AI-300 Practice Question 37: Managed Compute Quota Billing
Microsoft Machine Learning Operations Engineer Associate (AI-300) · Test 4 · question 7 of 10
A team wants to deploy an open-source model from the Hugging Face collection in Microsoft Foundry by using managed compute (preview). The subscription already has plenty of unused Azure virtual machine quota. The platform team asks how capacity is obtained and how the deployment will be billed, and wants billing to stop overnight when nobody uses the model. Which statement is correct?
Show answer and explanation
Answer:
The Foundry deployment overview describes managed compute (preview) as a managed GPU platform for open-source and custom models. It says that *"managed compute billing is hourly per accelerator SKU"* and that *"quota is granted per accelerator SKU per region through the Foundry quota process and is separate from Azure VM quota ... Existing Azure VM quota can't be applied to a managed compute deployment."* Auto-scale and scale-to-zero *"align cost with actual traffic so that billing stops immediately when instances scale down"*, and you can configure an idle timeout so the deployment scales to zero when no traffic arrives (option C). Managed compute is currently in public preview.
- A is wrong because VM quota cannot be used for managed compute, and its billing basis is the accelerator hour, not tokens. - B is wrong because serverless API deployments serve Foundry Models, while open-source models from the Hugging Face collection use managed compute. - D is wrong because provisioned throughput units belong to serverless API deployments of Foundry Models. They are reserved capacity that is billed whether or not requests arrive.