Azure Machine Learning: V6/V7 VM sizes not available in westus2 for compute instance or cluster

Mandeep, Mandeep 20 Reputation points
2026-06-30T22:50:47.5433333+00:00

I’m trying to create Azure Machine Learning compute resources using newer VM generations (V6/V7), but they are not available in my workspace (westus2).

When creating either a compute instance or a compute cluster in AzureML Studio, I want to use newer VM generations (V6 or V7 series CPUs). According to Azure VM documentation, these newer VM families have been available for some time.

However, in the AzureML Studio VM size dropdown, only up to V5 VM sizes are available.

What is the recommended way to enable or access newer VM generations (V6/V7) for Azure Machine Learning workloads?

Azure Machine Learning
Answer accepted by question author
Divyesh Govaerdhanan 11,725 Reputation points MVP Volunteer Moderator
2026-06-30T23:26:23.4166667+00:00

Hello Mandeep, Mandeep,

Welcome to Microsoft Q&A,

This is not a quota or region availability issue. Azure Machine Learning only supports a fixed allow-list of VM series for compute instances and compute clusters, and that list has not been updated to include V6 or V7 generation sizes yet.

The official supported series table for AML managed compute (compute instance and compute cluster) currently tops out at Dv3/DSv3 for general-purpose, Ev3/ESv3 for memory optimized, and NCadsH100_v5 / ND-H100-v5 / ND-H200-v5 for the newest GPU tier. There is no Dv5, Dv6, Dv7, Ev6, Ev7, or any other v6/v7 series in that list, for any region. This is separate from the general Azure VM size catalog. Even though V6/V7 series are GA at the Azure Compute level and show up fine for plain VMs, AML maintains its own allow-list and has to onboard each series before it appears as selectable in Studio, SDK, or CLI for compute instance or compute cluster.

So this is not something westus2 specifically is missing. No region currently exposes V6/V7 for AML managed compute, because the series itself is not yet on AML's supported list. Recreating the workspace in another region will not fix it.

For workloads that need V6/V7 today, use a Remote VM (provision the VM yourself, then attach it to the workspace as a training compute target) or Azure Machine Learning Kubernetes (Arc-enabled AKS) on top of a VM size you provision directly through Azure Compute. Both attach as unmanaged compute and bypass the AML managed-compute allow-list.

Please Upvote and accept the answer if it helps!!

Was this answer helpful?

1 person found this answer helpful.

0 additional answers

Sort by: Most helpful
Most helpful Newest Oldest

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.