The short answer: As of March 31, 2026, Backblaze had 345,662 drives in service. After excluding boot drives and groups that were too small, it evaluated 341,263 drives: the quarterly annualized failure rate (AFR) was 1.24%, while the lifetime AFR was 1.39%. The worst performer was the 14TB Seagate ST14000NM0138, with 4.90% for the quarter and 5.78% over its lifetime. Among large groups, the best were the 14TB WD Ultrastar WUH721414ALE6L4 at 0.33% and the 22TB WUH722222ALE6L4 at 0.38%. Three models recorded no failures during the quarter. However, these figures come from a data center where enterprise drives run continuously—for a home NAS using WD Red Plus or IronWolf drives, they show how failure rates behave, not which drive you should buy.
Backblaze has published drive failure statistics since 2013, and its dataset is the only public one containing hundreds of millions of drive days for specific model numbers. The latest report covers the first quarter of 2026. It was published on July 9, 2026, roughly two months later than usual—the reason is instructive in itself, as explained below—and at the time of writing in late August 2026, it is the most recently published quarter.
How Many Drives Backblaze Tracks and What a 1.24% AFR Means
Of the 345,662 drives tracked, Backblaze excluded 3,907 boot drives and 492 drives belonging to models below the threshold for quarterly evaluation—more than 100 drives and 10,000 drive days are required. This left 341,263 drives across 33 models, representing 30,203,180 drive days and 1,030 failures.
AFR, or annualized failure rate, expresses failures as an annual rate: the number of failures divided by drive days, multiplied by 365. For the 22TB WUH722222ALE6L4, that means 42 failures over 3,992,942 drive days, or 0.38%. Quarterly figures are sensitive to group size: Backblaze has 129 Seagate ST16000NM000J drives in its fleet, and a single failure during the quarter pushed the AFR to 3.61%. The lifetime table is more useful for long-term conclusions and has a threshold of 500 drives and 100,000 drive days: 529,968,464 drive days, 20,212 failures, and an AFR of 1.39%.
The trend is downward: the annual AFR was 1.36% in 2025 and 1.55% in 2024, while Q1 2025 came in at 1.42%. That is despite an aging fleet—the average age of individual models ranges from 2.5 months for the 26TB WD to 112.3 months for the 8TB Seagate ST8000DM002.
Models That Failed Most Often in Q1 2026
A selection from the 33 models: the worst performers, the most numerous models, and those with consistently low long-term rates. Source: Backblaze Drive Stats for Q1 2026, including the quarterly and lifetime tables; product families matched to model numbers using Western Digital and Seagate datasheets.
| Model | Capacity | Drives | Q1 2026 AFR | Lifetime AFR | Note |
|---|---|---|---|---|---|
| Seagate ST14000NM0138 | 14 TB | 1,251 | 4.90% | 5.78% | drives supplied in Dell servers, high failure rate since 2021 |
| Seagate ST10000NM0086 | 10 TB | 981 | 4.63% | 3.13% | average age of 98 months; Backblaze says they are reaching the end of their service life |
| Seagate ST12000NM0007 | 12 TB | 983 | 4.59% | 2.25% | older 12TB generation |
| HGST HUH721212ALN604 | 12 TB | 9,780 | 3.98% | 2.28% | 81 months old |
| Seagate ST24000NM002H (Exos X24) | 24 TB | 9,605 | 3.29% | 2.89% | average age of 9 months, yet the worst performer among high-capacity models |
| Seagate ST12000NM0008 | 12 TB | 18,650 | 2.83% | 2.09% | — |
| Toshiba MG08ACA16TEY | 16 TB | 4,885 | 2.18% | 2.36% | reached 16.95% in Q3 2025 because of replacements during a firmware rollout |
| Toshiba MG08ACA16TA | 16 TB | 40,036 | 1.05% | 1.07% | second-most common model |
| Toshiba MG07ACA14TA | 14 TB | 37,263 | 1.03% | 1.03% | 64 months old, stable |
| WDC WUH721816ALE6L4 (Ultrastar DC HC550) | 16 TB | 26,801 | 0.97% | 0.51% | lowest lifetime AFR among large groups |
| Seagate ST16000NM001G (Exos X16) | 16 TB | 34,729 | 0.52% | 0.68% | — |
| WDC WUH722222ALE6L4 (Ultrastar DC HC570) | 22 TB | 45,638 | 0.38% | 0.58% | most common model in the fleet |
| WDC WUH721414ALE6L4 (Ultrastar DC HC530) | 14 TB | 8,640 | 0.33% | 0.52% | 61 months old |
| Toshiba MG11ACA24TE | 24 TB | 7,203 | 0.22% | 0.42% | only 4 months in service |
There are two things worth remembering. „Manufacturer" is not a useful category: Seagate has one model with a lifetime AFR of 5.78% and another with 0.68%, while Toshiba has models at 2.36% and 0.42%. The model, generation, and sometimes firmware are what matter. Nor is age alone a death sentence: the ST8000DM002—the only remaining consumer drive in the table, deployed in 2016—has an AFR of 1.43% after 112 months, while the 24TB Exos X24 is at 3.29% after only nine months.
