cost, quota-limited workspace that “may not be used for commercial purposes”.[10]
- Support plans. Business, Production and Mission Critical plans are sold with an Order.[16]
The contract stack is covered in Databricks Master Cloud Services Agreement.
Compliance and measurement
Databricks meters consumption itself, so it does not rely on customer declarations. The retrieved MCSA lets Databricks collect “Usage Data”, meaning telemetry about use of the Platform Services, and contains no licence audit clause.[2] The practical compliance questions are financial rather than about entitlement. They include whether usage is burning down the commitment at the planned rate, which workspaces and SKUs drive spend, and whether usage on a given platform or Powered Service counts toward the commitment at all. The Cloud Provider Directory warns that usage of Powered Services “may not count towards your Universal Commitment / Minimum Commitment” unless Databricks billing confirms otherwise in writing.[3] The MCSA also bars reselling or giving third parties access to the Platform Services except as the Agreement allows.[2] Catalog proof: Databricks may collect Usage Data on platform use; Powered Services and some workspaces may not count toward the commitment; Platform Services may not be resold or provided to third parties.
Out of scope
This page does not cover open-source projects that Databricks contributes to, such as Apache Spark, Delta Lake and MLflow, which have their own open-source licences. It also leaves out Azure Databricks price lists and contract terms, which Microsoft sets, the SAP Databricks commercial terms inside SAP Business Data Cloud, third-party listings on Databricks Marketplace, and Advisory and Training Services.