What Used Enterprise Drives Are Worth: Walking One Mixed Lot Through a Buyer’s Valuation

I spend a lot of my working time reviewing hardware asset recovery programs, and the pattern I keep running into is not that sellers are greedy about what their retired drives are worth. It is that they are working from the wrong variables entirely. Some of them undervalue a pallet badly.

Others hold out for a number the market was never going to pay, and by the time they accept reality, the drives have aged another two quarters.

What sits underneath all of that is a supply problem nobody in the storeroom caused. When new drives cannot be scheduled, the installed base becomes the supply, and equipment that a finance team had already written down to zero starts behaving like inventory.

What follows is not a price list. Published prices for used enterprise storage go stale in weeks, and a fixed figure for a 16TB SAS drive printed in an article was true for someone else, somewhere else, at some point in the past.

What is durable is the set of factors behind the number and the order in which a buyer works through them. So instead of listing those factors, I want to put one lot on the table and walk it across.

Call it four hundred drives out of a decommissioned array estate. Mostly 3.5-inch nearline SAS, a tray of 2.5 inch 10K units, a couple of dozen NVMe drives pulled from a newer tier, some still in their carriers, some loose in bags, none of it powered on since it came out of the racks. That is close to what I see most weeks.

Below is what a buyer establishes about a lot like that, in the order they establish it, and where the number moves at each step.

The Unit Count Is the Number That Does Least Work

The first thing I do with a lot like this is set the count aside. Four hundred is a logistics figure. It tells a buyer how many pallets to plan for and how many hours of bench time to budget, and it tells them almost nothing about what the lot is worth.

Residual value, in the sense that matters here, is what a buyer will actually pay at a given moment, after accounting for what it will cost that buyer to test, sanitize, document, and resell the equipment.

That figure has no fixed relationship to the original purchase price, and it does not scale with how many units you happen to have.

Plenty of the organizations I review never get past the count, and the reason is structural, not careless.

Drive disposal falls between the infrastructure team that knows the hardware, the security team that cares only about sanitization, and the finance team that stopped tracking these assets three depreciation cycles ago.

Nobody owns the number, so somebody invents one, and the invented figure goes into the decommissioning budget as though it had been worked out.

First Cut: Capacity, and Where the Lot Sits Against the Installed Base

The buyer’s first real sort is by capacity, and this is the one variable most sellers have already thought about. It is a genuine driver, and it is also where the thinking usually stops.

Capacity is a strong driver because the buyer’s customer is solving for density: terabytes per rack unit, per watt, per dollar of chassis. Higher capacity serves that better, so older low-capacity drives get squeezed out even when they work perfectly.

Where the reasoning goes wrong is in assuming the relationship is linear or stable. It is neither. Value per terabyte tends to rise across the middle of the range and then flatten at the very top, where the largest drives compete directly with new units carrying a warranty.

At the bottom, per-terabyte value collapses toward zero, because testing, wiping, and shipping a 1TB drive costs much the same as testing, wiping, and shipping a 20TB drive, and that fixed handling cost eats the entire margin on the small one.

The other thing a capacity-only read misses is where the installed base actually sits, which is what decides whether our four hundred drives trade quickly or sit.

Backblaze’s Drive Stats report for 2025, published in February 2026, covers 344,196 drives and shows the fleet concentrated heavily in the 14TB to 16TB band, at 52.06 percent, with 22.81 percent at 20TB and above and 25.13 percent at 12TB or below.

That is a reasonable proxy for where second-hand demand and second-hand supply meet. Drives in that band trade quickly. Drives outside it trade slowly or not at all, and slow is expensive.

Second Cut: The Spec Sheet Behind Each Capacity Number

Having sorted by capacity, the buyer now prices the actual part, not the figure printed on the label. This is the step where I see most of the recoverable value sitting untouched, because these attributes routinely swing a drive’s worth further than a capacity step does.

Interface is the first of them. SAS and SATA drives are not interchangeable in the buyer’s world. SAS drives are dual-ported and are what enterprise arrays and most server backplanes expect, so they serve a different buyer than SATA nearline units, and the two markets can move independently.

NVMe is a third market again, with its own form factor complications: U.2 and U.3 are not universally cross-compatible with existing backplanes, and the newer EDSFF formats, E1.S and E3.S, only fit chassis designed for them.