Models with Zero Failures and Product Families That Remain Reliable over Time
Three models recorded zero failures during the quarter: the HGST HMS5C4040BLE640 (4 TB, the final 186 drives, averaging 106 months in service), HGST HUH728080ALE600 (8 TB, 1,026 drives and 86,921 drive days), and Seagate ST16000NM002J (16 TB, 466 drives). The Seagate ST8000NM000A, ST12000NM000J, ST14000NM000J, ST16000NM000J, and Toshiba MG09ACA16TE each recorded one failure.
Even zero is only a snapshot. The same HUH728080ALE600 had an AFR of 10.29% in Q4 2025. Backblaze investigated its sensitivity to vibration at one location and planned a migration; one quarter later, it recorded zero failures. A group of roughly one thousand drives is smaller than a single Vault, which contains 1,200 drives. It is located at one site, and a single environmental change can move the figure by an order of magnitude.
Over their lifetimes, WD Ultrastar helium-filled families stand out at the low end: the 16TB DC HC550 at 0.51%, 14TB DC HC530 at 0.52%, 22TB DC HC570 at 0.58%, and HGST HUH721212ALE600 at 0.62%. Among Seagate drives, the 16TB Exos X16 at 0.68% and ST12000NM001G at 0.99% remain below 1%, as does Toshiba’s MG10ACA20TE at 0.69%. Above 2% are the ST14000NM0138, ST10000NM0086, Exos X24, MG08ACA16TEY, HUH721212ALN604, and both older 12TB Seagate generations, the NM0007 and NM0008.
Same Capacity, Different Model: Two 14TB Families Side by Side
Two 14TB models have been running at Backblaze for almost exactly the same length of time: 60.3 and 61.3 months. The Seagate ST14000NM001G (Exos X16): 10,519 drives, 944,400 drive days, 21 failures, and an AFR of 0.81%. The WD WUH721414ALE6L4 (Ultrastar DC HC530): 8,640 drives, 776,557 drive days, 7 failures, and an AFR of 0.33%. Their lifetime AFRs are 1.38% versus 0.52%—the same capacity, data center, and age, but a 2.5-fold difference.
At 24 TB, the difference is greater, although both groups are young: the Seagate ST24000NM002H is at 3.29% across 9,605 drives, compared with 0.22% for 7,203 Toshiba MG11ACA24TE drives. At the same time, Backblaze is investing most heavily in high-capacity drives: of the 10,220 drives deployed during the previous quarter, 9,404—92%—had capacities above 20 TB. The entire group of drives with capacities of 20 TB or more, totaling over 86,000 units, has an AFR of 0.85%.
What Manufacturers Promise: An MTBF of 2.5 Million Hours Equals 0.35% on Paper
The enterprise product families purchased by Backblaze all specify an MTBF of 2,500,000 hours and a projected AFR of 0.35%: the WD Ultrastar DC HC570 and DC HC550, rated for up to 550 TB per year and covered by a five-year warranty; the Seagate Exos X16 and Exos X24, rated at 0.35% over 8,760 operating hours per year and up to 550 TB per year; and the Toshiba MG08. The calculation is simple: 8,760 hours in a year divided by 2,500,000 hours equals 0.35%.
