Hard Drive Failure Rates by Brand and Model: Backblaze's Measured Numbers for Q2 2026

Hard drive failure rates in Q2 2026: the fleet as context

Backblaze's Q2 2026 report covers 1 April to 30 June 2026 and was published on 29 September 2026. Its quarterly AFR across the fleet was 1.73%, and the lifetime AFR to 30 June 2026 was 1.41%.

Those are fleet figures, and a fleet figure moves with the mix of drives in it. The quarter-by-quarter record and how the reports are put together belong to Backblaze drive stats; the rest of this page is about the models underneath the fleet number.

Hard drive failure rate by manufacturer, model by model

Backblaze publishes its rates per drive model, not per manufacturer, and this page keeps it that way. We have not added rows together into a brand total. Average ages in just the rows below run from 3.4 months to 83.8 months, roughly seven years, so a brand figure would mix new and old cohorts and measure Backblaze's buying history as much as the drives.

The rows are selected from the Q2 2026 quarterly table, as Backblaze printed them, with each model's drive days for the quarter. Backblaze's quarterly table leaves out small populations, so many models have no row at all.

Read across the makers and each one's rows spread widely. Toshiba's run from 0.58% for the young MG11 24 TB to 1.94% for the MG07 14 TB at 67.1 months. The WDC rows sit between 0.55% and 1.16%, while the HGST 12 TB, at 83.8 months, is at 7.63%. Seagate's run from zero for the ST12000NM000J to 8.26% for the ST14000NM0138.

That is why the answer to hard drive failure rates by brand is a table rather than a number. A buyer choosing between makers is really choosing between models, and the row that matters is the one whose part number matches the drive in the basket.

Hard drive failure rate by manufacturer at the same capacity

Lining up rows of the same size is the fairest way to compare makers in this table, because capacity tends to travel with a drive's generation. At 16 TB all three makers appear: the Seagate ST16000NM001G at 0.65% (43.7 months), the Toshiba MG08ACA16TA at 1.01% (33.7 months) and the WDC WUH721816ALE6L4 at 1.16% (37.5 months), with the small Seagate ST16000NM000J group at 2.57%. Three large groups of broadly similar age, all between about 0.6% and 1.2%.

At 14 TB the comparison is about age and group size as much as maker. The Toshiba MG07ACA14TA and the Seagate ST14000NM0138 average 67.1 and 65.5 months, yet ran at 1.94% and 8.26%, and the Seagate group is about a thirtieth of the Toshiba's size: 1,235 drives against 37,171.

At 24 TB the Toshiba MG11ACA24TE ran at 0.58% and the Seagate ST24000NM002H at 3.3%, on almost identical drive counts of 9,620 and 9,631. The Toshiba group averaged 5.8 months and the Seagate 12.2, so neither row yet says much about middle age, but for groups of that size over the same quarter the gap is large.

None of this ranks a brand. It shows that within one quarter of one fleet, model, age and group size move the rate more than the name on the label.

Reading the failure rate table: outliers, small groups and young drives

If you came looking for a hard drive failure rate chart, the table above is built to work as one: each model's rate sits beside its age, drive count and drive days, which a bar of rates alone would hide. Backblaze flagged three models as outliers for Q2 2026. Two appear above, the HGST HUH721212ALN604 at 7.63% and the Seagate ST14000NM0138 at 8.26%, and both are old, at 83.8 and 65.5 months on average, roughly seven years and five and a half. The rule behind the flag is part of the report mechanics described on Backblaze drive stats.

Small populations swing. The Seagate ST16000NM000J row has 171 drives and 14,221 drive days, and its single failure produces a 2.57% AFR. The ST12000NM000J, with 97,746 drive days, had no failures at all in the quarter. Neither number says much on its own; the drive days beside them show why. The Toshiba MG08ACA16TA, with 3,614,263 drive days and a rate near 1%, is far stronger evidence than a small group at zero.

Very young cohorts need the same caution in the other direction. The WDC 26 TB row averages 3.4 months and the Toshiba MG11 24 TB row 5.8 months, so their rates describe the start of the drives' lives, not their middle.

Lifetime failure rates to 30 June 2026

The lifetime table counts every drive day a model has spent in the fleet, so one quarter's noise matters less. Four of the models above, from the lifetime table to 30 June 2026:

Set side by side, the quarter and the lifetime figure agree closely for these four: the MG08 at 1.01% against 1.07%, the ST16000NM001G at 0.65% against 0.68%, the WDC 22 TB at 0.55% against 0.58%, and the Seagate 24 TB at 3.3% against 3.24%. When the two line up, one quarter is not distorting the picture, and for the Seagate 24 TB the quarter was not a one-off.

Where they disagree, give the lifetime figure more weight for a model with tens of millions of drive days behind it, and treat the quarterly figure for a young model as a rate that is still forming.

Hard drive failure rate after 5 years

There is no single answer at five years, because the curve depends on the model. Backblaze says that in general the failure rate follows the bathtub curve as drives age, 'unless it doesn't'. The Q2 2026 rows show both sides at almost the same age: the Toshiba MG07ACA14TA at 67.1 months ran at 1.94% for the quarter, while the Seagate ST14000NM0138 at 65.5 months was at 8.26%.

Backblaze's fleet-wide age curve, and the age at which its peak falls, is the subject of hard drive reliability. For your own drive, the direct evidence at five years is its SMART record rather than any fleet average.

Data center hard drive failure rate: what Backblaze's fleet is

Backblaze's figures come from its own data centres. It takes a snapshot of each operational drive every day, and the published rates are built from that record, so they describe Backblaze's drives, workloads and environment.

Three limits follow for a UK buyer. The fleet holds the models Backblaze runs, so many drives listed on Amazon.co.uk have no row at all. A row describes one part number, so match the exact part number before applying a row to a listing. And a data centre is not a home NAS under a desk, so treat the rows as evidence about models rather than a promise about your machine.

Using failure rates when choosing a drive

• Compare models at similar average ages: an old model's high rate and a young model's low one are not like for like.
• Weight each row by its drive days: a group with a few thousand drive days can swing on a single failure.
• Compare at the same capacity where you can, as in the 16 TB, 14 TB and 24 TB pairs above.
• Do not turn a model's row into a brand verdict: in Q2 2026 one maker's models ran from zero failures to 8.26%.
• Match the exact part number before applying a row to a listing.

This page does not rank drives or name a most reliable brand. Maker ratings such as MTBF, AFR and workload, and how they relate to these measurements, belong to hard drive reliability.