On Data Storage Costs

For most of my professional career, I’ve worked in information technology. Though I’ve done quite a bit in the IT world, my primary focus has been infrastructure. That has historically meant servers, storage, backups, email, networks, phone systems, Wi-Fi, Internet, VPNs and firewalls, inter-site connectivity, and other hardware. I spent a lot of time focused on the infrastructure to run workloads—physical servers, virtual machines, and cloud technologies.

Recently I took a job managing a team of IT professionals who install and manage physical security systems. By this, I mean the systems that control whether a door opens by waving a badge or fob, intrusion detection, and video surveillance. These systems have become heavily dependent on the types of IT systems I managed for many years, so it’s a natural fit for me, even if the industry is a bit different than what I’m used to.

A few weeks ago we were discussing a server build for a customer, and the subject of disk storage came up. As you can imagine, storage is a key cost driver for a video surveillance system disk. We have customers with video retention requirements that run from 15 days to 120 days. In some cases the server we sell needs considerable storage space. We keep telling our customers that storage costs are outrageous right now, the highest they’ve ever been. I began wondering: “Are they really? I mean, I know server prices have been climbing lately, but have they really outpaced inflation?” This led me to do a little digging.

The first thing I wanted to know was the cost of a gigabyte of storage in inflation-adjusted dollars. Sometime in the latter end of the 2000s, we saw “SSD” storage arriving on the scene in what I’ll call “enterprise storage.” Enterprise storage is that data storage used by businesses to run their mission-critical workloads and store mission-critical data. So it’s not the same as that hard drive you bought at Costco to store your music or back up your MacBook. This distinction meant we had to look at two different categories. What I found was exactly what I thought I’d find: old-school spinning disks were already pretty cheap but declined over time, while SSDs were expensive by comparison but saw steep cost declines until they arrived at a price point just north of spinning disks (which nobody really uses anymore). This can be seen in the following graph:

Now, this doesn’t tell the whole story. As the footnote in the chart indicates, this is media-only cost. In other words, it’s the cost of a hard drive only. It doesn’t include all the costs attendant to that storage. See, in enterprise storage, the kind that lives in a company’s data center, we combine all of those hard drives into an array of disks. This array has hardware costs of its own, along with a control plane (the hardware that runs it), licensing and software costs, plus loss of raw storage due to redundancy (mirroring data to multiple disks, RAID configuration, etc.).

So this brought me to my next question: “What’s the all-up cost per gigabyte, factoring in a 5-year life cycle?” These arrays have a high purchase price followed by varied annual maintenance costs that tend to increase over time. Though an industry framework for calculating total cost of ownership exists, there isn’t a set of data that averages costs over the 25-year time span I was looking for. Nonetheless, data can be extrapolated from existing models to come up with a cost estimate, which runs as follows:

Notice that the chart looks a little different for spinning disk, but is roughly the same shape for SSD, and in both cases we see the price per gigabyte dropping substantially, to roughly the same by 2025. Again, these are inflation-adjusted dollars. Note, of course, the uptick in 2026, when we see storage costs on the rise. But the cost per gigabyte in inflation-adjusted dollars is still significantly lower than it used to be.

Another thought occurred to me as I analyzed the data and results: we don’t use data the same way we used to. Or, if I were to state it a little differently: we use more data now to do the same things we used to do. Photos are bigger now than they were 10 years ago. A single Excel spreadsheet has grown significantly while doing the same work. Then, of course, there’s data retention in systems. As storage has gotten cheaper, there’s less need to purge data—any particular ERP or accounting system or file server has more data in it than it used to. All because it’s just cheaper. That led me to the idea of “storage inflation.” That notion is that it actually takes more storage to do the same job as it did in the past.

If this is true, and my experience tells me it is, then what does that do to my storage cost chart? Does storage cost the same, more, or less now when factoring in both cost inflation and storage inflation? Well, this requires more assumptions, the primary one being the rate of inflation. The AI tool I used created a simplified storage inflation index: 100 x (GB required for today's workload / GB required for a 2002 workload). Obviously workloads will vary, but it came up with an average storage inflation rate of 20% per year. This seems high when compared to cost inflation, but I think it’s fair.

When you apply that rate to the cost data, something interesting happens: the cost per gigabyte simply goes flat.

While it is true that storage has gotten cheaper over the years, as it has gotten cheaper, the cost savings have been eaten up by...you got it…consumption. We are just fat (when it comes to how we consume storage).

Of course, this doesn’t cover storage costs in public cloud environments like Amazon’s AWS and Microsoft’s Azure. Nor does it analyze the costs that are rolled into “Software as a Service” models of workload delivery. Those environments are naturally affected by these costs, but they are often abstracted away from hardware costs.

Stay tuned next week for Spin’s Analysis of RAM costs…