Analyze Microsoft Foundry costs in Azure Cost Management

To analyze billed Microsoft Foundry costs, open cost analysis in Azure Cost Management for your Foundry resource, resource group, or subscription. Then group and filter costs by meter, resource, and tag to explain spending, allocate it to projects, and reconcile it with your invoice.

The cost pages in the Foundry portal show the same Cost Management data, organized by project, agent, and model. Use Cost Management in the Azure portal when you need meters, tags, budgets, exports, or scopes beyond one subscription. In both places, cost appears 8 to 24 hours after usage. For the Foundry portal views, see View subscription costs in the Foundry portal (preview).

Important

Features or capabilities marked (preview) in this article are in public preview and are subject to the Supplemental Terms of Use for Microsoft Azure Previews. Previews are provided without a service-level agreement, and Microsoft doesn't recommend them for production workloads. Certain features might not be supported or might have constrained capabilities.

Prerequisites

  • A Foundry resource with usage to analyze.

  • Cost Management Reader or an equivalent billing role at the scope you analyze, such as the subscription or resource group. To create Azure Cost Management budgets or exports, you need Cost Management Contributor or Contributor. For details, see Permissions for cost data and cost controls.

  • A supported Azure account type for Cost Management.

  • To open cost analysis from the Foundry portal, access to the Foundry resource, such as the Foundry User role.

    Important

    The Foundry RBAC roles were recently renamed. Foundry User, Foundry Owner, Foundry Account Owner, and Foundry Project Manager were previously named Azure AI User, Azure AI Owner, Azure AI Account Owner, and Azure AI Project Manager. You might still see the previous names in some places while the rename rolls out. The role IDs and core permissions are unchanged by the rename.

Open cost analysis for your Foundry resource

Cost analysis shows the costs for the scope that you open it from. Choose the scope that includes the costs you want to analyze.

Scope What it shows
Foundry resource Meters for Azure OpenAI models and Microsoft models, which are displayed under each Foundry resource. Claude consumption billed in Claude Consumption Units (CCU) is also associated with the Foundry resource. Use this scope to filter by the project tag.
Resource group Costs for the resources in the resource group. The cost meters for Foundry Models from partners and community display under the resource group instead of the Foundry resource.
Subscription Resources that are spread across multiple resource groups. This scope includes many other resources, so add filters to focus on Foundry usage.

You can open cost analysis from the Foundry portal or directly in the Azure portal.

Open from the Foundry portal

  1. Sign in to Microsoft Foundry. Make sure the New Foundry toggle is on. These steps refer to Foundry (new).
  2. Select Manage in the upper-right navigation.
  3. Select Resource details.
  4. Under View resource, select Manage this resource in the Azure portal.
  5. In the Azure portal, select Resource Management > Cost analysis in the left navigation.

Open from the Azure portal

  1. Sign in to the Azure portal.
  2. Go to your Foundry resource or to its resource group.
  3. Select Cost analysis. For a Foundry resource, it's under Resource Management. For a resource group, it's under Cost Management.

Break down costs by meter and resource

Group costs by meter to see which models and services generate charges. Then view costs by resource to see which resources generate each meter.

Group costs by meter

  1. Open cost analysis for the resource group where you deployed the Foundry resource. This scope also shows the cost meters for Models from partners and community, which display under the resource group instead of the Foundry resource.

  2. Set Group by to Meter. The table shows the cost of each billing meter, so you can see which model series generate costs.

    Screenshot of cost analysis for a resource group with Group by set to Meter, listing the cost of each billing meter, such as GPT-4o token meters and Azure Marketplace model meters.

  3. To filter by model deployment, select Add filter.

  4. Select Tag.

  5. Select the deployment tag.

  6. Select one or more deployments.

    Screenshot of cost analysis for a resource group with a tag filter set to the deployment tag, listing model deployment names to filter by.

You can also filter by agent: select the agent.name tag instead. Foundry adds this tag when an agent calls a model, starting in November 2026. You can filter by user-defined tags in the same way.

View costs by resource

  1. In cost analysis, select View > Cost by resource. The table lists the resources that generate each billing meter.

    Screenshot of the View menu in cost analysis for a resource group, with the Cost by resource view highlighted and Resources listed under Smart views.

