The Databricks Master Cloud Services Agreement (MCSA) is the master contract between Databricks, Inc. and its customers for the Databricks Services. These are the Platform Services, Support Services, Training Services and Advisory Services.[1] The retrieved version shows “Last Updated February 20, 2026”.[1] The Agreement consists of the MCSA (with the Data Processing Addendum and Security Addendum), the Service Specific Terms and any Orders. In a conflict, “the order of precedence shall be as follows: (a) the applicable Order, (b) the Service Specific Terms, and (c) this MCSA”.[1] Catalog proof: Order prevails over Service Specific Terms and the MCSA.
The MCSA’s Effective Date is the effective date of the first Order that references it. For a pay-as-you-go customer, it is the date of first access. For PayGo Customers, Databricks “may make changes to the MCSA and pricing”, and continued use constitutes consent to the then-current version.[1] Customers buying through a reseller are governed by End Customer Terms, which form part of their Agreement.[1]
Editions
The MCSA applies in the same way to every platform tier. Tiers and add-ons are commercial choices made in the Order or the account console, as described in Databricks DBU pricing and platform tiers. Other terms apply alongside or instead of the MCSA in four cases:
| Offering | Governing terms | Key limits |
|---|---|---|
| Paid platform (PAYG or Order) | MCSA, Service Specific Terms, Acceptable Use Policy for the paid version[3] | Order controls |
| Free trial | MCSA; trial services provided “AS-IS”[1] | 14-day credits; Databricks liability capped at USD 5,000 |
| Free Edition | Databricks Free Edition Terms; Free Edition Acceptable Use Policy[4] | Personal, not-for-profit use only |
| Azure Databricks | Microsoft’s terms; not governed by the MCSA[1][12] | Billed by Microsoft |
Metrics
The MCSA itself does not define a licence metric. Units and prices come from the Order and the Price List, mainly the Databricks Unit (DBU).[11] The MCSA does define who may use the service. An “Authorized User” is any of the “employees, personnel, or agents of Customer or its Affiliates provisioned by Customer to access and use the Databricks Services”.[1] There is no per-user fee. Users matter for responsibility, because the customer answers for its Authorized Users’ compliance and for any use under their accounts.[1]
Counting / floors
Use rights and restrictions
Databricks makes the Platform Services available “on a non-exclusive basis in accordance with the Agreement solely for Customer’s or its Affiliate’s business purposes”.[1] The customer and its Authorized Users may not do any of the following:
- violate the Acceptable Use Policy or applicable laws
- use the Platform Services other than as documented
- reverse engineer the Platform Services or Deliverables
- “sell, resell, license, sublicense, distribute, rent, lease, or otherwise provide access to the Platform Services to any third party except as expressly permitted under the Agreement”[1]
Affiliates may sign their own order forms under the same Agreement.[1] Catalog proof: Platform Services may not be resold or provided to third parties.
Usage Data instead of audits
The retrieved MCSA has no clause that lets Databricks audit the customer’s licence use. Databricks meters consumption itself. “Usage Data” is “data and telemetry collected by Databricks relating to Customer’s use of the Platform Services”, which may include queries entered by an Authorized User but not their results. Databricks may collect and use it “to develop, improve, operate, and support its products and services”.[1] In practice, compliance means reconciling metered usage with the Order rather than defending a deployment count, as described in software license audit. Catalog proof: Databricks may collect Usage Data on platform use.
Payment, suspension and termination
Fees are paid as the Order specifies. Unless the Order says otherwise, they are paid in U.S. dollars, invoices are due within 30 days, and past-due amounts accrue 1.5% interest per month.[1] Databricks may suspend Workspaces immediately if it reasonably suspects a breach of the customer responsibilities that may cause material harm. It may also suspend them after giving notice of delinquent, undisputed Fees.[1] Either party may terminate the MCSA on 30 days’ notice if no Orders are outstanding, or for an uncured material breach. Termination of the MCSA ends all Orders.[1] Cancelling a pay-as-you-go plan in the account console “is permanent and terminates your relationship with Databricks”. Customer Content is deleted within 30 days of cancellation.[8] Catalog proof: Invoices are due in 30 days in USD with 1.5% monthly interest; Cancelling a pay-as-you-go plan is permanent.
