Cloud capacity is no longer a given, and prices are starting to rise

We couldn’t create a Postgres server in East US

Recently, we tried to provision a PostgreSQL server in East US, the region where most of our other workloads run. We weren’t able to.

It wasn’t something a quota bump could fix. The region simply didn’t have the capacity.

We reached out to Microsoft support. Their answer wasn’t very promising: the region was experiencing capacity constraints, and our request would go to their capacity team for review.

We were able to get around it by moving to another VM family in another region.

But that isn’t free. It adds latency, egress cost and operational overhead we didn’t plan for.

Capacity used to be someone else’s headache

For years, those of us who operate in the cloud have assumed capacity is infinite. You don’t have to worry about it. It is someone else’s headache.

At most, you might run out of quota in your VPC or subscription. More often than not, an email, a ticket or a support case got the quota bumped.

Underneath those virtual resources are physical servers. When I worked on bare-metal capacity in the public cloud, new capacity took roughly 18 months to bring online.

We are not alone

If you have been operating in the cloud recently, chances are you have run out of capacity too, usually CPUs in a given region.

Several others have hit the same wall: a PostgreSQL upgrade blocked and a scale-up rejected in East US, VM shortages in North Europe, a “full” UK South, and similar reports on AWS and Google Cloud.

Memory and storage moved first

Some time back, I wrote a post about rising prices, primarily for memory and storage, and followed up on LinkedIn. In fact, in June Apple raised prices on Macs and iPads, saying it had “never seen a component price increase this much, this quickly.” The 1 TB MacBook Pro went from $1,699 to $1,999.

It shows up in hardware buying as well. One IT admin couldn’t find enterprise SSDs even with four times the previous year’s budget.

It comes down to supply and demand. AI has made processing and generation simple. Anyone can now produce code, content and analysis in seconds, so we need more processing and more storage than ever before.

AI has eaten up a lot of the supply of chips, memory and storage, and it takes years to build new supply.

When demand grows faster than supply, prices go up. And sometimes, as we found in East US, there is nothing available at all.

Cloud prices, which only went down, are now going up

Microsoft has announced that in specific regions, prices for selected VMs and storage will increase by 8% to 17%, starting February 1, 2027.

This is significant. For several years we have seen cloud costs go down, or stay the same while we got more capacity and more processing power.

×

Microsoft attributes this to regional market conditions, and it has adjusted regional prices before. But given the capacity shortages we are seeing, it wouldn’t be surprising if this extends to other regions and other clouds.

In fact, AWS has already increased its GPU capacity pricing twice in 2026.

Most storage is bigger than it needs to be

You can’t control cloud prices or regional capacity. You can control how much you provision compared with what you actually use.

Disks are usually sized for a peak that rarely comes. The gap between provisioned and used is paid for every month, and when prices go up, that gap costs more.

And is it on the right tier?

The other question is where the data lives. Block storage on premium disks that is barely touched. Blob data in the hot tier that is read once a quarter. Each is a small decision that adds up across thousands of resources.

That is the part we work on at Lucidity: keeping provisioned storage in line with real usage, and on the right tier.

What will it cost you?

Your bill can go up even if your usage doesn’t. Try it with your own numbers. Real bills mix SKUs, discounts and commitments, so treat this as a rough guide.