For those same models, Backblaze measures lifetime rates of 0.51% for the DC HC550, 0.58% for the DC HC570, 0.68% for the 16TB Exos X16, 1.07% for the MG08ACA16TA, and 2.89% for the Exos X24—between 1.5 and 8 times higher than the figures on paper, despite operating in an environment closer to datasheet conditions than anything found at home.
The reason lies in the footnotes. For the DC HC550, WD states that MTBF and AFR are estimates based on statistical measurements and acceleration algorithms at a workload of 220 TB per year and a temperature of 40°C. For the DC HC570, it adds that MTBF and AFR „do not predict an individual drive’s reliability and do not constitute a warranty.„ For the WD Red Plus, which most home NAS owners buy, the 1,000,000-hour MTBF is derived at 90 TB per year and 40°C, even though the permitted workload is 180 TB per year. Toshiba says the N300’s MTTF „is neither a guarantee nor an estimate of product life." MTBF describes a population under laboratory conditions, not the lifespan of your individual drive.
What people actually buy for NAS systems in the Czech market—we compare them in WD Red vs Seagate IronWolf vs Toshiba N300—includes the WD Red Plus WD20EFPX through WD120EFGX: CMR, 180 TB per year, 1,000,000 hours, and a three-year warranty; WD Red Pro WD102KFBX through WD260KFGX: CMR, 550 TB per year, 2,000,000 hours for the 10–12TB models and 2,500,000 hours for the 14–26TB models, with a five-year warranty; and Toshiba N300: CMR, 180 TB per year, 1,000,000 hours for the air-filled 4–10TB models (HDWG440 through HDWG71A) and 1,200,000 hours for the helium-filled 12–24TB models.
Why Backblaze Cannot Tell You Which Drive to Buy
- The drives you buy are not in the table. WD Red Plus, IronWolf, N300, WD Blue, and Barracuda drives are absent. Backblaze buys enterprise models by the thousand. The Exos X16, Ultrastar DC HC570, and Toshiba MG08 may be available from large Czech retailers, but hardly anyone buys them for a four-bay NAS.
- The environment is different. The drives run 24/7 in cooled racks, undergo qualification testing, and are subject to active risk management. In Q1 2026, Backblaze described a model affected by two concurrent mechanical faults—one during writes and another during power cycling. The solution was to reduce power cycling and switch the affected Vaults to a no-upload mode. A NAS shut down every night or a drive kept in a drawer experiences precisely the kind of stress the data center deliberately eliminated.
- The definition of failure is operational. A failed drive is one that Backblaze removed from service according to its own criteria. The Q1 2026 report acknowledges that the program does not detect first-day failures because a drive must have been in the inventory on the previous day, meaning day-one failures are undercounted. Drives removed while still operational during migrations to higher capacities do not enter the failure statistics at all.
- The numbers belong to a cohort, not an entire product family. The Toshiba MG08ACA16TEY had an AFR of 16.95% in Q3 2025 because the figures included drives removed during the rollout of firmware supplied by Toshiba. It fell to 4.14% in Q4 2025 and 2.18% in Q1 2026. Its sister TA and TE variants, which differ in sector format—4Kn versus 512e—and optional secure erase features, have AFRs of 1.05% and 1.38%. Anyone who had „ruled out Toshiba" in November 2025 would have been reacting to firmware, not mechanical design.
- Small groups mislead in both directions. One failure among 129 drives produces 3.61%, while zero failures among 186 drives produces 0%—and the same 8TB HGST model can swing between 10.29% and 0%.
What the data can tell you is that generation and model matter more than the logo; enterprise CMR drives from all three manufacturers generally remain below 1% under good conditions; and some models are consistently poor performers. Backblaze, Seagate, and Dell were already investigating the ST14000NM0138 in 2021. It recorded 6.86% in Q3 2025 and 4.90% in Q1 2026.
What This Means for a Home NAS or Small Business in the Czech Republic
A drive will fail—you just do not know which one. In a four-drive NAS, if the fleet-wide AFR of 1.24% applied, the probability of at least one drive failing within a year would be 1 − 0.9876⁴ ≈ 4.9%. At 0.38% for the best large model, it would be about 1.5%; at 4.9% for the worst, approximately 18%. At home, the risk is probably higher. A backup outside the NAS therefore matters more than your choice of brand—our article on the 3-2-1 rule explains how to set one up.