A drive that only fits equipment nobody is still buying is worth what the scrap value of its enclosure suggests, regardless of how fast it is.

Form factor carries similar weight, and in our example lot it is the tray of 2.5-inch 10K units that tells the story. Those drives filled performance tiers a decade ago, lost that market to solid state, and never recovered it.

The 3.5-inch nearline units in the same lot, the dullest hardware on the pallet, have held up far better, because bulk capacity is exactly what is scarce.

Then there is the enterprise versus consumer split, which sellers usually get directionally right and specifically wrong. Enterprise drives do hold more value, and brand prestige has nothing to do with why.

They carry what a professional buyer needs, namely higher-rated workloads, power loss protection on the SSD side, dual porting, longer firmware support, and predictable failure characteristics. A consumer drive of identical capacity is a different product to the person buying it, and it goes into a thinner channel.

Two spec-sheet details cost sellers more money than any of the above, and I have rarely seen either one checked before a lot is offered:

• Sector formatting. Drives pulled from certain storage arrays are formatted to 520 or 528 byte sectors instead of 512. They are perfectly good drives, but they need reformatting before a general buyer can use them, and some tools and firmware revisions make that awkward or impossible. An unreformatted 520-byte lot gets discounted for the labor and the risk.
• Vendor firmware and locking. OEM-branded drives can be worth more inside their own ecosystem and less outside it, because the population of buyers narrows to people running that vendor’s hardware. Self-encrypting drives whose keys are gone are worse still. If nobody can release the lock, resale value approaches zero, and the sanitization conversation becomes a destruction conversation.

Carriers and caddies belong in this list too. They are cheap individually, and they are missing constantly, and the part of our example lot that arrived in its original trays is worth more per drive than the part that arrived loose in anti-static bags.

Third Cut: Grading the Specimen, Not the Model

At this point the buyer stops describing what kind of drive it is and starts describing this particular drive. That means health data, and it is the part of the walkthrough where mechanical drives and solid-state drives require genuinely different reasoning.

For hard drives, the SMART attributes that carry weight are not the ones most dashboards highlight. Power-on hours (attribute 9) sets the general expectation, but on its own it is a weaker signal than people assume.

What buyers actually inspect are the defect indicators: reallocated sector count (attribute 5), current pending sector count (197), and offline uncorrectable sector count (198).

A drive with 40,000 hours and a clean defect profile is a normal enterprise drive that did its job. A drive with 9,000 hours and a growing pending sector count is a problem someone is about to inherit.

The industry data supports treating age with more nuance than fear. Backblaze’s 2025 figures put the annualized failure rate for its fleet at 1.36 percent, down from 1.55 percent in 2024, across a population that is largely composed of drives that have been running for years. High hours are a discount factor, not a disqualification.

One practical trap catches sellers with lots shaped like our example, because most of it is SAS. SAS drives do not expose the ATA-style SMART attribute table that consumer tools read.

Their health data lives in SCSI log pages, and a seller who runs a SATA-oriented utility across a SAS lot will report that the drives have no data available. Buyers read that as untested.

The two dozen NVMe drives in the lot invert all of that logic. Hours barely matter. Writes do. The figure buyers care about is the percentage of rated endurance already consumed, reported in the NVMe SMART and health information log as Percentage Used, alongside Data Units Written and the available spare figure against its threshold.

Endurance ratings themselves are not marketing numbers pulled from the air: they are derived from defined workloads under the JEDEC standard JESD219A, which specifies the workloads used to rate and verify SSD application classes. That is why a drive rated at 3 drive writes per day, and one rated at 0.5 are describing something measurable.

The misjudgment I see most often is a seller applying hard drive instincts to an SSD lot: reporting power-on hours proudly, having never pulled the endurance figures, then being surprised when the offer comes in low because the buyer had to assume the worst about wear.

Untested and undocumented gets priced as the bad case, because for the buyer it sometimes is.

Fourth Cut: Pricing the Lot as a Single Object

Only now does the buyer put the four hundred drives back together and price them as one thing, because a lot is the unit this market actually transacts in.