  2. Expand a resource to see its cost by meter.

  3. To see how each meter changes over time, set Group by to Meter.

  4. Change the chart type to Line.

  5. To see the cost of one Foundry resource split across its model deployments, select View > Resources under Smart views.

  6. Expand the Foundry resource.

    Screenshot of the Resources view in cost analysis with a Foundry resource expanded to show its cost by service and meter, such as GPT-4.1 input and output token meters.

Understand where model costs appear

Where a model's charges appear depends on how the model is billed. For more information about each offer, see How Foundry Models are billed.

Model offer Billed through Where its meters appear
Foundry Models sold by Azure, including Azure OpenAI Microsoft, directly Azure OpenAI models and Microsoft models appear as meters under each Foundry resource. You typically see meters that account for model input and output usage.
Models from partners and community Azure Marketplace The cost meters display under the resource group instead of the Foundry resource. Some providers' models appear as meters under Global resources.
Claude models billed in CCU Azure Marketplace CCU-billed Claude consumption is associated with the Foundry resource. For details, see Analyze Claude costs.

The following example shows the meters for models sold by Azure, with costs grouped by meter.

Screenshot of cost analysis grouped by meter for a resource group, highlighting Azure OpenAI token meters with the publisher type Microsoft and the service name Cognitive Services.

Some providers' models appear as meters under Global resources, but CCU-billed Claude consumption doesn't. The word Global isn't related to the SKU of the model deployment, for instance, Global Standard.

If you have multiple Foundry resources, your bill contains one entry for each model for each Foundry resource. The resource meters use the format model-name-GUID, where the GUID is an identifier associated with a given Foundry resource. You see billing meters that account for inputs and outputs for each model you consumed.

Screenshot of cost analysis grouped by resource for a resource group, highlighting Azure Marketplace model resources of type SaaS in the global location, next to a Foundry resource with Azure OpenAI token meters.

Allocate costs by project (preview)

For supported models, Foundry supports chargeback at the project level. FinOps teams and admins can allocate shared Foundry spend back to the business unit, team, or workload that incurred it. Use project-level attribution when multiple projects share the same Foundry resource and you need to split the bill accurately.

Foundry automatically applies a project tag to each project's usage of supported models, so you don't need to add tags manually. Project-level attribution has these limitations:

  • It supports models sold by Azure, including Azure OpenAI.
  • It doesn't support models served through Azure Marketplace.
  • It doesn't support Claude models yet.

To view costs by project:

  1. Open cost analysis for your Foundry resource. For steps, see Open cost analysis for your Foundry resource.

  2. In the filter bar, select Add filter.

  3. Select Tag.

  4. Select the project tag.

  5. Select one or more projects to view their attributed cost for the selected time range.

    Screenshot of cost analysis for a Foundry resource filtered by the project tag, showing accumulated cost over time for one selected project.

Use the results to allocate shared Foundry resource costs to individual projects for chargeback or showback, and to track each project's spend trends over time.

Analyze Claude costs

Claude deployments in Foundry use CCU billing, except deployments that existed before CCU billing, which keep their per-model plan. Cost Management associates CCU-billed Claude consumption with your Foundry resource, so scope cost analysis to the Foundry resource that's associated with your Claude deployments.

To allocate Claude costs, filter or group consumption costs by the Azure tags assigned to that resource. Claude supports these resource-level tags for cost allocation, but Foundry project-level cost attribution isn't supported for Claude yet.

To understand input, output, and cache costs, use the model and usage-type breakdowns that Cost Management provides for CCU-billed consumption. If your reports, exports, or chargeback rules use ResourceId, review them so they account for consumption charges that appear under the Foundry resource.

Cost Management shows Claude charges at the prices you pay, including any private-offer pricing. Cost for Claude deployments that use per-model token billing is also available in Cost Management. To learn how token usage converts to CCU, see Claude consumption units (CCU) billing.

Reconcile provisioned throughput reservations

If you use provisioned throughput with Azure Reservations, reconcile reservation costs with your provisioned deployments. You pay hourly for each deployed provisioned throughput unit (PTU).

A reservation applies a discounted rate to the PTUs it covers. You pay the hourly rate for deployed PTUs that go beyond the reservation quantity. The cost pages in the Foundry portal don't show reserved costs, so use Cost Management for these costs.

