AllocationFailed. The requested VM size is currently not available in the selected region due to capacity restrictions.Operation results in exceeding quota limits.
If any of those look familiar, you’re not alone, and it’s not a one off. These messages are turning up more and more often across Azure, and while it’s a global trend, it’s particularly prevalent in Europe at the moment. Nowhere am I seeing it more than in North Europe, my home region and the default choice for so many Irish and EU customers, and as I’ll get into below there are some genuinely local factors that make it a particularly acute example of a much wider trend. This post is my attempt to explain what’s actually happening, why North Europe is such a pressure point right now, and more importantly, what you can practically do about it.
What we’re actually seeing
The symptoms vary depending on what you’re trying to deploy, but the pattern is consistent across customers I’m working with. VM creation or scale-out operations failing with capacity errors, particular VM series being completely unavailable in a region, zone-redundant deployments failing because a specific availability zone doesn’t have stock, and quota requests that would previously have been approved in minutes now being instantly refused, with almost no exceptions made.
Critically, this isn’t just a VM problem. It affects PaaS and SaaS services too, since everything ultimately runs as compute somewhere, whether that VM is one you manage or one Microsoft manages on your behalf under the hood makes no difference. If you’re waiting on quota for something like Microsoft Fabric capacity, don’t assume being a platform managed service gets you out of the queue. The same economics of supply and demand apply either way.
Why is this happening?
There’s a lot going on here so I won’t turn this into an energy policy blog post, but at a high level, this isn’t really a Microsoft specific problem. It’s fundamentally a datacentre power and hardware supply problem, and it’s affecting every major public cloud provider to some degree. AWS and GCP customers are reporting very similar capacity friction in their own European datacentres. The common thread everywhere is a combination of the extraordinary growth in demand for compute driven by AI workloads, and the physical limits on how quickly power and hardware supply chains can keep pace with that demand.
This is where North Europe specifically becomes a really good case study in just how tight things can get, because on top of that global demand pressure, we have a very local, very Irish complication sitting right on top of it. Dublin is now Europe’s second largest datacentre cluster (after London), and that growth has put serious strain on the national energy grid, to the point where the regulator effectively paused new datacentre grid connections in the greater Dublin area back in 2021. However, recent policy changes have effectively reopened the door to new developments. Rather than a blanket restriction, Ireland has moved to a conditional growth model where new datacentres can connect to the grid provided they can demonstrate measures such as onsite generation, energy storage, and a strong commitment to renewable energy. While the regulator has reopened a pathway for growth, this doesn’t suddenly create spare capacity. The combination of planning, power infrastructure, renewable generation and datacentre construction means it will take several years before these policy changes translate into significant new compute capacity on the ground.
Unfortunately, this means that we will need to work within these capacity constraints for a number of years to come and plan accordingly.
Microsoft’s approach to managing these constraints
To manage this, Microsoft are leaning heavily on subscription quotas as the control mechanism, and it’s worth understanding just how far that now extends in North Europe specifically. New customers are currently not being allocated any quota at all in North Europe. If you’re an existing customer, the expectation is that you work within whatever quota you already have allocated, because quota increase requests for North Europe are almost certainly going to be refused while these constraints remain in place. There are exceptions, of course, but unless you are a very large customer or have a really compelling business justification for it, you can expect quota requests to be denied here for the foreseeable future.
In fairness to Microsoft, from what I’ve seen they are trying to honour quota where the underlying capacity genuinely exists to support it, and they’re not just cutting people off arbitrarily.
One useful lever if you’re an EA customer is to make use of Azure Quota Groups, which let you share quota allocation across your subscription estate, effectively moving quota from one subscription to another rather than requesting new quota from Microsoft. It’s a genuinely handy option if you have headroom sitting unused in one subscription and a need in another. The catch is that it only supports IaaS today, so it won’t help you with PaaS or SaaS quota constraints, and it isn’t available at all outside of EA agreements, so if you’re on a CSP or pay-as-you-go subscription this particular tool isn’t in your kit.
A caveat here: quota does not equal capacity. Quotas are Microsoft’s internal mechanism for managing capacity and allocating it to customers. Just because you have quota, it does not mean that capacity has been assigned to your subscription or that you are in any way guaranteed to have that capacity available to you for deployment.
What can customers do to mitigate these constraints?
A few practical things I’d suggest considering, depending on your situation:
- Look at your regional strategy early rather than at the point of deployment. Don’t assume North Europe or West Europe capacity, or quota, will simply be there when you need it. Bake capacity risk into your design conversations from day one, not as an afterthought when a deployment fails.
- Work within your quotas. If you can free up some quota by scaling down some workloads to a lower SKU then this might free up enough compute cores for that extra VM that you need. This does not guarantee capacity availability, but it might get you out of a tight spot.
- If you’re an EA customer, look at Azure Quota Groups before you go anywhere near a quota increase request. Redistributing quota you already own across your subscriptions is a far more reliable path right now than asking Microsoft for more of it in North Europe.
- Use on-demand capacity reservations where it makes sense. If you have a known future need for specific VM sizes, reserving that capacity ahead of time is a more dependable path than hoping it will be there when you go to deploy. I haven’t had any customers actually use this yet as you are committing to paying for capacity before you even deploy. That said, it’s worth knowing this option exists if you really need it. Just note, this also does not fully guarantee capacity availability – that’s why it comes with a service SLA offering service credits where capacity cannot be fulfilled.
- Build flexibility into your architecture where you can. Avoid hard dependencies on a single VM family or a single region wherever possible. The more flexible your workload is about where and on what it can run, the less exposed you are to any one region’s capacity position.
- Multi-region has now become an essential component for every Azure tenant, not simply for DR purposes, but for capacity purposes. If you are already multi-region but using the typical North/West Europe region pair, then you will probably need to add a third region to that strategy.
Consider Sweden Central as your third region (or your second, if you aren’t already multi-region). At the time of writing, Sweden Central is the region Microsoft are recommending as the go-to alternative for EU customers currently using North Europe or West Europe. This may change, but today this is the recommendation. Importantly, and unlike some alternative EU based Azure regions, it has full capacity available across all three availability zones at the moment, which matters a lot if you’re deploying zone-redundant services. - Engage early with your Microsoft account team or partner if you have a large or time sensitive deployment coming up. A conversation well in advance of a project deadline gives Microsoft the best chance of planning capacity for you, rather than you finding out about a constraint at go live.
Key Takeaways
Capacity and quota friction in North Europe isn’t a sign that anything is fundamentally broken with Azure, and it’s certainly not unique to Microsoft. It’s the practical, physical reality of a huge and sudden surge in global compute demand meeting the hard limits of power, cooling and hardware supply, landing especially hard here because of how concentrated our own datacentre footprint already is and on top of a grid that’s been under strain for years. The regulatory and grid picture in Ireland is genuinely moving in the right direction to unlock further growth, but that takes time to land, and in the meantime the quota picture for North Europe is going to stay tight.
My advice is to plan for it rather than be surprised by it. Understand your regional exposure, and if you don’t already have quota in North Europe, assume you won’t get it in the short term rather than waiting to find out. Use Azure Quota Groups if EA gives you that option, and have Sweden Central (or another suitable alternative) baked into your design architecture from the outset, rather than keeping it in your back pocket as a fallback for anything North Europe or West Europe can’t currently support.
azurealan.ie 