Cannot start my VM

Question

Cannot start an existing, previously working GPU VM due to persistent

zonal allocation failure.

Region/zone : West US 2, Availability Zone 3

VM size : Standard_NC144ds_xl_RTXPRO6000BSE_v6

Error : Allocation failed. OverconstrainedZonalAllocationRequest

Timeline: the VM was deallocated normally and has failed every start

attempt since 06:24 UTC+1 on 2026-08-04 - more than 50 consecutive

attempts over 5+ hours, retried automatically every 5 minutes.

Troubleshooting already performed on our side:

• Disabled Accelerated Networking on the NIC (the error message then dropped the "Networking Constraints" line, so that constraint is cleared).
• Confirmed the VM has NO proximity placement group, NO availability set, NO dedicated host / host group, NO capacity reservation group, and IPv4-only (no IPv6 configuration).
• Confirmed via az vm list-skus that this SKU is offered ONLY in zone 3 of West US 2, so relocating to zone 1 or 2 is not possible.
• Confirmed there are no subscription-level restrictions on the SKU.
• Quota note: our ncdsrtxpro6000bsev6xl family vCPU quota is 144/144, fully consumed by this single VM.

What we are asking:

1. Is there a known capacity shortage for NC_RTXPRO6000BSE_v6 in West US 2 zone 3, and what is the expected time to availability?
2. Can the capacity team flag our subscription for allocation when capacity frees up?
3. Would an On-demand Capacity Reservation be grantable for this SKU in this region/zone, and if so for which sizes?
4. Region expansion plans for NCv6 beyond West US 2 and Southeast Asia (we require a US region).

Answer 1

Hello @Jose Macedo,

VM Size = NC144ds_xl_RTXPRO6000BSE_v6 - 8x RTX PRO 6000 Blackwell 96GB, 144 vCPUs. This family went GA in June 2026 only in West US 2 and Southeast Asia. Single-cluster offering = extremely limited supply.

1. Is there a known shortage in West US 2 Zone 3? Yes, by design. NCv6 Blackwell is a new, supply-constrained GPU. Microsoft lists only 2 GA regions as of GA announcement. West US 2 Zone 3 is the only zone in the US where list-skus returns it for you. When that one cluster fills, every customer trying to Start/Resize/Create that SKU in that zone gets AllocationFailed. There is no public ETA, Capacity frees only when someone else in that cluster stops/deletes.

2. Can the capacity team flag your subscription? Azure does not have a waitlist / flag for best-effort allocation. The platform does not queue your request. You need to open a Support Request > Service and subscription limits (quotas) > Virtual Machine > Capacity Assistance or Technical > VM Allocation Failure.

Include: Subscription ID, VM name, region/zone, SKU, exact error code OverconstrainedZonalAllocationRequest, timeline since 2026-08-04 06:24 UTC+1, 50+ retries, business impact, and that this is a stopped existing production VM, not a new deployment.

Support can check internal cluster utilization and in some cases do a live capacity placement assist, but they cannot guarantee time.

3. Would an On-Demand Capacity Reservation be grantable? This is your only durable fix. Yes, ODR is supported for N-series in West US 2, but you have 2 blockers right now:

• Your quota is 144/144 consumed by the stopped VM. ODR consumes quota separately. You need to request a quota increase to e.g. 288 for Standard NCsRTXPRO6000BSEv6 Family vCPUs in West US 2 before you create the reservation.
• The reservation creation itself will also fail if there is no capacity at that moment - same as VM start. So you will need to attempt it in a loop or via Support.

Steps to try:

Bash

# 1. Create reservation group in zone 3az capacity reservation group create -g <rg> -n ncv6-wus2-z3 --location westus2 -z 3
# 2. Try to create reservation for the same SKUaz capacity reservation create -g <rg> --capacity-reservation-group ncv6-wus2-z3 \  -n ncv6-144 --sku Standard_NC144ds_xl_RTXPRO6000BSE_v6 --capacity 1
# If that fails with allocation, try smaller slices - sometimes 24/48/96 is grantable when 144 is not:# Standard_NC24ads_A10_v4 style does not apply - for this family sizes are:# Standard_NC24ds_v6, NC48ds_v6, NC96ds_v6, NC144ds_xl_RTXPRO6000BSE_v6

If reservation succeeds, associate your VM: Stop (deallocate) VM, set capacityReservationGroup property to the group ID, Start.

For future Blackwell VMs, best practice: Never Deallocate. Use az vm stop (not deallocate) to keep allocation, or keep an ODR. Deallocated = no hardware held.

4. Region expansion for NCv6 beyond West US 2 and SE Asia? No public roadmap dates yet. GA announcement is only West US 2 + Southeast Asia. Historically N-series expands to East US, East US 2, West US 3, North Central US in 1-2 quarters after initial GA. If you have a Microsoft account team, ask them to add you to the NCv6 expansion private preview waitlist and to file a Region Expansion customer impact statement - that is what influences prioritization.

---

If this answers your query, do click Accept Answer and Up-Vote for the same. And, if you have any further query do let us know.

Answer 2

Hi @Jose Macedo ,

it looks like a lot of Azure regions having capacity issues these days. Especially with AI/KI related workloads and compute resources.

As far as I know:

There is no detailed plan with expected time for increased capacity in the Azure regions.

Regarding the flag for your subscription the best option might be to get in contact with your Microsoft account teams for quota escalation .

The Standard_NC144ds_xl_RTXPRO6000BSE_v6 isn't listed for On-demand Capacity Reservation -> https://learn.microsoft.com/en-us/azure/virtual-machines/capacity-reservation-overview

---

(If the reply was helpful please don't forget to upvote and/or accept as answer, thank you)

Regards

Andreas Baumgarten