Most Reliable Enterprise Hard Drive: Backblaze Failure Rates by Part Number

Most reliable by what measure: observed failures or a maker's rating

Two kinds of number get called reliability, and they answer different questions. The first is an observed failure rate: drives counted in service, failures counted as they happen. The second is a maker's rating, printed on a data sheet before any particular drive has run. This page keeps them in separate tables and never averages one with the other.

The observed figures here come from Backblaze, which publishes failure data for the drives in its own storage fleet. Its Q2 2026 report covers 1 April to 30 June 2026. Over that quarter Backblaze monitored 359,101 drives used to store data, and its fleet-wide annualised failure rate (AFR) for the quarter was 1.73%. The lifetime AFR to 30 June 2026 was 1.41%, which Backblaze describes as a slight uptick.

For comparison, the Q1 2026 report, covering 1 January to 31 March 2026, gave a quarterly AFR of 1.24% and a lifetime failure rate of 1.39%. A single quarter moves more than the lifetime number, which is why we print both wherever Backblaze's lifetime table has the model.

Backblaze also sets a floor before a model appears at all. Its quarterly table includes a model only with more than 100 drives and more than 10,000 drive days, and its lifetime table only with more than 500 drives and more than 100,000 lifetime drive days. A model with a tiny population therefore never reaches either table, but a model just over the line can still carry a figure resting on a handful of failures.

Most reliable enterprise hard drive models in Backblaze's Q2 2026 data

These are enterprise part numbers from Backblaze's Q2 2026 tables, with the population behind each figure. Read the drive count and the average age before the percentage.

Two models pair a low figure with a large, established fleet. WDC WUH722222ALE6L4 shows 0.55% for the quarter across 45,665 drives averaging 19.1 months, and 0.58% over its lifetime in the fleet. Seagate ST16000NM001G shows 0.65% across 35,100 drives at an average age of 43.7 months, the oldest population in this table, and 0.68% over its lifetime. Those are the strongest records here, because a low rate across tens of thousands of drives, at average ages of a year and a half and more than three and a half years, is hard to produce by chance.

Set against the fleet, six of the seven sit below Backblaze's fleet-wide quarterly AFR of 1.73% for Q2 2026, and one sits well above it. Some are also steady from quarter to lifetime: Toshiba MG08ACA16TA reads 1.01% for the quarter and 1.07% over its lifetime across 39,932 drives, a middling record but a consistent one. Toshiba MG10ACA20TE, at 1.10% across 25,270 drives averaging 12.1 months, sits in the same band with a younger fleet.

Toshiba MG11ACA24TE has the lowest lifetime figure in the table, 0.50%, but it rests on 23 failures at an average age of 5.8 months. That is a young population, not a proven one: Backblaze has not yet seen these drives through the 19.1 and 43.7 months of average service the WD and Seagate fleets above have behind them.

At the other end, Seagate ST24000NM002H is the 24 TB model with a 3.30% quarterly AFR and 3.24% over its lifetime, from 78 failures in the quarter. Backblaze reports that in Q2 2026 ten of its 31 models had an AFR above 3%, so this drive is not alone in that band, but it is the one in this table. WDC WUH721816ALE6L4 shows why a single quarter is not the whole story: 1.16% in Q2 2026 against 0.56% over its lifetime.

Backblaze defines an outlier, by a quartile analysis, as an AFR above 6.95%, and its Q2 2026 outliers included an HGST 12 TB and a 10 TB Seagate. Backblaze gives clear reasons: the HGST drive is nearly seven years old, and the 10 TB Seagate was nearly 8.5 years old. Age belongs in any reading of these figures, and the enterprise HDD lifespan page takes that question further.

Maker AFR ratings: why they cannot pick a winner

All five documents in that table land on the same 0.35% figure, so the rating cannot separate one enterprise line from another. It is a design target stated under conditions, and Toshiba spells its conditions out for both the MG08 and the MG11: power on all year, a set annual workload and a surface temperature ceiling.

Western Digital's HC580 sheet is the plainest about what the figure is not. It says MTBF and AFR ratings do not predict an individual drive's reliability and do not constitute a warranty. The same caution, in our view, applies to an observed fleet rate: a population figure tells you how a model behaved across many drives, not what your one drive will do.

We do not set the observed figures against the ratings as if they were the same test. We have not found a statement of the workload or environment Backblaze runs, so a Backblaze AFR above or below a maker's 0.35% is not evidence that the rating was wrong. The hard drive reliability page explains how to read MTBF, AFR and workload in more depth.

Related drives on Amazon.co.uk

These tables show current offers from three of the model families above. The Backblaze figures describe Backblaze's own fleet only, not any offer below. The Toshiba tables are filtered by generation, capacity and interface, so they can also hold sibling part numbers. Prices are pounds sterling inc VAT.

The ST16000NM001G table above holds 2 listings right now. Check the condition on each row, new or used, before you buy.

There is no table here for the two WD part numbers or for the Exos X24 part number named above.

How to use reliability data when choosing an enterprise drive

• Match the exact part number. A Backblaze figure belongs to one part number, and a one-letter difference is a different row in Backblaze's table.
• Read the population before the percentage. Tens of thousands of drives over years of service carry more weight than a few thousand drives a few months old.
• When a quarter and the lifetime figure disagree, give more weight to the lifetime figure, and note how old the fleet is.
• Treat a maker's 0.35% as a rating under stated conditions, not as a forecast for your drive.
• Keep observed and rated figures apart. Averaging them produces a number neither source published.

Whichever model you choose, a failure rate across a fleet does not tell you whether your drive will be among the failures. Redundancy is what protects the data. The RAID calculator below shows what a parity level costs in usable capacity for the drives you are weighing.

For the full quarter, with every model Backblaze tracks and its own commentary, see our Backblaze drive stats digest. For the fleet-wide picture across makers, hard drive failure rates collects the figures in one place.