How Long Do Unplugged SSDs, USB Drives and Hard Drives Keep Data? Heat, Wear and Backup Tips

Why flash drives forget​

SSDs and USB sticks both use NAND flash. NAND stores each bit as an electrical charge locked inside a microscopic cell. The insulation around that charge isn't perfect, so the charge leaks slowly and the cell's voltage drifts. That makes stored bits harder to read reliably.

National Instruments' technical documentation explains why power matters. A powered SSD refreshes data in the background, and an unpowered one can't. NI's document says charge loss is accelerated by higher storage temperatures and slowed by lower ones. It also says that for most drive endurance specifications, reliability means 12 months of storage at 40°C at the end of usable drive life. That's the worst case, a worn-out drive.

Wear matters because every write-and-erase cycle damages the insulating layer a little more. A heavily written drive has leakier cells than a fresh one.

Flash types, briefly​

• SLC stores one bit per cell. It's the most durable, and you rarely find it in consumer drives.
• MLC stores two bits per cell. It's mostly found in enterprise gear now.
• TLC stores three bits per cell. It's common in consumer SSDs.
• QLC stores four bits per cell. It has the lowest endurance and is also common in consumer drives.

MakeUseOf notes that most consumer SSDs use TLC or QLC because those types are cheaper and offer higher capacities. More bits per cell means telling more charge levels apart, and that leaves less margin. Cell type alone doesn't tell you how long a particular drive will hold data. Look for a product-specific retention spec.

What the JEDEC rule says​

JEDEC's JESD218 is the standard behind the often-quoted "one year" figure. Several sources describe it the same way. It sets a minimum of one year of retention at 30°C for consumer drives, measured at the end of the drive's rated write life. Enterprise drives get a different profile: three months at 40°C.

This number is easy to misread.

• It's a floor for a worn drive, not a typical lifespan. One WindowsForum guide puts it this way: the JEDEC figure is a minimum for a worn drive, not a typical lifespan.
• It's tied to a usage profile. The client class lists active use at 40°C for 8 hours a day, with retention at 30°C powered off for 1 year.
• Temperature changes the result a lot. One analysis of the standard says a client SSD must hold data for only 500 hours at 52°C or 96 hours at 66°C. Those are harsh conditions, not a normal desk or closet.
• A fresh, lightly used drive stored somewhere cool should do far better than the minimum. Heat and heavy wear are what shorten it.

NI's table shows how steep the temperature effect is under its model. At the end of drive life, storage time falls from 365 days at 40°C to 104 days at 50°C, and 2 days at 85°C. The table also shows longer times at lower temperatures, such as 3.8 years at 30°C and 7.8 years at 25°C. These are modeled figures for a worn drive. They aren't promises for any particular product.

Small tests, mixed results​

Three hobbyist experiments get cited a lot. All are small and none is a controlled study, so treat them as illustrations.

Ten Kingston sticks, no rot (so far)​

In 2020, Zachary Vance bought ten Kingston DataTraveler SE9 32GB drives and filled them with random data. He planned to test more of them every few years and to rewrite the good ones afterward. MakeUseOf reports that by early 2026 he had tested five drives and found no bit rot. Tom's Hardware, covering the same test in March 2026, called the results quite promising and said the drives were treated as a small-scale test. The caveat is the sample. It's one model, probably from one batch, stored in presumably benign conditions. It's encouraging for those drives, not proof about USB sticks in general.

Three cheap drives, three outcomes​

Reddit user Carnildo put data on three inexpensive USB drives. One sat in an attic, one sat indoors unpowered, and one sat indoors and was read periodically. After a year, MakeUseOf reports these counts of corrupted images:

In nearly all cases, whole 4KB blocks were randomized. That's why images tended to cut off partway through. The odd part is that the passive indoor drive did worse than the attic drive. That's a good reason not to over-read three non-identical drives. The one result that holds up is that the drive that was read periodically fared best. That's consistent with the idea that regular power-on and reads help, but it doesn't guarantee that every consumer drive refreshes its cells that way.

HTWingNut's worn SSD test​

YouTuber HTWingNut loaded four 128GB TLC SSDs with 100GB of random data and checked it with MD5 hashes. Two drives stayed lightly used. The other two were written to 280TB, far past their 60TBW rating. Then all four were stored unpowered in a climate-controlled room and checked once a year.

• After year one, all four passed.
• After year two, one of the heavily worn drives had four corrupt files. It also took more than 42 minutes to verify, against about 10 minutes the year before.

How-To Geek's coverage of the same experiment adds the right caveat. A sample that small isn't indicative of a larger trend. The result fits what the physics and the JEDEC rule predict: wear shortens retention. It doesn't say how long a typical SSD lasts.

Hard drives fail differently​

Hard drives store data magnetically, so the flash leakage story doesn't apply to them. A hard drive's magnetic recording doesn't need power to persist. That's why many people prefer them for shelf storage, with conditions. The risks are different:

• Heat and strong magnetic interference can damage data.
• The mechanics can fail. MakeUseOf's main warning is that long periods without power can leave the bearings seized, so the drive won't spin. Another guide says a hard drive stored cool and dry usually keeps its data for many years, but its motor can fail to spin up after long storage.
• Once a large area is damaged, recovery is hard and often means paying a data recovery specialist.

I found no source that gives a fixed number of years before an idle hard drive seizes, so don't treat any specific figure as fact.

Which brand?​

MakeUseOf mentions a September 2026 peer-reviewed study of Backblaze data covering about 443,000 drives. It reportedly found HGST ahead of Seagate and Western Digital. HGST is the Hitachi drive business that Western Digital bought in 2012 and later stopped using as a brand for new drives. The study measured failure rates across a data-center fleet. It doesn't speak to how long data survives in a drawer. It also can't predict how one consumer drive will do. MakeUseOf's point that Samsung is missing from Backblaze's data because of capacity, not quality, is a fair one.

What to do about it​

The evidence supports a conservative plan:

1. Keep more than one copy of anything important. The 3-2-1 rule is three copies, on two kinds of media, with one off-site. That's the standard advice, and every source here repeats it.
2. Verify your backups. A backup you haven't read back is a hope. Copy files back, open them, and compare checksums for the irreplaceable ones. How-To Geek suggests a full read and verify pass and compares checksums.
3. Keep drives cool and dry. Avoid hot spots like attics, car glove boxes, and spots beside heaters. Heat is the biggest accelerator in the models.
4. Don't use a worn SSD or a cheap USB stick as your only archive. Wear and heat stack up.
5. Plug in archive drives periodically. One guide suggests at least once a year, checking the files each time. That gives you a chance to catch problems early, though it isn't a guaranteed refresh for every drive.
6. Check health data. SMART tools such as CrystalDiskInfo can show wear and errors. On Windows, a SMART warning is a cue to copy data off the drive now.

The bottom line​

The headline claim overreaches if it means all drives lose files on a predictable schedule. The evidence supports narrower points:

• Unpowered flash can lose data, faster when it's hot and worn.
• Hard drives have their own failure modes, mostly mechanical.
• Small DIY tests range from no errors after six years to hundreds of corrupted images after one.
• No single drive or brand is a safe sole copy.

Your best protection is a verified second copy somewhere else.