Do not build an array from a single shipment of one model unless you have to. Backblaze says that when buying at scale, a defective batch can fail all at once. In a four-drive RAID 5 array built from one box, the second failure is likely to occur during the rebuild—we explain why rebuilds kill RAID arrays in a separate article.
Age is a reason to make a plan, not to panic. According to the bathtub-curve analysis from October 2025, the fleet has an AFR of around 1.30% during the first year, while the failure peak has shifted to 10 years and 3 months, when it reaches 4.25%. In data from 2013, the peak was 13.73% at around 3 years and 3 months; in 2021, it was 14.24% at 7 years and 9 months. A five-year-old drive is not automatically finished—but an array of five-year-old drives from the same shipment needs a replacement schedule, not a policy of waiting for failure.
CMR is essential. Backblaze’s tables contain enterprise CMR models; the statistics say nothing about SMR. Our CMR vs. SMR overview explains why SMR drives do not belong in an array.
What the Lab Sees: How Drives Actually Die
Statistics tell us how many drives failed, not why. In the lab, most HDD recovery cases involve mechanical or electronic damage: damaged read/write heads following a fall or wear, a seized spindle motor, burned electronics after a power surge, damage to the firmware service area, and—especially in older drives—bad sectors that an array may fail to detect for years until a rebuild begins. Replacing the heads in a helium-filled enterprise drive with ten platters is more demanding than doing the same in a 4TB air-filled NAS drive: it requires a precisely matched donor drive and longer imaging. High capacity may reduce AFR on paper, but it raises the cost of a mistake. Because known weaknesses—firmware, adaptive head parameters, and encrypting bridges in external enclosures—are tied to specific drive families, we always ask for the exact model, not merely the brand.
You Own a Drive from a Product Family with a Higher AFR: What to Do
- Find the exact model number on the label or using a SMART utility, then look it up in the Backblaze table or downloaded dataset. If it is not listed, the statistics say nothing about it—which is true for most consumer drives.
- Back up first. If the data exists only on that drive, copy it elsewhere today, onto another medium and ideally into the cloud as well. Copy the entire contents in one go, not piecemeal across several sessions.
- Use SMART, but do not overestimate it. Reallocated, pending, and uncorrectable sectors are reasons to replace a drive; „PASSED" is not proof of good health. The process is explained in how to tell whether a drive is dying.
- Plan the replacement. Replace drives from a product family with a consistently high AFR, or an array of drives from the same shipment that are five or more years old, one at a time. Let each rebuild finish before replacing the next drive—the procedure is covered in our article on upgrading a NAS with larger drives.
- When a drive starts acting up, back it up—do not repair it. Read errors, disappearing from the system, or slowing down mean it is time for one final copy, not CHKDSK.
What Not to Do
- Do not repeatedly power on a drive that clicks or fails to spin up. Every startup with damaged heads scrapes the platter; follow our first aid guide for a clicking drive.
- Do not replace drives according to brand rankings. An array built from „Backblaze winners" but lacking a backup is still an array without a backup.
- Do not power-cycle your NAS unnecessarily. Backblaze addressed a fault linked to power cycling by reducing it; continuous operation is gentler on mechanical components than daily startups.
- Do not ignore errors on a young drive. Backblaze says its own figures undercount failures in the first year; a warranty replaces the drive, not your data.
- Do not treat MTBF as a lifespan. A figure of 2,500,000 hours does not mean 285 years of operation; it means 0.35% per year across a population under laboratory conditions.
Frequently Asked Questions
Is Seagate Worse Than WD or Toshiba?
Not as a brand. In the same fleet, Seagate has a model with a lifetime AFR of 5.78%—the ST14000NM0138—and another at 0.68%—the 16TB Exos X16. WD has the DC HC550 at 0.51% and the WUH721816ALE6L0 variant at 1.04%. The model and generation are what matter.
Which Drive Should I Buy for a NAS According to Backblaze?
