How to Sell a Decommissioned Hardware Lot

Not the chassis. The configuration inside it, the completeness of it, and how many identical ones there are.

Two racks of the same server model can be worth very different amounts. Processor generation and core count, total memory and how it is distributed, drive count and capacity, and whether the drives and memory are still installed all move the number more than the model badge does. A node stripped of its processors and memory is a chassis, and a chassis is priced as a chassis.

Completeness is the second lever and it is underrated. Rails, bezels, drive caddies, fan modules and power supplies are all part of what makes a node usable, and a machine missing them needs those parts bought before it can be racked at all. Trays for some platforms cost more than a used drive, and rail kits for a discontinued generation can be genuinely hard to buy.

Consistency is the third. A run of identical nodes from one cluster is easier to place than the same count assembled from odd models, and it is priced accordingly.

A range comes back when the buyer has to hedge against what your list left out. A specific number comes back when it does not.

What to write down, by category. Servers: model and generation, processor part number and count per node, total memory per node, drive bays populated and with what, and whether rails and caddies are fitted. Drives: capacity, interface, form factor and rough quantity per capacity. Flash: the exact part number with a count against each, because capacity alone does not price flash. Memory: module part number, capacity, speed and rank code, with a count per part number. Processors: exact part or spec number with a quantity, and confirmation the pins are intact. Networking: switch model, port count and speed, power supply count, and transceiver part numbers with quantities.

That is the whole job for most lots and it fits on one page.

Where something is unknown, write unknown. An honest gap is easy to price around. A wrong specification found at collection is what forces a valuation to be revised.

The most common way a good lot is described badly is as a list of assets rather than a list of configurations.

An asset register has one row per machine, keyed on an asset tag and a serial number, and it usually says almost nothing about specification. A hundred rows of that tells a buyer far less than five rows saying twelve of this build, twenty of that one, and eight oddments listed individually.

So group before you send. Put the identical machines together, describe one of them properly, and put a count against it. If you have a fleet where most machines match and a handful do not, describe the majority build once and list the exceptions separately.

This is also the fastest route for you. Describing one node properly takes five minutes. Typing a hundred asset rows takes an afternoon and produces a worse valuation.

And send what you already have rather than building something new. A rack elevation, an inventory export or a half finished spreadsheet is a perfectly good starting point.

The instinct is to pick out the attractive items, place those individually at a good number, and offer the remainder to whoever will take it. It is the most expensive habit on the sell side.

What happens is predictable. The parts carrying the value leave first: the current generation nodes, the accelerators, the high capacity drives, the recent switches. What remains has to find its own buyer on its own merits, and it does not add back up to what the whole would have been worth. The remainder often sits for months, and storage, handling and the time it consumes are real costs.

It is also work. Listing, packing and shipping items individually consumes staff time that the improvement rarely covers.

There are genuine exceptions, normally where one line is both high value and easy to place on its own. The way to find out is to have the whole lot valued first and then ask which parts, if any, do better separately. That order costs you nothing and tells you the answer. The reverse order cannot be undone.

Sellers routinely strip things out on the way to the pallet that were adding to what the lot was worth.

Leave the rails, bezels and drive caddies fitted. They are commonly pulled and thrown away, and they add value rather than getting in the way of it. If drives are already out of their sleds, keep the sleds and the screws with them, because a sled missing its fixings is noticeably less useful than a complete one.

Leave the memory and the drives in the servers unless you have a reason to remove them. GTR Supply takes memory either installed in the servers or pulled and boxed, both equally, so there is nothing to gain from the labour of pulling it. The same is true of drives, subject to whatever your data policy requires.

Do not test, sort or grade anything first. That is the work the buyer does, and a lot that has been picked through is usually worth less than the same lot left intact.

Do separate new sealed stock, used stock and customer returns on the list, because they are valued differently.

The step that most often decides whether an internal disposal gets approved has nothing to do with price.

Every drive is securely wiped and a certificate of destruction is available on request, so the chain of custody is documented for your compliance team. Switches and routers are reset to factory defaults, which clears stored configuration, logs and credentials.

Raise your specific requirement while the valuation is still being put together rather than after terms are agreed. Organisations need different things. Some need certificates issued against specific serial numbers. Some need drives destroyed physically rather than wiped. Some need drives pulled and retained on site before anything else is allowed to leave the building. All three are workable, and each one changes both what the lot is worth and how the collection has to be organised.

If you have a written disposal policy, send it with the list. Working from the document is faster than working through the requirements one question at a time, and it prevents the version where everything is agreed and then reopened.

Most delays are not price disagreements. They are these, and all five are avoidable.

Waiting to build a perfect manifest before contacting anybody. Send what you have. A list that gets the high value items right beats a complete inventory that describes everything vaguely.

Guessing at a specification instead of marking it unknown. A wrong figure found at collection forces a valuation to be revised, which is the conversation nobody enjoys.

Nobody owning the physical work. Disconnecting, de-racking and moving equipment to a loading area is separate work with its own owner, and projects slip when everyone assumes somebody else is doing it.

Site access discovered late. Whether there is a loading dock, whether a tail lift is needed, whether access runs through a goods lift, and whether there are time windows or contractor security procedures all change how a collection is booked.

Data requirements raised after terms are agreed. It is the one item that can reopen everything, so it belongs at the start.