The sellers I work with who recover the most have stopped valuing drives at all for the same reason. Composition changes the number in ways individual drive specs cannot explain:

• Homogeneity carries a premium, and our example lot loses on this one. Four hundred identical 16TB SAS drives from a single array, in carriers, on a matching firmware revision, is a product. Four hundred assorted drives spanning six models, three capacities, and two interfaces is a sorting job, and sorting is labor.
• The bad tail drags the good drives down. A lot where fifteen percent of the units are failed or unknown forces the buyer to test everything and to price in the units that will not sell. Pulling the known failures out before quoting almost always nets more than leaving them in for the count.
• Freight economics are real. A palletized lot that ships on a truck costs a fraction per drive of the same drives shipped as parcels, and that difference is part of the buyer’s math whether or not it is itemized.
• Documentation is worth money. A manifest listing model, serial, capacity, interface, and health readings shifts risk off the buyer and shortens the settlement cycle.

All of which explains why the specialist buyers in this segment quote per lot instead of publishing a price list.

Of the firms I have dealt with on drive lots, Big Data Supply is the one I reach for when somebody wants to see what that looks like on paper, because a bulk submission gets priced against the composition of the specific lot and against the market the week it is offered.

That is the only defensible way to put a number on something that is half nearline SAS and half a sorting job. Most of the used hard drive lots I see never get that treatment, mainly because nobody thought to ask for it.

The question to put to a buyer is which parts of your lot moved their figure, and in which direction. A buyer who cannot answer that has not looked at your lot. They have looked at your count.

The Gate That Sits Outside the Valuation Entirely

I should be clear about where sanitization fits, because sellers often expect it to add something to the number. It sits outside the valuation altogether. It is the gate you pass through before a number exists at all, since no legitimate buyer will take a lot they cannot certify.

The method has to match the media, and I still watch this go wrong at sites that consider themselves careful. Degaussing addresses magnetic storage, which means platters and tape.

\Run the same equipment over a solid-state drive and the contents survive intact, because those bits are held as trapped charge in flash cells, not as magnetic orientation. Solid-state media needs a manufacturer-secure erase, an NVMe format, or a cryptographic erase instead.

Getting it wrong in the other direction carries its own cost, because degaussing a working platter drive also erases the servo information the drive needs to position its heads, so a drive you meant to sell becomes a drive you have to scrap.

NIST Special Publication 800-88 Revision 2, finalized in September 2025 in place of the withdrawn Revision 1, sets out the framework most buyers and auditors work from.

The value implication is straightforward. A verified logical wipe preserves the asset and its resale value. Physical destruction removes both. Destruction is the correct answer for some data classifications, but I would rather see it chosen deliberately than applied as a default, and priced as the decision it is.

The Variable I Cannot Grade For You

The last thing sellers misjudge is time, and no amount of careful grading recovers what it takes.

Used storage pricing tracks the new-equipment market, and that market is currently being shaped by AI infrastructure demand.

In September 2025, TrendForce reported that lead times on new nearline hard drives had stretched from just a few weeks to over 52 weeks, a year of waiting for a part that used to arrive next week, and that single figure explains most of what has happened to the used drive market since.

TrendForce’s assessment that same month noted that enterprise SSD contract prices were expected to rise five to ten percent quarter over quarter in the fourth quarter of 2025, and we noted that hyperscale buyers were reconsidering high-capacity QLC SSDs for cold data partly because those drives draw roughly thirty percent less power than nearline hard drives. Conditions like these do not hold still.

My own judgment, offered as judgment rather than as measured data, is that the timing penalty is larger than the grading penalty for most organizations. Drives sitting in a storeroom for a year while approvals grind through lose value on two fronts at once: the market moves, and the technology moves past them.

Getting a quote early, even if the handover happens later, at least tells you what you are holding.

That is the part I would keep if I could keep only one. Everything else in this walkthrough stays recoverable. You can go back and pull the defect counters off the SAS units, pull consumed endurance off the NVMe drives, strip the known failures out before quoting, palletize what is left, and write a manifest that says honestly what is on the pallet.

Each of those raises an offer; none of them is expensive, and all of them work just as well next month as they do today, because a lot that is documented and sorted is a product wherever it is priced. What you cannot go back and recover is the quarter you spent deciding.

In a market where new drive lead times have run past a year, the value of a decision made this quarter is not the value of the same decision made next year, and no amount of careful grading closes that gap afterwards.

Frequently Asked Questions