Are recertified hard drives safe? What to know and how to test one

What the labels mean

Drive listings use a handful of words that sound alike but mean very different things:

A lot of recertified drives are enterprise models, like Seagate Exos or WD Ultrastar, that came back from data centers. Those were built for heavy use around the clock, which is part of why they hold up well in a home server. They're also louder than desktop drives, so keep that in mind if the server lives in a bedroom.

Warranty and price

The warranty tells you how much the seller trusts the drive. Typical terms in 2026:

Price is the whole reason to consider one. A computingforgeeks comparison in October 2026 found recertified drives at about $29 to $34 per terabyte, against about $46 to $48 for new. Always compare per terabyte, and I track current disk prices on my personal site if you want a quick reference.

One more thing about the shortage: recertified stock is getting harder to find, because more people are buying it and hard drive makers have most of their new production booked into 2028. If you see a good drive with a solid warranty, don't count on the same deal next month. There's more on that in what the memory shortage changes.

Check the SMART data

Every hard drive keeps a log of its own health, called SMART. Read it the day the drive arrives, before you copy anything onto it. On Windows, CrystalDiskInfo is the easy way. My guide to checking drive health walks through it. On Linux, install smartmontools and run:

sudo smartctl -a /dev/sdX

Replace sdX with your drive (run lsblk to see which one it is). These are the lines that matter:

That last one deserves a warning. The regular power-on hours count can be reset, so a low number doesn't prove a drive is new. Seagate drives keep a second count in a log called FARM, which is much harder to wipe. Recent versions of smartctl can read it:

sudo smartctl -l farm /dev/sdX

This isn't paranoia. In 2025, buyers found Seagate drives sold as new that had 10,000 to 22,000 hours in their FARM logs. That's one to two and a half years of running around the clock. On a recertified drive, plenty of hours is normal and expected. What you're checking for is a drive sold as new that isn't.

Run a full surface test

SMART only reports problems the drive has already found. A full test reads or writes every part of the disk, which brings out weak spots while you can still return it. It takes hours, and on a big drive it can take a few days, so start it and leave it alone.

1. Run the drive's own long test. On Linux, sudo smartctl -t long /dev/sdX starts it, and the output tells you roughly how long it'll take. Check the result afterward with sudo smartctl -a /dev/sdX.
2. Write to every sector. On Linux, badblocks writes test patterns across the whole drive and reads them back. It erases everything on the drive, so triple-check the drive name first:

   sudo badblocks -b 4096 -wsv /dev/sdX

   For drives larger than 16 TB, use -b 8192 instead of -b 4096.
3. On Windows instead, a full format does a similar job. In Disk Management, create a partition and format it with Perform a quick format unchecked. Windows then writes to every sector, which takes many hours on a large drive.
4. Check SMART again. Compare attributes 5, 197, and 198 with what you saw before. If any of them went up during the test, return the drive while the return window is open.

A drive that passes all of that has made it past the riskiest stretch. Drives tend to fail either early in their life or late, the "bathtub curve" that Backblaze's long-running drive reports keep showing. Testing up front is how you catch the early ones.

Where a recertified drive belongs

Recertified drives make the most sense where one failing is an inconvenience, not a loss:

• A second backup copy, like an external drive you back up to once a week.
• A mirrored pair or a NAS with redundancy, where one drive can die and the other keeps your files. Mix drives from different batches if you can, so they don't wear out together.
• A home server for media, or for photo backups with something like Immich, as long as there's another copy somewhere else.

Where I wouldn't use one: as the only copy of family photos or important documents, or as the main drive in someone's everyday PC. For a PC, a new SSD is faster and quieter anyway. The usual rule applies to any drive, new or old: three copies of what matters, on two different kinds of storage, with one kept somewhere else.

Why I don't sell drives from my pickups

None of the hard drives from my e-waste pickups get resold. They're destroyed, or wiped and scrapped. So nothing on this page is a pitch for drives I sell. It's the checklist I'd want anyone to use when buying one from someone else.

If you're getting rid of a computer yourself, wipe the drive first, or take the drive out and destroy it before the computer goes to recycling. And if you're buying a used drive from a stranger, assume it's in the "highest risk" row above and test it like it is.