The 60-Month Rule: Two Decades of Enterprise Hardware Support Tails

Finding 1 — the number is exactly 60, not "about five years"

A five-year support tail is folklore in enterprise IT. The published dates show something sharper than folklore: not a cluster around five years but a spike on the exact month. 1,530 of 2,775 pairs (55%) are exactly 60 months; 62% land within ±3 months. The interquartile range is 60–60 months — that is, the middle half of the entire dataset is a single value. The full distribution, per vendor →

The spread is where the planning risk lives. 38% of pairs miss the window, and they miss it widely: the observed range runs from 1 months to 203 months, with a 5th percentile of 24 and a 95th of 92. 21% run short of the convention and 17% run long — running long means more support than expected, not less. 286 pairs give under 36 months. Planning on 60 months is right most of the time and wrong by years the rest of it, which is why the exact part number decides.

Finding 2 — it has not moved in 24 years

Median support tail of each annual end-of-sale cohort with at least 15 pairs. The rule marks 60 months on a fixed 0–120 month axis.

23 of the 25 cohorts has a median of exactly 60 months. That is the finding: hardware sold out in 2003 and hardware sold out in 2027 carry the same support tail.

And the vendor mix inverted underneath it. A flat median would be unremarkable if it were the same few vendors throughout. It is not — the composition of the sample changes completely across the period, and the median holds anyway:

Finding 3 — the manufacturer predicts the tail far better than the equipment type

It is intuitive that storage arrays would be supported on a different schedule from wireless access points, and in the aggregate the data appears to agree — the share of tails landing on the 60-month convention differs by 48 percentage points between the highest and lowest equipment type:

That spread is mostly a vendor effect wearing an equipment-type costume. Each type is dominated by a different set of manufacturers, so comparing types compares vendors. Holding the manufacturer constant collapses most of it — below, every vendor publishing enough pairs to span two or more equipment types:

The typical within-vendor gap is 8 percentage points — 4 of the 7 vendors sit at or below it. Between vendors, the same measure runs from 0% to 98%, a spread of 98 points. The exception is Cisco Meraki, whose own equipment types differ by 32 points — so this is a strong tendency, not a law, and Cisco Meraki is the case that proves it needs stating that way. The practical form: a storage array from a manufacturer with a rigid policy is more predictable than a switch from one without. Ask who made it before you ask what it does.

Finding 4 — the support clock belongs to the product, not to your purchase

A support tail is measured from end of sale, which invites the assumption that buying late in the sales window still buys 60 months. It frequently does not: manufacturers set one end-of-service-life date for a product line and let variants reach end of sale at different times underneath it. In 137 cases here — 330 pairs, 12% of the dataset — a single end-of-service-life date covers products with different end-of-sale dates, a median gap of 17 months between the longest and shortest tail under the same deadline. In 20 of them the last-sold variant is left under 36 months.

Concentration, stated plainly: IBM accounts for 57 and Dell accounts for 48 of the 137 cases. This is a pattern in how those manufacturers publish, not a universal rule — but where it applies, counting 60 months from a purchase date is the wrong arithmetic. Check the exact part number →

Method, and what this cannot tell you

A support tail is the number of months between a SKU group's vendor-published end-of-sale date and its vendor-published end-of-service-life date. A pair is included only where the manufacturer published both dates for that group. End-of-service-life dates EOSL.ai derives from a vendor's published support-life policy are excluded — they would all equal the policy length and fabricate the consensus this page reports. Tails of zero or fewer months, or of 240 months or more, are treated as data errors and dropped. Medians throughout, never means: single-bulletin outliers are common and a mean would chase them. Annual cohorts are reported at 15+ pairs and vendor × equipment-type cells at 15+; below those thresholds a figure is withheld rather than estimated.

Equipment types are EOSL.ai's classification, not a vendor taxonomy: each product's published category string is matched against a fixed keyword list — Firewall / security, Wireless, SAN / storage, Server, Router / WAN, Switch, and Other — first match winning. The rule is deterministic, so any cell can be re-derived from the open dataset.

What this cannot tell you. It measures what manufacturers publish, not what they do — a vendor may extend support past its stated date, or offer paid extensions this dataset does not see. It describes the 1,785 products EOSL.ai tracks with complete published date pairs, which is not a random sample of all hardware ever sold and over-represents equipment that has already reached end of life. It says nothing about product quality, reliability, or the quality of any vendor's support. And because a vendor absent from a table is absent for want of published data, absence is never evidence about that vendor. EOSL.ai does not grade, score or rank manufacturers.

Cite this research: “EOSL.ai, The 60-Month Rule: Two Decades of Enterprise Hardware Support Tails, Aug 2026” — https://eosl.ai/research/. Press and research enquiries: [email protected]. Figures recompute on every database refresh, so a citation should carry the date it was read.