The Backblaze hard drive reliability study by Christoph Siemroth of the University of Essex and Yeomyung Park of Sungkyunkwan University finds that HGST drives had the lowest estimated failure rate in a large data-center sample, while Toshiba’s was the highest. The result comes from an adjusted statistical comparison, not from lining up brands’ raw annual failure reports and declaring a winner.
The researchers analyzed Backblaze’s public records for 443,156 hard drives from 2013 through the second quarter of 2025, representing more than 1.66 million drive-years. Their IEEE-published study used duration-regression models to compare manufacturers while controlling for drive age, capacity, form factor and average operating temperature.
What did the Backblaze hard drive reliability study find?
- HGST’s estimated failure rate was about 41% of Seagate’s.
- Western Digital’s was about 52% of Seagate’s.
- Toshiba’s was about 107% of Seagate’s, placing it slightly above Seagate in the models.
That makes the gap between HGST and Toshiba greater than twofold in this dataset after the study’s adjustments. HGST no longer sells new drives as an independent brand: WD acquired the company in 2012 and later wound it down.
The adjustment is doing real work here. Backblaze’s fleet did not add each manufacturer’s drives at the same time. HGST was the most common brand among newly installed drives in 2014; Seagate held that position during much of the later 2010s; Toshiba led in 2023; and WD led in 2024. A raw fleet comparison can therefore end up comparing older drives from one maker against younger drives from another, which is a lousy proxy for a manufacturer effect.
The researchers also had to account for drives that disappeared from the records without a reported failure, known in survival analysis as censored observations. Tom’s Hardware reported that 146,943 drives, about 31% of the sample, left the dataset without a recorded failure, largely during capacity upgrades. The study says this feature matters because removal does not establish that a drive failed.
Does the ranking tell buyers which hard drive to buy?
Only within fairly firm limits. The sample consists of enterprise hard drives operating in Backblaze data centers across several model generations. It is evidence of manufacturer-level differences in that environment, not a verdict on every consumer, NAS or current retail model. Backblaze did not have comparable workload measurements for all manufacturers, and very few sampled drives had run for more than a decade, so the estimates chiefly describe short- to medium-term reliability.
The paper also found that higher average drive temperature was associated with higher failure rates and that higher-capacity drives tended to be more reliable. Those are findings from these models, not a license to treat capacity as a reliability setting. For buyers weighing spinning disks against flash storage, the broader SSD versus HDD trade-off still includes price, capacity, performance and backup planning. A brand average cannot rescue data that exists on only one drive.
This story draws on original reporting from Tom's Hardware.
