This issue happens when I stop/deallocate the VM and later try to start it again.

Question

This issue happens when I stop/deallocate the VM and later try to start it again.

KUPS 0

Hello,

I am using an Azure GPU virtual machine, but recently I have been unable to start it.

Whenever I click the Start button, Azure returns the following error message:

“Allocation failed. We do not have sufficient capacity for the requested VM size in this region.”

This issue happens when I stop/deallocate the VM and later try to start it again.

Because of this issue, I would like to ask whether Azure provides a way to reserve or secure GPU VM capacity in advance for a specific future date or time.

What I want to achieve is the following:

I want to make sure that I can use the GPU VM at a later time without running into this capacity error.
I would like to know whether Azure supports reserving capacity in advance for the same VM size/SKU.
If advance reservation is possible, I would like to know what option or process I should use.
If it is not possible, I would like to know the best recommended alternative.

My questions are:

Is it possible to reserve Azure GPU VM capacity in advance for a future timeslot?
Can I secure the same GPU VM SKU ahead of time so that the VM can be started successfully later?
If advance reservation is not supported in this scenario, what is the recommended workaround? For example, should I use Capacity Reservation, keep the VM allocated instead of deallocated, move to another region, or use a different VM size?

My goal is to avoid repeated allocation failures caused by insufficient regional capacity when starting the VM later.

Thank you.

Himanshu Shekhar 5,225 Reputation points Microsoft External Staff Moderator

2026-03-19T13:15:28.9166667+00:00

KUPS - Thank you for the details. This is a regional capacity limitation for the requested GPU VM size. When the VM is stopped and deallocated, its underlying hardware is released, and Azure attempts a fresh allocation when you start it again.

If sufficient GPU capacity is not available in the region at that time, the start operation can fail with this error. This behavior is expected for high‑demand GPU SKUs and is not caused by any issue with your VM configuration.
Himanshu Shekhar 5,225 Reputation points Microsoft External Staff Moderator

2026-03-19T13:33:10.96+00:00
Is it possible to reserve Azure GPU VM capacity in advance for a future timeslot?

Azure does not support reserving GPU VM capacity for a specific future date or time window. There is no time‑based or scheduled reservation mechanism for VM allocation.

Can I secure the same GPU VM SKU ahead of time so the VM can be started successfully later?

Yes. Azure supports On‑Demand Capacity Reservations, which allow you to reserve physical compute capacity for a specific VM size and region. Once the reservation is successfully created, that capacity is held exclusively for your subscription and can be used at any time while the reservation exists.

This is the only supported way to guarantee GPU capacity availability and prevent allocation failures caused by regional shortages. [learn.microsoft.com]

Important clarification: Azure Reserved VM Instances (1‑year / 3‑year) provide cost savings only and do not guarantee capacity availability, so they will not prevent this issue. [azure.microsoft.com]

If advance reservation is not used, what is the recommended workaround? If capacity reservation is not an option, the recommended alternatives are:

Keep the VM allocated instead of deallocated Stopping the VM without deallocating it retains the GPU hardware and avoids reallocation.

Use Azure On‑Demand Capacity Reservation (recommended for critical workloads) This provides guaranteed capacity for the selected GPU SKU and region. [learn.microsoft.com]

Be flexible with region or GPU VM size Try a different region, availability zone (if supported), or an alternative GPU SKU with better availability.

Additionally reached for details via private messages.