Need Help Escalating Unexpected Azure Databricks Serverless Charges for Learning Account

Hello Nirmal v,

Greetings! Thanks for raising this question in the Q&A forum.

What likely happened is that a Model Serving endpoint in the Serving tab of your Databricks workspace was left in a Ready state. Serverless real time inference endpoints keep at least one provisioned or scale to zero enabled compute unit warm in the background once created, and they continue to accrue Serverless Real Time Inference DBU charges as long as the endpoint exists in Ready state, even if you are not actively sending requests to it. Deleting the notebook or closing the tutorial does not remove the endpoint itself, it has to be explicitly stopped or deleted from the Serving tab. This is why charges kept accumulating after you stopped actively using the workspace.

Check for any remaining Model Serving endpoints. In your Azure Databricks workspace, go to Machine Learning in the left sidebar, select Serving, and look at every endpoint listed. Any endpoint showing Ready is live and billing. Select it and choose Delete to remove it completely, or Stop if you only want to pause it. If you cannot see or manage an endpoint that belongs to a colleague or another identity, you will need workspace admin permissions to remove it.

Check for other running compute. Go to Compute in the sidebar and review all-purpose clusters, job clusters, and SQL warehouses. Terminate anything still running. Also check Workflows for any scheduled jobs that might spin up compute automatically on a recurring basis.

Confirm no further accrual using Cost Management. In the Azure portal, go to Cost Management + Billing, open Cost analysis for the subscription, set the group by option to Meter or Resource, and filter to Azure Databricks. This lets you see day by day whether the Serverless Real Time Inference charges have actually stopped since you deleted the endpoints, rather than relying on the invoice which lags by a day or more.

Set a budget and cost alert going forward. Still in Cost Management + Billing, go to Budgets, create a budget for this subscription with a low threshold, for example €20, and set an alert so you get notified by email before charges reach a level like this again.

Open a dedicated Billing support request rather than continuing through general chat support. Go to Help + support in the Azure portal, select Create a support request, choose Issue type Billing, and select the affected subscription. Billing support requests are free regardless of your support plan, including on a free or basic account, so this is the correct channel and there is no cost barrier to using it. In the problem description, include both invoice amounts you mentioned (the approximately €140 already invoiced and the approximately €240 upcoming charge), state clearly that you are an individual learner, that the charges came from Serverless Real Time Inference endpoints left running unintentionally while following tutorials, and that you are requesting a one time goodwill credit or partial waiver given the circumstances. Whether a goodwill adjustment is granted is entirely at Microsoft's discretion and is decided case by case by the billing team, there is no guaranteed outcome, but this is the correct and only channel for that decision to be made.

If you keep getting redirected between agents without resolution, explicitly ask in the same support ticket thread for the case to be escalated to a supervisor or a senior billing engineer, and reference your ticket number in that request. There is no separately named public facing Azure Billing Escalation team you can contact directly, escalation happens by requesting it within your existing ticket rather than opening a new one, since starting over tends to reset context and cause the same redirection loop.

If this answer helps you kindly accept the answer which will help others who have similar questions.

Best Regards,

Jerald Felix.