Liability
Each party’s aggregate liability is capped at the amount paid for the Databricks Services giving rise to the liability in the twelve months before the first incident. This is the General Cap. Unauthorized disclosure of Customer Content caused by Databricks’ breach of confidentiality, the Security Addendum or the Data Processing Addendum is capped at twice that amount, the Super Cap. Databricks’ liability for free-trial and beta services is limited to USD 5,000.[1] Databricks defends customers against third-party IP claims about the Databricks Services. The defence excludes claims about the public open-source versions of software such as Apache Spark, unless the claim specifically concerns the Platform Services.[1]
Virtualization & partitioning
Not addressed. The MCSA licenses access to a managed cloud service, not installable software, so it contains no host, partition or core rules.
Cloud / BYOL
The MCSA defines a “Cloud Service Provider” as the third-party provider of cloud infrastructure on which Databricks provides the Platform Services. It treats “Databricks Powered Services” separately. These are third-party services powered by Databricks and listed in the Cloud Provider Directory, which the MCSA “does not govern”.[1] The Directory lists AWS and Google Cloud for direct Platform Services and Azure Databricks, provided by Microsoft, as the Powered Service.[12] The customer remains responsible for securing its own cloud environments used with the Databricks Services.[1]
Programs
Service Specific Terms
The Service Specific Terms, last updated 2026-07-14, incorporate Databricks’ service level agreements and set rules for AI features.[2]
- Inputs and Outputs. These are Customer Content.
- Customized models. A customer that customizes AI models has exclusive use of them.
- Training. Databricks “shall not use Customer Content, Inputs, or Outputs to train, re-train, or fine-tune GenAI Models that Databricks makes available for use by third parties”.[2]
- GenAI indemnity. Databricks defends claims that text Output of a “Covered AI Service” infringes third-party IP, without a damages cap. Covered AI Services are generally available Databricks AI assistive features, and Model Serving with listed foundation models used with safeguards (system prompt, AI Gateway logging and filtering, red-team testing). The indemnity excludes free-trial use.[2]
See generative AI licensing. Catalog proof: GenAI indemnity covers only qualifying generally available AI services.
Support
If an Order specifies no Support Services, “support shall be limited to the Documentation”.[1] The Support Policy offers Business, Production and Mission Critical plans. They differ in response times, from one business day down to 15 minutes for mission-critical severity, and in support hours and access to Spark experts. Support “Contacts” are unique named users, and accounts may not be shared.[10] See Databricks support plans.
Free trial
The free trial gives “credit for free Databricks usage, valid for 14 days”. Trial accounts created with personal email addresses may face limits, such as one SQL warehouse scaling to at most 50 DBUs per hour.[7] At the end of the trial, the account is automatically subscribed to the plan used during the trial.[11] See Databricks free trial.
Free Edition
Databricks Free Edition is “the no-charge version of portions of the cloud-based Databricks Data + AI Platform”.[4] It replaced Community Edition, which was retired in 2025.[6] Its terms differ sharply from the MCSA:
- Use “must be personal and not-for-profit”.
- Databricks receives a “royalty-free, perpetual, irrevocable, and worldwide license” to the user’s content, including to train AI models.
- Databricks may stop the service or delete content at any time.
- Databricks’ liability is capped at the greater of USD 100 or the amounts paid.[4]
The documentation adds that Free Edition accounts “may not be used for commercial purposes”. Accounts are serverless-only and quota-limited, have no account console, SSO, SCIM, support or SLA, and cannot be upgraded to the full platform.[5][8] Free Edition workspaces used inside an organization for work purposes therefore fall outside both the free terms and any paid Order. Catalog proof: Free Edition is limited to non-commercial use.
U.S. public sector
The U.S. Public Sector Services Schedule, last updated 2026-05-07, modifies the MCSA for federal, state and local government customers, including public hospitals and universities. For Federal Customers, the Platform Services are “commercial computer software” under the FAR. Disputes follow the Contract Disputes Act. Commits, support and subscriptions are payable upfront on Net 30. Auto-renewal, late-payment interest and customer indemnities do not apply where the Anti-Deficiency Act restricts them.[9] Under the schedule’s reseller clause, Databricks may share consumption metrics with the reseller that fulfils an order.[9] See U.S. Public Sector Services Schedule.
Out of scope
This page does not cover the Data Processing Addendum, the Security Addendum, the EU Data Act addendum or the End Customer Terms for reseller purchases in detail. It also does not cover Microsoft’s terms for Azure Databricks or the open-source licences of Apache Spark, Delta Lake and MLflow.