Hardware refresh cycle: three, four or five years, and how to decide per device

Where the three, four and five year rules came from

Most refresh calendars inherited their number from a lease term or a warranty. Forrester, in its February 2024 Total Economic Impact study of Windows 11 Enterprise commissioned by Microsoft, recorded the organizations it interviewed as refreshing "typically three to four years." Sage Sustainable Electronics, in its 12th Annual ITAM Benchmarking Report released in January 2026 across 56 organizations and 2.5 million assets, found the expected retirement age for laptops had fallen to 4.0 years from 4.3, with desktops at about 4.6 years. The report named Windows 10 end of support, tariffs and interest in AI hardware as the reasons the laptop number moved down.

Then the price of memory moved the number the other way. Gartner wrote in February 2026 that it expects PC lifetime to increase by 15 percent for business buyers by the end of 2026, because higher prices will prompt buyers to hold on to devices for longer. Within twelve months the industry benchmark for a laptop had moved from 4.3 years to 4.0 and then been told to stretch again. A fleet policy pinned to any one of those numbers was wrong for part of the year whichever one it picked.

Why one number for the whole fleet costs money both ways

A calendar rule makes two mistakes at once. It replaces machines that are fine because they crossed a date, and it keeps machines that are failing because they have not. The first mistake is visible in the capital budget. The second is spread across service desk tickets, lost hours and the manager who quietly buys a replacement outside the program.

Device classes make this worse. Our starting assumptions assign a five-year life to Task Worker devices and a three-year life to Developer devices. These are configurable defaults that must be checked against your workload and repair data. Gartner, in a September 2026 note on recommended lifespans, says rising costs, memory constraints and AI PC evolution are disrupting traditional device life cycle strategies and recommends multitiered policies. In plain words, one cycle for the fleet is the wrong unit; the class is the right one, and within the class the individual device is the right one.

Gartner also says, in a December 2025 note on giving PCs a second life, that extending laptops to five years and beyond requires careful planning, product selection and, in almost all cases, a midlife refurbishment. That is the honest cost of a long cycle: a battery, a solid-state drive or a memory upgrade halfway through, planned and budgeted, instead of a device that limps to year five on its own.

A per-device test you can run from your inventory

We assess five signals using the inventory and telemetry your organization has collected; some signals need additional exports or collection policies. The weights below are the defaults; a customer can change any of them, and the report shows what the change does to the split before it applies.

What Microsoft itself says the performance data is for

The resource performance report in Intune Advanced Analytics shows CPU and RAM usage by device, model and manufacturer. Microsoft describes its purpose as optimizing hardware replacement costs by extending the life of devices with no performance problem, or replacing underperforming devices sooner. The battery health report says that for devices with high battery scores you should consider extending their lifecycle usage to reduce costs. Those two sentences, from the vendor whose management platform most fleets already run, are the case for deciding per device rather than by date.

Microsoft added Advanced Analytics to Microsoft 365 E3 from July 2026, according to its Intune blog and pricing page, so many tenants that never bought the add-on now have the report. If yours is one of them, open Reports, then Endpoint analytics, then Resource performance, and look at the model view first. A model that is two years old and spiking on RAM in half its fleet needs a different decision from a model that is five years old and idle.

Setting class lives for your own fleet

Start with the defaults and adjust from your own repair data. The classes we use and their default lives are on the verdicts page: Task Worker at 5.0 years, Knowledge Worker at 4.0, Executive and Engineer at 3.5, Developer and 3D Engineer at 3.0. The life is the point where age alone starts to weigh toward replacement; it is never the point where a device is replaced regardless of the other four signals.

Two thresholds from Microsoft are worth copying into your policy. The battery health report treats below 60 percent of design capacity as most impacted and 60 to 80 percent as moderate. The startup performance page notes that hard disk boot drives typically boot three to four times slower than solid-state drives, and recommends the upgrade. A laptop at 65 percent battery with a spinning disk is a refurbishment candidate, and a far cheaper one than a replacement.

What to do with the calendar you already have

Keep it, and change what it triggers. The calendar should start a review, never a purchase order. Every device that crosses its class life gets scored, and the score produces one of four verdicts: keep, redeploy, right-size or replace. The purchase the calendar would have made is deferred when the device is fine, downgraded to a smaller class when the user never needed the bigger one, and made at once when the signals agree.

On the fictional 7,400-device fleet in our sample report, the four-year calendar had 4,030 devices due and the evidence said 234 needed replacing now, with an identified savings floor of about a third of the budget the calendar asked for. The gap is the budget conversation. If you want the same split for your own fleet, the initial report runs it from an export, and the readout walks through which class lives you should change first.

Common questions