Cloud Storage Pricing: Why the Bill Never Matches the Sticker Price

Cloud storage looks like a commodity and is priced like a puzzle. Two services advertising "unlimited storage" can differ by a factor of five in what you actually pay, because they are counting different things: seats, terabytes, active users, or API calls.

The point of this piece is not to rank providers. It is to give you the arithmetic to work out which pricing model punishes your particular shape of usage, before you move several hundred gigabytes and discover it.

The three pricing models

Almost every cloud storage product uses one of these, sometimes in combination.

Per seat. You pay for each user, and storage is pooled or generous. This is the model sold to companies, because companies think in headcount.

Per terabyte. You pay for capacity, and users are free or cheap. This is the model sold to individuals and to teams who store a lot and log in rarely.

Per workspace or per shared drive. A flat fee for a container with a cap. This is the model that looks cheapest at the start and becomes the most expensive at scale, because the next tier up is a big jump.

The mismatch happens at the edges. A team of three people storing eight terabytes is badly served by per-seat pricing. A team of forty people storing four hundred gigabytes is badly served by per-terabyte pricing. Both are common.

The arithmetic worth doing

Take your actual numbers — number of people who need access, and total bytes stored — and run them through both models.

An illustrative example. Suppose a per-seat plan costs the equivalent of 10 units per user per month and includes 2TB pooled. A per-terabyte plan costs 8 units per terabyte per month with unlimited users.

The numbers are made up, but the shape is not. The crossover point is roughly "how many people per terabyte do you have." Above about five people per terabyte, per-seat wins. Below one, per-terabyte wins by an order of magnitude.

Most organisations have no idea which side of that line they are on, because nobody ever adds up the two numbers. It takes ten minutes and it is usually the largest single saving available in a software budget.

The costs that are not on the pricing page

Four things consistently show up later.

Egress

Some providers charge for data leaving the platform, and the rate is not always obvious because normal use never triggers it. Migration does. So does serving files to a website, or syncing to a second provider.

Before you commit to a large archive, find out the egress rate and multiply it by your total size. That number is the real cost of ever leaving.

Versioning and trash

Version history and deleted-file retention are features until you notice they are counted against your quota. A heavily edited folder of design files can occupy three to five times its apparent size once versions accumulate.

The fix is a retention policy: keep versions for 30 days, not forever. Almost nobody needs a version from eight months ago, and almost everybody is paying for it.

Minimum seat counts and annual commitments

Business tiers frequently require a minimum number of seats and an annual commitment to reach the advertised rate. If you are four people, the advertised per-seat price may not be available to you at all.

The upgrade cliff

Per-workspace plans have hard caps, and the next tier is rarely a small increment. Going from "up to 2TB" to "up to 10TB" can triple the price for capacity you need one-tenth of. That cliff is where teams end up paying for a tier they will not fill for two years.

What to do about it

Separate hot storage from cold storage

Most of what organisations store is never read. Old project files, exports, recordings, backups. Putting that in the same tier as active working files is the most common way to overpay.

Two habits cover most of it:

• Move anything untouched for 90 days to a cheaper archive tier, if the provider offers one. Check the retrieval cost first — cheap storage with expensive retrieval is a trap for backups you actually restore.
• Compress before archiving. Text-heavy exports, logs, and raw captures often shrink by 70% or more with no loss that matters.

Keep the client-side copy

The cheapest storage is the disk you already own. A local copy of your archive on an external drive is not a backup strategy on its own, but it means you can keep cloud storage for the working set rather than everything.

Do not store what you can regenerate

Build artefacts, node modules, caches, rendered previews. If it can be rebuilt from source in under an hour, it does not need to be in cloud storage at all. This sounds obvious and is routinely ignored, because the folder is already there and nobody wants to be the person who deleted it.

Audit access, not just capacity

Per-seat pricing means dormant accounts are pure cost. Every quarter, list the users who have not signed in for 90 days and decide whether they need a licence. This is the single highest-yield storage saving in most companies, and it requires no migration at all.

When self-hosting makes sense, and when it does not

Self-hosting storage — a NAS, or a server running an open-source sync tool — is genuinely cheaper per terabyte at scale. It is also genuinely more expensive in every other dimension: hardware replacement cycles, the person who maintains it, the backup strategy you now own, and the fact that a disk failure is your problem at 11pm.

The honest threshold: self-host if you have someone who enjoys this and you are storing more than a few terabytes. Do not self-host to save money on two hundred gigabytes. You will spend more on your own time than the subscription cost, and you will find out during a restore.

The short version

• Work out your people-per-terabyte ratio. It determines which pricing model fits.
• Find the egress rate before you commit to a large archive.
• Set a version retention policy instead of paying for history you never read.
• Compress and archive anything untouched for a quarter.
• Audit dormant seats quarterly. It is the cheapest saving available.

The sticker price is the beginning of the calculation, not the answer. The providers know this, which is why every pricing page has a "contact us" tier at the bottom.