The cheapest petabyte is the one you never buy.
How to find, price, and reclaim the storage you already own, before you pay 2026 prices for more.
Read the full guide
Enter your work email to open all eight chapters, the cost worksheet, and the sprawl score.
You have full access.
Start with chapter 1, or jump to the worksheet to see what your reclaimable storage costs you each year.
Storage stopped being cheap.
Demand is climbing, supply is committed, and prices have moved faster than budgets.
Three forces met at the same time.
Demand keeps growing.
Gartner forecasts data center storage demand to grow 19.5% a year through 2029, and unstructured data already makes up 70% to 90% of what organizations hold. In a Seagate survey, 99% of IT leaders expect AI to increase their storage needs, and only 38% feel fully prepared.
Supply is spoken for.
Drive makers sold their 2026 output to hyperscale AI buyers. Nearline lead times grew from a few weeks to more than 52 weeks.
Prices moved on every tier.
Flash took the biggest jump, but disk did not escape: a 30TB hard drive went from $495 to $1,216 in one year. IDC expects no SSD price relief before 2028.
Price of a 30TB SSD, compared with a 30TB hard drive
Source: VDURA Flash Volatility Index, August 2026. The gap peaked at 23.2x in Q1 2026.
A 2026 refresh planned against 2024 prices no longer exists at that price. The capacity you order arrives next year. The capacity you reclaim is available this quarter.
yearly growth in data center storage demand through 2029, to 3.19 zettabytes (Gartner)
flash shortages expected to last into the second half of the year (Gartner)
the earliest year IDC expects any relief in SSD prices
Most of what you store, nobody opens.
The reclaimable share is rarely in one place. It is spread thin across every tier you pay for.
Less than 10% of stored capacity was accessed in the previous 90 days.
Measured, not estimated.
When CTERA scanned 856 file shares holding more than 16 PB, it found that 88.3% of files had not been opened in more than a year, and less than 10% of capacity had been touched in the last 90 days.
Some of that data must be kept for legal or research reasons. But it does not need to live on the most expensive tier, and a large share of it does not need to exist at all. A 2016 Veritas study put redundant, obsolete, and trivial (ROT) data at 33% of what organizations store.
Compare when a file was last modified with when it was last read. The gap between those two dates is where reclaimable capacity hides.
Duplicates
Copies of the same file across shares, projects, and clouds, each one billed at full price.
A global advertising holding company found 39% of 1.6 PB of Amazon S3 storage was redundant, obsolete, or trivial.
Cold data
Files no one has opened in years, still sitting on fast, premium tiers.
88% of files had not been opened in over a year (CTERA).
Orphaned data
Files whose owners left the company, with no one left to say keep or delete.
88% of organizations still have enabled accounts for people who left (Varonis).
Abandoned AI output
Models, checkpoints, embeddings, and test runs left behind when a project moved on.
Gartner predicts organizations will abandon 60% of AI projects that lack AI-ready data.
Scratch and temporary files
Debug output, core dumps, render caches, and logs that were never cleared.
A Fortune 500 semiconductor company surfaced 1.9 PB of abandoned and ephemeral data.
Dark data
Data no one has classified, so no one knows its value or its risk.
36% of stored data is dark or unclassified (Ponemon).
Storage is only the first line on the bill.
Unmanaged data costs money in five places. Each one sits in a different budget, so the total belongs to no one.
Storage
You pay 2026 prices to keep capacity filled with files no one reads. Every refresh is sized by what you hold, not by what you use.
People
Without one view across storage, every search and every cleanup is manual. Teams lose time hunting for files, recreating the ones they cannot find, and doing by hand what a policy could do on its own.
Cloud
Moving data to the cloud without knowing what it is moves the waste too, and the fees follow it.
Risk
You cannot protect or delete what you cannot find. Only 34% of organizations know where all their data is (Thales).
AI
AI pipelines copy and transform data at every step, which multiplies every other cost on this page.
Individually, each cost looks manageable. Together, they grow every quarter the data is left alone.
Weak data foundations waste $108 billion in AI investment each year, according to Hitachi Vantara.
Put a price on your waste.
Ten minutes, six numbers, and one total you can take to finance.
You do not need a perfect number to start. You need a defensible one. The worksheet opens with a worked example. Replace each value with your own figures, and the totals update as you type.
Enter your own numbers in the white boxes. The example shows the method, not a benchmark. For B, it uses storage prices only.
Not sure about B? Build it from your storage tiers.
Your cost per PB is an average of what each tier costs, weighted by how much data sits on it. Enter the PB on each tier and what each tier costs you per PB per year. The starting prices are the industry averages behind the Diskover Cost of Data calculator. See where these numbers come from. Replace them with the prices on your own invoices and contracts.
| Storage tier | PB on this tier | Cost per PB per year | Yearly cost |
|---|---|---|---|
| Total | 0 PB | $0 per PB | $0 |
Starting prices are storage costs only. To make B fully loaded, add support contracts, power and cooling, backup copies, and admin time.
Score your sprawl.
Six statements. Two minutes. An honest view of where you stand.
Rate each statement from 1 (never true) to 5 (always true). Your score adds up as you go.
Reclaim before you buy.
Four steps. The first three make you lean. The fourth keeps you there.
Connect
Point Diskover at your storage. Distributed indexers read metadata from each system you choose: SMB, NFS, S3, Azure Blob, tape, and archive. Nothing is copied or moved to get started.
One inventory of what you own, across every tier and site.
Which systems hold the most capacity, and which ones no one has looked at in years?
Catalog
Every file lands in one searchable catalog with rich metadata, such as owner, size, location, last modified date, and last access date. Plugins add business context from the systems you already run.
A price on every share, project, and team.
Which teams or projects drive the largest share of your storage bill?
Curate
Rank what is duplicate, stale, orphaned, or abandoned by what it costs to keep. Decide what stays, what moves to a cheaper tier, and what retires. Any authorized team can ask questions in plain language through the AI Connector, and every action it proposes waits for human approval.
A sized list of candidates, largest and most expensive first, with the evidence for each decision.
Who needs to approve each type of decision, and what proof will they ask for?
Activate
Turn each decision into a standing rule. Files move, tier, retain, or clear on the schedule you set, and the catalog records what happened and when.
Savings that keep going, and an audit trail you did not have to build.
Which cleanup did you do once that should now run on its own?
A one-time cleanup helps once. Data keeps arriving, so the waste grows back unless the rule keeps running. Policy is what turns a project into a habit.
What good looks like.
Results from organizations that chose to manage their data instead of buying around it.
$180B entertainment leader
Life science research institute
Major global energy company
Global advertising holding company
Fortune 500 semiconductor company
$314B semiconductor company
None of these teams started by buying storage. They started by seeing what they had. The savings came from capacity they already owned.
Figures reported by Diskover customers. Customer names are not shown.
Build the case finance will sign.
During a shortage, reclaim before you buy is a story finance wants to hear. Here is how to tell it.
A one-page outline.
Lead with the number, not the technology. Finance approves a cost it can see and a plan it can check.
Diskover customers typically recover their investment in 3 to 6 months, from reclaimed capacity and time returned to their teams.
Before your proof of concept
0 of 5 ready
“We hold 2 PB across five systems, at a storage cost of $317,375 per PB per year. At least 30% of it has not been opened in years. That is $190,425 a year we pay to keep data no one uses, before we count support, power, admin, or search time. We propose a 30-day proof of concept on our own data to confirm the number, then policies to reclaim it before our next storage purchase.”
Know what you are paying to store before you pay to store more.
Start with a 30-day proof of concept on your own data, and recover your investment in 3 to 6 months.
