Backblaze Q2 2026: Drive Failures Climb to 1.73%

Backblaze's Q2 2026 Drive Stats put the quarterly annualized failure rate at 1.73%, the highest in recent quarters. The scary headline hides a split: the increase comes from aging 8 to 14TB drives, while the 20TB and larger models that now make up over a quarter of the fleet are its most reliable.

• Backblaze analyzed 354,415 drives over 31,553,350 drive days and logged 1,498 failures, for a quarterly AFR of 1.73%. The lifetime AFR is 1.41%.
• Seven of 31 models exceeded a 3.0% AFR, and the worst offenders are older 10 to 14TB drives.
• The 20TB+ models in the report run between 0.55% and 1.1%.
• Backblaze says there are no SMR drives in its fleet today, which matters if you are shopping for a NAS.

Did drives actually get worse this quarter?

Partly. A quarterly AFR of 1.73% against a lifetime figure of 1.41% is a real uptick, and Backblaze describes the quarter as its highest in recent ones. But an average across 354,415 drives is a blend of very different animals, and this quarter the blend is being dragged by a handful of models, not by a broad decline in quality.

RelatedAMD buys Taalas, betting on AI models etched into silicon

Seven of the 31 models tracked crossed 3.0%. A few of the ugliest: Seagate's ST10000NM0086 (10TB) at 9.33%, the Seagate ST14000NM0138 (14TB) at 8.26%, and HGST's HUH721212ALN604 (12TB) at 7.63%. Below those sit the ST12000NM0007 (12TB) at 4.17%, WDC's WUH721816ALE6L0 (16TB) at 4.07%, Toshiba's MG08ACA16TEY (16TB) at 3.66% and the ST8000NM0055 (8TB) at 3.44%. Notice the pattern. These are mostly the smaller capacities, which in a fleet that keeps growing means the older purchases, and old drives are where wear-out failures start to show. That last step is my inference, not something the figures prove directly.

Which drives are the reliable ones?

The big ones, by a wide margin. Every 20TB+ model in the report lands at or near the bottom of the failure table, and they do it with serious volume behind them. That is what makes the numbers credible: 45,665 units of one WDC model is not a lucky sample.

Backblaze notes that deployment "continues to concentrate at the larger end of the range," so the reliable drives are also the ones it keeps buying. Three Seagate models, the ST8000NM000A, ST12000NM000J and ST14000NM000J, posted zero failures in the quarter. That is a good quarter, not proof of immortality. A zero is only as meaningful as the number of drives behind it, and I would weigh it lightly next to a 25,000-drive model with a steady 1.1%.

Is this a Seagate problem, or a WD one?

Neither, and this is the lesson I would take from the table. Seagate owns three of the worst figures above and also the three zero-failure models. HGST shows up in the bad list; so do WDC and Toshiba 16TB drives. The same brand is on both ends of the chart. What separates good from bad is the specific model and how long it has been spinning, not the logo on the label.

There is a caveat even for the big-drive story. Seagate's 24TB ST24000NM002H came in at 3.3%, so "bigger is safer" is not a law. It is a trend across the models Backblaze has enough data on, and newer models with less time in service can still surprise.

What about SMR drives in a NAS?

Backblaze's report spends a long stretch on SMR versus CMR. Shingled magnetic recording overlaps tracks to squeeze 10 to 25% more capacity out of the same platters, at the price of needing workloads that suit it. Its verdict on its own fleet is short: "There are no SMR drives in the Backblaze fleet today."

RelatedAMD's Threadripper Halo Station Targets Trillion-Parameter AI

So none of the failure rates above say anything about SMR reliability. For a home NAS, that absence is a useful signal about what a company with hundreds of thousands of drives chooses to run, but it is not a test of SMR. If you are building a RAID or ZFS array that rewrites data often, check whether a drive you are eyeing is SMR before you buy, and lean toward conventional CMR models. Backblaze's whole fleet is evidence that CMR works at scale.

If you are still weighing spinning disks against flash for that build, our earlier piece on SSDs versus hard drives covers the trade.

Our take

A 1.73% headline will get shared as "hard drives are failing more," and that reading is lazy. The accurate one is that an aging cohort of 8 to 14TB drives is hitting the far side of its life, while the newest big drives are doing the job Backblaze bought them for. The fleet is also moving toward them: deployment keeps concentrating at 20TB and up.

For a buyer, that means two rules. First, shop by model and by age, not by brand loyalty, because every major maker appears on both sides of this chart. Second, buy current high-capacity CMR drives, and read the model number before you trust a reputation. Backblaze's numbers come from its data-center conditions, so your home mileage may differ, but the direction is hard to argue with.

Original analysis by GenZTech. Source: Backblaze.