The real cost of keeping an aging PC, with the numbers that hold up

The figure everyone quotes, and what it actually measured

A large per-device dollar figure for the cost of a PC older than four years circulates in vendor decks and refresh business cases, usually described as a cost per year. It comes from a 2018 Techaisle study of 2,156 small and mid-size businesses across Asia Pacific, commissioned by Microsoft. Microsoft's own releases describe it as the cost of keeping a PC more than four years old, or the total cost of owning such a PC, per device. They never say per year. The number has also crossed eight years, a region and a market segment on its way into enterprise refresh cases in North America and Europe. We do not use it, and we would advise you not to.

The pattern is worth naming because it is common. A sponsored study produces a striking number, the sponsor rounds it into a slide, and the unit quietly changes as it is repeated. Finance teams notice, and once they do, the whole business case loses credibility. Better to use smaller numbers with their units intact.

The figures that survive a careful read

Techaisle ran a second study for Intel and Microsoft in Canada in 2019, with 175 small and mid-size businesses. That release does state a per-year figure: the average cost of keeping a PC more than four years old is CA$1,710 per device per year, with such PCs 4.6 times more likely to undergo repairs, costing an average of 196 hours of lost productivity per device. It is still vendor-sponsored and still small business, and we quote it only with those labels, but at least its unit is clear.

An older Intel release from 2013, again from Techaisle, surveyed 736 small businesses across six countries and found that repairing a PC four years or older cost $427, that repair and maintenance was 1.5 times more frequent on such PCs, and that 42 working hours were lost when an older PC needed repair. The per-repair cost and the frequency multiplier are the two numbers a service desk can check against its own tickets, which is their real use.

What Microsoft measures on the device itself

The most defensible aging signals come from the management platform, because they are measured rather than surveyed. Endpoint analytics records startup time, restart count and blue screen count per device. Microsoft's startup performance guidance states that devices booting from a hard disk typically take three to four times longer to start than devices with a solid-state drive, and recommends the upgrade. The battery health report in Advanced Analytics treats capacity below 60 percent of design as most impacted and runtime under three hours the same way.

There is a new age-related risk in 2026 as well. The Secure Boot certificates Microsoft issued in 2011 began expiring in June 2026, with the Windows Production PCA 2011 certificate expiring on 19 October 2026, and Microsoft's guidance in KB5062710 explains how devices move to the 2023 certificates. Older devices whose manufacturer no longer ships firmware updates may have trouble taking the new certificates. The per-model impact is not something we can quantify from public sources, so treat it as a flag to check against your BIOS inventory rather than a number.

The direct cost that has a Microsoft price tag

For a device still on Windows 10, the cost of doing nothing has a published component. Microsoft Learn prices commercial Extended Security Updates at $61 per device for year one, doubling each year for at most three years, and cumulative, so a device enrolling now in year two pays $183 and a device kept through year three pays $427 in total. Year two runs from 14 October 2026 to 12 October 2027. A device that cannot take Windows 11 and stays in service is either an ESU line or an unpatched machine, and both belong in the cost of doing nothing.

Pricing it from your own service desk

The honest way to price an aging cohort is to use your own tickets. Pull incidents by device for the last 12 months from the service desk, join them to the inventory export by serial number or device name, and compare the ticket rate for devices past class life against devices inside it. Multiply the excess tickets by your hours per ticket and your loaded service desk rate, and you have a per-device, per-year figure with your own name on it. In our sample report the aging cohort is shown as a count of devices with reliability or Windows 11 flags, and the report says the support cost is priced from the customer's own ticket data. We do not substitute an industry figure.

Add the lost user hours only if your finance team accepts them; many do not, and the business case should stand without them. The verdicts page shows where the aging cohort sits in the model: the replace verdict claims no savings, and its cohort appears under the cost of doing nothing instead. The initial report produces that cohort and the ESU exposure for your fleet from an export, and the readout call is where we agree the ticket data you want to price it with.

Common questions