New vs Refurbished Hardware: Total Cost of Ownership

Comparing new and refurbished hardware on purchase price answers the wrong question. The number that matters is what the equipment costs to own over the years you will run it, and that calculation frequently reverses the obvious answer — in both directions.

What goes into the real cost

Six components, and purchase price is only the first.

Purchase price. Where refurbished has an obvious advantage, frequently substantial.

Support. A new server usually comes with vendor support included or available. Secondary-market hardware may not be eligible. That gap has to be filled with spares and internal capability, and both cost something.

Spares. On a supported new platform, the vendor supplies parts. On an unsupported or discontinued one, you hold them. Our spares guide covers building that position.

Power and cooling. Older hardware is generally less efficient, and that difference runs continuously for years. In a colocation facility it is billed directly; in your own building it consumes circuit and cooling capacity you may need.

Service life remaining. A newer platform stays supported by operating systems for longer, which frequently sets the replacement date rather than hardware failure — see our compatibility list guide.

Staff time. Sourcing, testing and troubleshooting unsupported hardware takes hours that a support contract would have absorbed.

Where refurbished wins clearly

Spares for equipment already in service. The strongest case by a distance. The platform is running, the compatibility is proven, and a matching part is exactly what you need. New parts frequently do not exist for discontinued platforms at all.

Capacity tiers. Backup targets, archives and bulk storage where cost per terabyte governs and there is no performance requirement. Our backup guide covers sizing these.

Standardising an existing estate. Buying more of what you already run keeps configurations, firmware testing, staff knowledge and spares pools uniform. That operational saving frequently exceeds the hardware saving, and it is why demand for discontinued models stays strong.

Test, development and disaster recovery. Environments where an outage is tolerable and hardware currency does not matter.

Components with no wear mechanism. Memory and processors either work or fail in ways testing detects, with nothing comparable to a mechanical drive’s wear-out. Our memory guide covers this.

Where new is the better answer

Being direct about this matters, because a supplier who says refurbished is always right is not being useful.

Where vendor support is a requirement. Some organisations mandate it, and some workloads justify it. If a four-hour response contract is genuinely needed, that usually means new.

Long deployments. If the hardware must serve for many years, starting on a platform that is already several years old shortens the useful window — particularly when operating system support ends before the hardware does.

High-density or power-constrained environments. Where a rack is near its circuit limit, efficiency differences translate directly into how much equipment fits.

Performance tiers. Where the workload genuinely needs current-generation capability rather than adequate capability.

Where nobody has time to hold spares. If there is no capacity to manage a spares position and no internal hardware capability, a support contract is doing real work.

The mixed position most organisations reach

Not a compromise — usually the correct answer.

New for production performance tiers and anything under a support requirement. Refurbished for spares, capacity tiers, test environments, disaster recovery and expanding platforms already in service.

That split appears repeatedly because it puts money where risk is and saves it where risk is low.

It also has a practical consequence worth planning for: keep decommissioned hardware that matches equipment still running. It is the cheapest known-compatible spares source available. Our decommissioning guide covers this.

Doing the calculation

For each option, over the years you expect to run it: purchase price, plus support contract cost or spares holding cost, plus estimated power difference, plus any staff time you can reasonably attribute.

Then divide by the years of useful service you expect — and be realistic about that for an older platform, since operating system support is frequently the binding constraint rather than hardware condition.

Two things people leave out consistently. The cost of an outage, which is what support contracts and spares are actually buying. And parts availability, which determines whether the platform stays maintainable for the period you are planning.

That last one is worth asking about before buying either way. A platform whose parts market is thinning has a shorter practical life than its condition suggests.

Common questions

Tell us the workload, the years you expect to run it and whether support is required, and we will advise honestly on which makes sense.