For anyone storing photos, work files or backups on a drive, the practical question is not rated lifespan but whether a second copy exists. Hard drives usually last for years; SSDs consume write endurance and report what remains; microSD cards for constant recording come in endurance versions. None gives a scheduled failure date. Your latest backup date matters more than any advertised service life.
- Cloud fleet snapshot: Backblaze’s second-quarter 2026 report, released on 29 September, covered 354,415 hard drives. The quarter’s annualised failure rate was 1.73%, which the company called the highest “en bastante tiempo”. Its all-time cumulative rate stood at 1.41%.
-
The bathtub peak shifted older: Backblaze’s October 2025 analysis put the failure peak at 13.73% around three years and three months in 2013, 14.24% at seven years and nine months in 2021, and 4.25% at ten years and three months in 2025. In 2025, drives under one year were only just above 1.30%. Summary:
“están mejorando y duran más”
.
- SSD life is measured in writes: Samsung rates the 990 PRO at 600 TBW for 1TB and 1,200 TBW for 2TB, each with a five-year limited warranty, “lo que ocurra primero”. That 600 TB over five years equals roughly 329 GB written daily. Samsung says endurance tests follow JEDEC’s JESD218 standard.
- Endurance cards are their own category: Samsung’s PRO Endurance 128GB advertises up to 70,080 hours — about eight years — of Full HD recording at 26 Mbps; the 256GB version advertises 140,160 hours.
What the Backblaze Numbers Reveal
The best public reliability data comes not from a drive maker but from a large customer: Backblaze, a cloud-storage firm that publishes fleet failure figures quarterly. Its full-year 2025 report recorded an annual rate of 1.36%, down from 1.55% in 2024. Age matters when a model deteriorates. In Q2 2026, Backblaze flagged an HGST nearly seven years old at 7.63%, a Seagate about eight and a half years old at 9.33%, and a 14TB Seagate at 8.26%. The two Seagates had few fleet units, so a handful of failures inflated the percentages.
Backblaze repeats that its conditions differ from yours. Its drives sit in a data centre; a drive carried in a bag faces impacts, power cuts and heat. The figures help compare trends and models, not promise a consumer drive ten years.
TBW, MTBF and What the Labels Mean
An SSD has no moving parts, but its cells wear with writes, so vendors quote TBW instead of years. Backblaze’s mid-2023 report on 3,144 server boot SSDs found a 0.90% cumulative failure rate since late 2018. That sample is old and specialised, so treat it as a reference, not a definitive hard-drive comparison.
MTBF (mean time between failures) estimates a population, not one drive. Samsung lists 1.5 million hours for the 990 PRO — more than 170 years, which nobody expects. Western Digital’s WD Red Plus sheet gives 1 million hours, calculated at 90 TB per year and 40°C, and states:
“no predice la fiabilidad de un disco individual ni constituye una garantía”
. The same sheet’s 180 TB per year workload and three-year warranty are better clues; warranty beats MTBF.
MicroSD cards are flash too, so writes wear them. For surveillance and dashcam use, endurance cards give figures: Samsung’s 128GB PRO Endurance promises up to 70,080 hours of Full HD at 26 Mbps, and its 256GB version promises 140,160 hours. A dedicated line signals what a standard card is not meant for.
S.M.A.R.T. Warnings and the Backup Rule
Almost every drive carries S.M.A.R.T. self-diagnostics. On Windows, CrystalDiskInfo is common; its developer published a July 2026 guide to “Caution” and “Bad” states that should be read in full. On many SSDs, “Caution” appears at 10% remaining life or less, an estimate based on the maker’s TBW, not a day countdown. Hard drives most often trigger warnings through reallocated (05), pending (C5) and uncorrectable (C6) sectors. “Good” does not rule out sudden failure. The rule:
“Back up first. Investigate afterward.”
On Mac, Disk Utility’s First Aid can warn of imminent failure; Apple’s help says to copy data and replace the drive.
Longevity ratings describe populations, not the unit on your desk. Data-centre hard drives often last years, and consumer SSDs rarely exhaust TBW, but averages cannot protect the individual unit. The real question is whether your data survives losing one device. If not, a verified backup routine matters more than any endurance rating; S.M.A.R.T. monitoring is a secondary check, not a substitute.