Compute limits

Escalate the denied GPU quota request through an Azure support request, including the rejected request details, exact GPU VM family, region, and minimum required vCPU limit.

Escalate the denied request

A quota of 0 prevents deployment, and GPU quota is enforced separately for each VM family, subscription, and region. Azure also checks both the VM-family quota and the total regional vCPU quota.^1^

1. In the Azure portal, search for Quotas.
2. Select Compute, then select the subscription and target region.
3. Locate the required GPU family—for example, Standard NCasT4 v3 Family vCPUs.
4. Open the rejected request or select its support/information icon, then choose Create a support request.
5. For a non-adjustable quota, select Quota type → Other Requests.
6. Include:
   • The exact VM SKU and family required.
   • Target region.
   • Requested vCPU limit. This must be at least the vCPU count of the VM size.
   • The denied quota request details.
   • Expected workload and deployment timeline.
   • Whether another GPU family or region is acceptable.
7. Submit the request for manual review.

This is the documented escalation path when an adjustable request is not fulfilled or the quota is unavailable for self-service adjustment.^2^

Request only the capacity needed for the first deployment rather than a large speculative limit. Approval remains subject to subscription eligibility and available capacity.

Check viable GPU alternatives

While the support request is reviewed, identify a GPU SKU available to the subscription in another acceptable region:

# Replace the region and SKU prefix as needed.
az vm list-skus \
  --location westeurope \
  --resource-type virtualMachines \
  --query "[?starts_with(name, 'Standard_NC')].name" \
  --output table

az vm list-usage \
  --location westeurope \
  --output table

In the usage output, check both:

• Total Regional vCPUs
• The specific NC, ND, or NV family quota

A VM deployment must fit under both limits. Quota and physical capacity are separate, so even approved quota does not guarantee that a particular GPU size is currently deployable in that region or availability zone. If necessary, test another current GPU series, zone, or region.^1^

Do not substitute families without checking workload compatibility:

• NC generally targets GPU compute and AI/ML.
• NV generally targets visualization and graphics.
• ND targets deep-learning workloads.

Check subscription eligibility

If the requested SKU is in the NC family, confirm that the subscription is not an Azure Trial or Visual Studio Subscriber subscription. Microsoft documents that some NC SKUs are unavailable to those subscription types and that the subscription level may prevent selection or deployment.^3^

If that restriction applies, support cannot turn an ineligible offer into an eligible one merely by changing quota. Move the workload to an eligible paid subscription, and then request the family quota there.

If support cannot grant the requested family

Ask support which currently available GPU family and region can be approved for the subscription. Then:

1. Verify that SKU appears in az vm list-skus for that region.
2. Request the minimum family quota required by the selected size.
3. Confirm that the regional vCPU quota is also sufficient.
4. Deploy while explicitly specifying the approved SKU and region.

A reservation should not be purchased as a workaround for a zero quota: quota permission and regional capacity are distinct controls.

References

1. Check vCPU quotas
2. Increase VM-family vCPU quotas
3. 'NC' sub-family GPU accelerated VM size series