Task Article
See what percentage of your reserved PTUs are in use across your deployments. View Azure reservation utilization
Understand the amortized cost of your reservations for clearer per-deployment billing visibility. View amortized benefit costs
Distribute reservation costs across teams or projects for internal cost attribution. Charge back Azure Reservation costs

To check that your deployments are covered, and to learn how reservations match deployments, see Provisioned throughput billing and cost management.

Find Azure OpenAI usage across a subscription

At subscription scope, cost analysis includes many resources that aren't related to Foundry. Use the full cost analysis tool in Cost Management, and filter by service tier to focus on Azure OpenAI usage.

  1. Sign in to the Azure portal.

  2. In the top search bar, search for Cost Management.

  3. Select Cost Management to open the full service experience. It includes more options, such as creating budgets.

  4. If Scope doesn't point to the subscription or resource group that you want to analyze, select change.

  5. Select the subscription or resource group.

  6. Select Reporting + analytics > Cost analysis.

  7. On the All views tab, select Accumulated costs.

    Screenshot of Cost Management with Reporting + analytics > Cost analysis selected, the All views tab open, and the Accumulated costs view highlighted.

    The dashboard shows the accumulated costs for the selected scope.

    Screenshot of the Accumulated costs view scoped to a subscription, with an accumulated cost chart and cost breakdowns by service name, location, and resource group.

  8. Select Add filter.

  9. Select Service tier.

  10. Select Azure OpenAI.

    Screenshot of the Accumulated costs view for a subscription filtered by Service tier set to Azure OpenAI, showing only Cognitive Services costs.

The Service tier filter isolates Azure OpenAI usage only. Azure OpenAI isn't listed when you filter by service, because its usage appears under the broader Cognitive Services service classification.

To analyze costs for other models, such as Models from partners and community, scope cost analysis to your Foundry resource's resource group.

Set budgets and alerts in Cost Management

Azure Cost Management budgets track your spending against an amount you set and send alerts when costs approach or exceed thresholds. They alert on billed cost and can't block requests. To alert on or block Foundry usage as requests run, use cost controls (preview) in the Foundry portal.

To create budgets and alerts, see Tutorial: Create and manage budgets and Monitor usage and spending with cost alerts. As you set them up:

  • Create budgets and alerts for your subscriptions and resource groups as part of an overall cost monitoring strategy.
  • For more granular monitoring, create budgets with filters for specific resources or services, so each budget tracks only those costs. For filter options, see Group and filter options.
  • To take more advanced actions, start automation from action groups in your budget notifications. This approach requires your own custom development.

Export cost data

Export your cost data to a storage account when you or others need to analyze it further. For example, finance teams can analyze exported data in Excel or Power BI. Exports are the recommended way to retrieve cost datasets.

You can schedule exports daily, weekly, or monthly, and set a custom date range. For steps, see Tutorial: Create and manage Cost Management exports.

Troubleshoot

  • You can't see cost data: Confirm that you have cost visibility permissions, such as Cost Management Reader, and Foundry access permissions at the correct scope.
  • Partner model costs are missing from your Foundry resource: Open cost analysis for the resource group instead. Except for CCU-billed Claude consumption, the cost meters for Models from partners and community display under the resource group instead of the Foundry resource.
  • Azure OpenAI isn't in the service filter list: Filter by Service tier instead. Azure OpenAI usage appears under the Cognitive Services service classification.
  • You see unexpected meter charges: Group by Meter and Resource to identify which service generated the charge. Then compare it with your deployment and traffic patterns.
  • Tag filters return incomplete results: Verify that required tags are applied to all participating resources and inherited consistently from your deployment process. The project tag covers only Models sold by Azure.
  • Cost doesn't match the Foundry portal: Both show the same Cost Management data. Check that the scope, filters, and date range match. The Foundry portal assigns cost to projects, model deployments, and agents through the project, deployment, and agent.name tags. Cost controls (preview) keep their own running total, which isn't reconciled with your bill during public preview. For details, see Understand the cost data.
  • Azure Cost Management budget alerts are noisy or delayed: Recalibrate alert thresholds after you observe normal usage for a full trend window. Then separate warning and critical thresholds.
  • Scope or policy changes affect cost visibility: Confirm that your selected scope and policy assignments still include all the workload's resources.
  • Data appears delayed after test runs: Wait for ingestion, and then recheck the same time window before you conclude that there's a billing discrepancy.