A drive rated for 1.2 million hours does not last 137 years. It just has a very confident data sheet.
The short answer
MTBF (mean time between failures) is a statistical figure in hours for a whole product family. AFR (annualized failure rate) expresses the same idea as the percentage of drives that fail per year [1]. What does it tell you about your own drive? Very little. Seagate states that both are population statistics that are not relevant to individual units [1]. Our assessment: use AFR to compare families and to plan spares, and use backups to protect your data.
How it works
- MTBF in hours. Seagate describes MTBF as the probable average number of service hours between failures, and notes that ratings of 300,000 to 1,200,000 hours might suggest 30 to 120 years of continuous operation, which is not what the figure means [1]. It is estimated from many drives running for a limited time, not from one drive running for decades [1]. Seagate's technology paper treats MTBF, AFR and the workload rate limit as separate metrics with different meanings [4].
- AFR in percent. AFR is the probable percentage of failures per year across installed units [1]. Seagate's data sheet example pairs an AFR of 0.73 percent with an MTBF of 1.2 million hours, assuming 8,760 power-on hours per year [1]. Our arithmetic: 8,760 divided by 1,200,000 gives 0.73 percent, so for small rates the two figures are the same statement in different units.
- Operator data: drive days. Backblaze reports failures per model and counts "drive days", the days all drives of a model were in service [3]. The table's own numbers fit the rule AFR = failures divided by (drive days / 365). Our check: 56 failures over 3,162,117 drive days give 0.65 percent for the Seagate ST16000NM001G, matching the table [3].
- Small samples swing wildly. In the same quarter, Backblaze lists a Seagate model with 171 drives, 14,221 drive days and one failure, which yields an AFR of 2.57 percent [3]. One failure moved the figure by more than two percentage points.
- Models differ a lot. In the same Q2 2026 report (April to June 2026), two 12 TB HGST models show 1.08 and 7.63 percent [3]. A family-wide number hides that spread. Our assessment: look at the model you actually bought, if an operator publishes it.
- Field versus data sheet. A study of about 100,000 disks found annual replacement rates that typically exceed 1 percent, with 2 to 4 percent common and up to 13 percent on some systems, while data sheets implied at most 0.88 percent [2]. It also found that replacement rates grew with age and that failures were correlated in time [2].
- What to do with it. A rate is a fleet average. It does not say when your drive fails and does not replace a tested backup.
A homelab example
The Bytearium homelab runs rack servers (HPE ProLiant DL360 Gen9 and DL380 Gen10) and a QNAP NAS, as of 14 to 18 Oct 2026. We have not recorded drive models or failure history, so we say nothing about them. A general calculation for your own NAS with four drives: at a field rate of 3 percent per drive and year (within the range in [2]), the expected number of failures is 4 x 0.03 = 0.12 per year, roughly one failure in eight years. That is an average, not a schedule: failures can cluster, and rates rise with age [2]. Keep a spare plan and test a restore.
Checklist
- Read the data sheet to see whether it gives MTBF, AFR or both.
- Convert MTBF to AFR: 8,760 divided by the MTBF in hours.
- Check which assumptions the figure rests on, such as temperature and power-on hours [1].
- Compare operators' tables by model, drive count and drive days.
- Distrust rates built from few drives or few failures.
- Use field data (2 to 4 percent) for your own planning, not only the data sheet [2].
- Keep a tested backup and a spare plan whatever the figure says.