I'm seeing persistent allocation failures in Azure Sweden Central. Nearly all Bv1 VM deployments fail, and moving to Bv2 doesn't help. Dv5 instances have now started running into similar capacity problems, which is especially concerning because they're among the last options that support Hyper-V Generation 1. Is this a temporary capacity incident, a surge in demand for the region, or part of a broader Azure capacity shortage? Are other users seeing the same thing?
5 Answers
Yes, Sweden Central appears to be under increasing pressure. It’s popular across the Nordic countries, and it has also become a recommended fallback when larger European regions such as North Europe, UK South, or West Europe can’t provide capacity. That may be concentrating even more workloads there.
There may also be extra pressure from newer AI and model-hosting services. Sweden Central is one of the European regions where some partner model deployments are available, so that could be consuming additional compute. More broadly, Azure appears to have capacity constraints in several European regions, not just Sweden.
Hopefully those services become available in more regions soon, instead of sending even more customers to Sweden Central.
B-series capacity problems have been reported for a couple of weeks, and they can happen globally because those sizes rely on shared or constrained capacity. I’ve even seen a VM fail to start again after being deallocated for only a short time.
That’s the dangerous part: once a VM is deallocated, you may need to resize to a much more expensive SKU just to get it running again. Reservations generally don’t solve an SKU that currently has no allocatable capacity.
The Hyper-V Generation 1 requirement is probably making the situation harder. If possible, migrate older Windows VMs to Generation 2 or Trusted Launch so you can use newer VM families such as Dv6. That gives you more options, although the migration needs careful planning because deallocating the original VM could leave you unable to restart it if capacity disappears.
The Gen 1 dependency is mostly technical debt from older machines and images. Migration should be possible, but it involves manual disk or image work, and there’s a real risk of losing the original allocation during the process.
Capacity reservations and reserved instances are easy to confuse. Reserved instances mainly provide a pricing discount; they don’t guarantee that a VM can be allocated. A capacity reservation can help only when the requested VM size and region are already available, so it won’t fix an SKU that is currently exhausted.
It seems to have gotten noticeably worse only recently, so demand growth or a wider hardware shortage could both be contributing.