No warning for deployment of Provisioned Models leading to unexpected Billing charges.

Delete the provisioned deployment in Microsoft Foundry immediately to stop its hourly billing, then update your existing Azure billing support request to ask for a discretionary credit or refund.

Stop further charges

Hourly billing starts when a provisioned deployment is created and stops when it is deleted.^1^

Deletion interrupts any application using the deployment. Before proceeding, confirm it is not serving production traffic; record its deployment name, model, deployment type, PTUs, resource, region, and creation time for the billing case.

1. Open the Microsoft Foundry portal.
2. Navigate to the affected resource, then open Deployments.
3. Select the provisioned deployment.
4. Select Delete, and confirm.
5. Refresh Deployments and verify that it no longer appears.
6. Record the deletion time and retain a screenshot or confirmation for support.

The documented CLI alternative is:^2^

az cognitiveservices account deployment delete \
  --deployment-name <deployment-name> \
  --name <azure-openai-resource-name> \
  --resource-group <resource-group-name>

Do not merely delete the parent Azure resource first. Charges for deployments on a deleted resource can continue until that resource is purged; deployments should be deleted before the resource.^1^

If the parent resource was already deleted without first deleting the deployment, follow the documented recovery/purge process referenced by that page. Purging is permanent and removes recoverability, so do it only after confirming that the resource and dependent configuration are no longer needed.

Check for a reservation or commitment

Deleting the deployment does not cancel or modify an Azure Reservation. In Azure portal → Reservations, check whether a Provisioned Throughput reservation was purchased. If one exists, review its cancellation or exchange options; fees might apply.^1^

If this is an older commitment-based provisioned deployment rather than the current hourly/reservation model:

1. Set its renewal policy to Don’t autorenew.
2. Delete the provisioned deployment.

Both actions are required to discontinue provisioned throughput under that model; leaving the deployment after commitment expiry can generate hourly overage charges.^3^

Update the existing billing case

Only Azure Billing Support can review and approve a reversal; deletion prevents new deployment usage charges but does not automatically reverse charges already incurred. Billing and subscription support is available without requiring a technical support plan.^4^

Go to:

1. Azure portal → Help + support → All support requests.
2. Open your existing request.
3. Add a message requesting review of the accidental provisioned deployment charges.
4. Include:
   • Subscription and affected resource name.
   • Deployment name, deployment type, region, and PTU count.
   • Creation and deletion timestamps, including time zone.
   • Screenshot of the deployment or activity record.
   • Cost Analysis export or screenshot showing the relevant charges.
   • Confirmation that this was a first-time experimental deployment created unintentionally.
   • A request for a one-time discretionary billing credit/refund.
   • A note that the deployment has now been deleted and whether any reservation or commitment exists.

Do not post the support-request number or subscription details publicly.

To document the amount, use Azure portal → Cost Management + Billing → Cost Management → Cost Analysis. Set daily granularity, filter by the relevant service and meter, then group by Resource; Cost Analysis supports service and meter filters for isolating the resource producing the charges.^5^

Microsoft documents that the current deployment flow includes Confirm Pricing, which displays the hourly list price before deployment.^6^ If that confirmation was absent, unclear, or did not display pricing in your session, state that explicitly and attach any screenshots. A refund is not automatic or guaranteed; the billing team must assess the request.

References

1. Operate provisioned deployments in production
2. Quickstart: Create a provisioned throughput deployment
3. Azure OpenAI provisioned Managed offering updates
4. Create an Azure support request
5. Common cost analysis uses
6. Get started using provisioned deployments on the Azure OpenAI in Azure AI Foundry Models