# Missing Databricks Commit Unit Prepurchase for Steady Azure Databricks Usage Canonical: https://www.pointfive.co/efficiency-hub/inefficiencies/missing-databricks-commit-unit-prepurchase-for-steady-azure-databricks-usage Azure Databricks charges DBUs for every workload on top of the virtual machines, storage and networking it runs on. By: PointFive Updated: 2026-09-28 [Cloud Efficiency Hub](https://www.pointfive.co/efficiency-hub) The short version Azure Databricks charges DBUs for every workload on top of the virtual machines, storage and networking it runs on. PointFive Research Cloud cost research at PointFive Azure service [Azure Databricks](https://www.pointfive.co/efficiency-hub/cloud-services/azure-databricks) Category [Compute](https://www.pointfive.co/efficiency-hub/service-category/compute) Reference CER-0452 Type Suboptimal Pricing Model ## Explanation Why the waste happens and who it affects. Teams that run Databricks at a steady, predictable level for months often pay the pay-as-you-go DBU rate for all of it, because commitment reviews focus on VM reservations and savings plans, which do not cover DBUs. Microsoft offers a Databricks commit unit (DBCU) prepurchase for one or three years that discounts DBU usage across all Azure Databricks workloads and tiers. Without it, a stable DBU baseline is billed at list price every month. The Azure Well-Architected guide for Azure Databricks recommends DBCU prepurchase for predictable workloads with stable usage. Because purchases are final and the discount depends on the tier size, the purchase has to be sized carefully rather than skipped. ## Billing model The pricing dimensions that drive this cost. DBU charges and the underlying Azure infrastructure are billed separately, and the prepurchase discounts only the DBU portion. DBU The Databricks processing unit billed per DBU-hour at a price that depends on workload type and tier; enabling Photon increases the DBU count Databricks commit unit (DBCU) A prepaid pool bought for one or three years in fixed purchase tiers, with a larger discount for larger tiers Drawdown at list-price ratio Each DBU consumed deducts DBCUs equal to that workload's DBU list price, so one pool covers all workloads and tiers Term expiry DBU usage draws down the pool until it is exhausted or the term ends; usage beyond the pool is billed at pay-as-you-go rates No cancellation or exchange Databricks prepurchase plans cannot be canceled, exchanged, split or merged; all purchases are final ## How to detect 5 checks to find it in your estate. - In Azure portal Reservations, check whether any Databricks prepurchase exists for the billing scope that owns the workspaces - In Cost Management, filter to the Azure Databricks service and review monthly DBU cost over at least three to six months to confirm a stable baseline - Query system.billing.usage grouped by sku\_name and usage\_date to see DBU consumption by workload type, then multiply by the list price per DBU to normalize the baseline into DBCU - Compare the normalized annual DBCU baseline against the purchase tiers on the Azure Databricks pricing page; on the US list (checked September 2026) 1-year tiers run from a 4% discount at 12,500 DBCU to 33% at 2,000,000 DBCU, and 3-year tiers reach 37% at 6,000,000 DBCU, so small estates gain little - Account for planned changes, such as moving jobs to serverless or enabling Photon, that will change future DBU consumption ## How to fix 4 ways to remove the waste. - Buy a DBCU prepurchase sized to a conservative share of the projected DBU baseline for the term, since unused units are lost at term end and the purchase cannot be canceled or exchanged - Choose shared or management group scope so the pool covers every workspace in the billing context rather than one subscription - Cover the underlying VMs separately with VM reservations or a savings plan for compute; the DBCU discount does not apply to VM, storage or networking charges - Track pool drawdown against the term and plan a top-up or renewal only after reviewing actual consumption, instead of buying a large three-year tier up front ## Documentation Vendor references for pricing and configuration. - [Optimize Azure Databricks costs with a prepurchase - Microsoft Cost Management learn.microsoft.com](https://learn.microsoft.com/en-us/azure/cost-management-billing/reservations/prepay-databricks-reserved-capacity) - [Azure Databricks Pricing azure.microsoft.com](https://azure.microsoft.com/en-us/pricing/details/databricks/) - [Architecture Best Practices for Azure Databricks - Microsoft Azure Well-Architected Framework learn.microsoft.com](https://learn.microsoft.com/en-us/azure/well-architected/service-guides/azure-databricks) - [Billable usage system table reference - Azure Databricks learn.microsoft.com](https://learn.microsoft.com/en-us/azure/databricks/admin/system-tables/billing) ## Related inefficiencies [Browse the library](https://www.pointfive.co/efficiency-hub) - Azure Databricks CER-0068 ### [Inefficient Private Link Routing to Azure Databricks](https://www.pointfive.co/efficiency-hub/inefficiencies/inefficient-private-link-routing-to-azure-databricks) In Azure Databricks environments that rely on Private Link for secure networking, it's common to route traffic through multi-tiered network architectures. This often includes multiple VNets, Private Link endpoints, or peered subscriptions... Networking - Azure Virtual Machines CER-0240 ### [Azure Hybrid Benefit Not Enabled on Virtual Machines](https://www.pointfive.co/efficiency-hub/inefficiencies/azure-hybrid-benefit-not-enabled-on-virtual-machines-a8f4b) Many organizations purchase Software Assurance or subscription-based Windows and SQL Server licenses that entitle them to use Azure Hybrid Benefit. However, if the setting is not applied on eligible resources, Azure continues charging... Compute - Azure Virtual Machines CER-0104 ### [Suboptimal Architecture Selection for Azure Virtual Machines](https://www.pointfive.co/efficiency-hub/inefficiencies/suboptimal-architecture-selection-for-azure-virtual-machines) Azure offers x86-64 VMs (Intel and AMD) and Arm64 VMs (Ampere Altra and Azure Cobalt 100), but default provisioning often leans toward Intel-based SKUs due to inertia or pre-configured templates. AMD and ARM alternatives offer lower list... Compute --- Source: the public page above. Product screenshots and illustrative interfaces are examples, not live customer data.