Note
Access to this page requires authorization. You can try signing in or changing directories.
Access to this page requires authorization. You can try changing directories.
To estimate Microsoft Foundry costs, add each Azure service that your solution uses to the Azure pricing calculator. Then validate the estimate with a small test workload. Before you roll out to production, reconcile it with the billed costs in Azure Cost Management.
Use this workflow to set a realistic spending baseline and avoid billing surprises.
Prerequisites
An Azure subscription for a small test workload.
Cost Management Reader at the subscription or resource group scope of the test workload, to view its costs. For other roles, see Permissions for cost data and cost controls.
A supported Azure account type for Cost Management.
The models and features that you plan to use are available in your target regions. For details, see Feature availability across cloud regions.
To deploy and call the test workload, roles that let you create Foundry resources, manage models, and build and develop in a project. Foundry Owner includes all three. You can also combine Foundry Account Owner or Contributor with Foundry User. For details, see Role-based access control for Microsoft Foundry.
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.
Estimate costs with the Azure pricing calculator
Foundry doesn't have a dedicated page in the Azure pricing calculator, because Foundry is composed of several optional Azure services. Add each service that your solution uses to one estimate. The following steps start with model usage.
- Go to the Azure pricing calculator.
- Search for the product that serves your model, such as Azure OpenAI.
- Select the product to add it to your estimate.
- Choose the region where you plan to deploy.
- Choose the model that you plan to use.
- Choose the deployment type, such as Global Standard.
- Enter your expected monthly input tokens.
- Enter your expected monthly output tokens.
- Review the estimated monthly cost that the calculator shows for the product.
- Add each other product that your solution uses, such as Azure Speech in Foundry Tools, Azure Language in Foundry Tools, or Foundry IQ (Azure AI Search).
- Review the total estimated monthly cost for your solution.
- Save or export the estimate so that you can compare it with billed costs later.
Your estimate shows an estimated monthly cost for each product and a total for your solution.
Model usage is typically billed per token, and the billing unit and rate can vary by model, deployment type, and meter. For how each billing model works, see Billing concepts for Microsoft Foundry.
Note the scope that each part of your estimate covers: subscription, resource group, or resource. Use the same scope when you compare your estimate with billed costs. As you add resources to your project, return to the calculator and update your estimate.
Validate your estimate with a test workload
Before you roll out to production, validate your estimate against real billing meters. This lightweight workflow helps you map each estimate assumption to observed charges.
- Deploy a small test workload that uses the services in your estimate. To deploy a model, see Deploy Microsoft Foundry Models in the Foundry portal.
- Generate representative traffic for the test workload.
- In Azure Cost Management, open cost analysis at the same scope as your estimate.
- Wait until cost analysis shows charges for your test resources. Cost and usage records can appear with a delay, depending on service ingestion timing. For typical timing, see Cost and usage data updates and retention.
- Group costs by Resource and then by Meter. For steps, see Break down costs by meter and resource.
- Compare the meter charges with your estimate assumptions.
- Adjust your spending baseline to match what you observe.
When you finish, you can map each major estimate assumption to one or more observed billing meters. You can also explain any material variance before production rollout.
While the test runs, you can watch its token usage and runs in the Foundry portal within minutes. For standard deployments, its cost appears on the project's Cost tab, by agent and model deployment, 8 to 24 hours after usage. For details, see View project, agent, and model costs (preview). For the meter-level comparison with your estimate, use Cost Management.
Reconcile estimates with billed costs
Use this checklist after each test cycle:
- Confirm that the scope you evaluate (subscription, resource group, or resource) matches the scope of your estimate.
- View or export meter-level charges for the same date range as your test traffic. Compare trends over that window instead of minute-by-minute values.
- Verify that required tags are present and consistently applied to participating resources.
- Compare your estimate assumptions with the observed meters, and record the variance by service.
- Update your Azure Cost Management budgets and alert thresholds only after you validate the trend over at least one full billing cycle.
Azure Cost Management budgets alert on cost and can't block requests. To alert on or block usage as requests run, set a budget in a Foundry cost control.
If your costs don't match your estimate, check these common causes:
- Dependent resources are missing from your scope: Include all dependent resources, such as storage, networking, and Azure Marketplace resources, in your Cost Management scope.
- Partner model meters aren't under the Foundry resource: The cost meters for Foundry Models from partners and community display under the resource group instead of the Foundry resource. The exception is Claude consumption billed in Claude Consumption Units (CCU), which is associated with the Foundry resource. Scope cost analysis to the resource group.
- Monitoring costs appear elsewhere: Azure Monitor Logs and alerting costs appear under those services and at the subscription level, not under your Foundry resource.
- You deployed in different regions: Validate region and model availability before deployment, and recheck your assumptions if you deploy in different regions.
- Data is still arriving: Wait for ingestion, and then recheck the same time window before you conclude that there's a billing discrepancy.
Before production rollout, ensure that operations and finance users have the built-in or custom roles they need to see costs. For details, see Permissions for cost data and cost controls.
Clean up resources
When you finish testing, delete the test deployments that you no longer need. Provisioned throughput deployments, and fine-tuned models with an hourly hosting fee, accrue charges until you delete them.
- Sign in to Microsoft Foundry. Make sure the New Foundry toggle is on. These steps refer to Foundry (new).
- Select Build in the upper-right navigation.
- Select Models in the left pane.
- Select the test deployment to open its details page.
- Delete the deployment.
If you created other resources only for this test, such as a Foundry IQ (Azure AI Search) service, delete them too.
Related content
- To understand how each charge accrues, see Billing concepts for Microsoft Foundry.
- To analyze billed costs by meter, resource, and project, see Analyze Foundry costs in Azure Cost Management.
- To alert on or block usage as requests run with a cost control budget, see Cost controls overview (preview).