None directly—the consumer NAS product families are not present in the dataset. The useful conclusion is to choose a CMR drive from a NAS or enterprise family with a verified model number, avoid building the array from a single shipment, and keep a backup outside the array.
Does an AFR of 1.24% Mean That One Drive Breaks for Every Hundred People?
It means that, under Backblaze’s workloads and operating conditions, roughly 1.24 drives out of every hundred in its fleet would fail over a year. At home, temperature, vibration, power cycling, and age shift that figure—and it says nothing about a particular drive.
Should I Throw Away an Old Drive Even If It Still Works?
Not because of age alone. Backblaze operates 8TB drives from 2016 with an AFR of 1.43%, as well as 4TB drives that have been running for more than 106 months. Replace drives based on SMART data, behavior, and the state of your backups, not the calendar.
A Drive from a Product Family with a High AFR Has Started Clicking. What Now?
Power it off. Do not try again, write to it, or freeze it. If it contains the only copy of your data, it belongs in a recovery lab; every additional startup increases the damage.
When a drive—whether it appears in the table or not—stops responding, starts clicking, or disappears from the system, power it off and contact us. At the ITHOPE laboratory in Brno, we handle HDD data recovery, RAID and NAS array recovery, and external drive recovery. For businesses that want drive replacements and backups monitored before a failure occurs, we also provide IT management and backup services.
Sources
- Backblaze Drive Stats for Q1 2026 — Q1 and lifetime tables, criteria, two mechanical faults, and undercounted day-one failures
- Backblaze: Q1 2026 press release — 92% of new drives above 20 TB, 0.85% AFR among more than 86,000 drives of 20 TB or more, and Q1 2025 AFR of 1.42%
- Backblaze Drive Stats for 2025 — annual AFR of 1.36% (2024: 1.55%), HUH728080ALE600 at 10.29%, MG08ACA16TEY at 4.14%, and 1,200 drives per Vault
- Backblaze Drive Stats for Q3 2025 — MG08ACA16TEY at 16.95% and Toshiba firmware, ST14000NM0138 at 6.86%
- Backblaze Drive Stats for Q3 2021 — ST14000NM0138 in Dell servers and the Seagate and Dell investigation
- Backblaze: Are Hard Drives Getting Better? (bathtub curve) — peaks of 13.73% (2013), 14.24% (2021), and 4.25% at 10 years and 3 months (2025)
- Backblaze Hard Drive Stats for 2016 — ST8000DM002 as a consumer 8TB drive
- Backblaze Hard Drive Test Data — raw data available for download
- WD Ultrastar DC HC570 datasheet (PDF) — WUH722222ALE6L4, 2.5 million-hour MTBF, 0.35% AFR, and estimation disclaimer
- WD Ultrastar DC HC550 datasheet (PDF) — WUH721816ALE6L4, estimated at 220 TB per year and 40°C
- WD Ultrastar DC HC530 datasheet (PDF) — WUH721414ALE6L4
- WD Red Plus product brief (PDF) — 180 TB per year, 1 million-hour MTBF at 90 TB per year and 40°C, and a three-year warranty
- WD Red Pro datasheet (PDF) — 550 TB per year, 2.0–2.5 million-hour MTBF, and a five-year warranty
- Toshiba N300 datasheet (PDF) — 180 TB per year, MTTF of 1.0 and 1.2 million hours, and the MTTF disclaimer
- Toshiba MG08 Series product overview (PDF) — TA/TE/TAY/TEY variants, 2.5 million-hour MTTF, and 0.35% AFR
- Seagate Exos X16 product manual (PDF) — ST16000NM001G and ST14000NM001G, 0.35% AFR at 8,760 hours per year
- Seagate Exos X24 datasheet (PDF) — ST24000NM002H, 2.5 million-hour MTBF, 0.35% AFR, and a five-year warranty
Where to find us
ITHOPE is a company based in Brno, Czechia: we do data recovery from disks, RAID and NAS, PC and laptop service and IT for business, all in our own lab at Viniční 3587/216, 615 00 Brno.
Phone +420 775 556 063 (24/7 line). You can bring the device in person, or send it by courier (Zásilkovna, PPL).