A USB drive left in a drawer for three years may come back with files that will not open. Nothing was dropped, nothing was corrupted in use, and the drive reports itself healthy.
Flash memory stores data as trapped charge, and charge leaks. That is a physical property rather than a defect, and it is why flash is a transfer medium rather than an archival one.
What changed in 2026
- Higher-density flash spread further. More bits per cell continued to reduce endurance and retention per cell.
- Counterfeit drives stayed prevalent. Fake capacity remained a persistent problem in low-cost channels.
- Controller quality diverged. The gap between reputable and generic drives widened.
- Prices fell. Cheap capacity made drives more disposable and expectations no more realistic.
Why charge leaks
Flash stores data by trapping electrons in a cell. Reading the cell measures the charge level. Over time, charge leaks away, and when it drifts far enough the value reads incorrectly.
Error correction handles a certain amount of drift. Beyond that, data is unrecoverable.
Three factors accelerate it:
Cell density. Storing more bits per cell means more distinguishable charge levels packed into the same range, so less drift is needed to cross a boundary. Denser flash retains data for less time.
Wear. Each write cycle degrades the insulating layer holding the charge. Worn cells leak faster, which is why an old heavily-used drive retains data far worse than a new one.
Temperature. Charge leaks faster when warm. A drive stored in a hot car or a warm loft loses data considerably faster than one stored cool.
Retention figures in the industry are specified for enterprise drives at end of rated life under defined conditions. Consumer USB drives are not specified at all, which is itself informative.
Cheap drives are worse
Inexpensive USB drives frequently use flash that failed qualification for better uses — dies with more defects, or cells already partially worn. Retention is the first property to suffer and the last to be noticed, because it fails silently and long after purchase.
Cheap drives also have simpler controllers with weaker error correction and less wear levelling, which compounds the problem.
Counterfeit drives are a separate issue: a drive reporting 1TB with a fraction of that in real flash writes data that goes nowhere. It appears to work until you read back files written past the real capacity — see fake SD cards for how to detect this, since the same fraud pattern applies to both.
| Use |
Suitable? |
| Transferring files between machines |
Yes |
| Bootable installation media |
Yes |
| Temporary working storage |
Yes, with a copy elsewhere |
| Long-term archive |
No |
| The only copy of anything |
No |
| Backup rotation media |
No — use disks |
Practical implications
Do not archive on flash. Photographs, documents, and anything you want in ten years belong on spinning disks with copies, or on cloud storage, or on both — see NAS backup strategy.
Power on occasionally if flash is holding data long term. Reading a drive lets the controller detect and rewrite marginally degraded cells. Once a year is a reasonable habit and it is not a substitute for a proper copy.
Store cool and dry. Heat is the accelerant, and a drawer indoors is dramatically better than a car or a loft.
Verify after writing. Checking a hash of the copy against the original catches both counterfeit capacity and write failures immediately.
Buy from reputable sources. The price difference is small and the failure difference is not.
Treat SSDs differently. SSDs in use are powered regularly and have far better controllers and wear levelling, so the same concerns apply much less. An SSD unpowered in a drawer for years has the same underlying physics, though better flash and better management.
Common mistakes
- Using flash as archival storage. Retention is finite.
- The only copy on a USB stick. Least reliable common medium.
- Storing in hot places. Accelerates charge loss substantially.
- Buying the cheapest available. Worst flash, weakest controllers.
- Not verifying after copying. Misses counterfeits and write errors.
- Never powering long-stored drives. No opportunity to refresh cells.
- Assuming SSD reasoning applies to USB sticks. Different quality entirely.
FAQ
How long does data last on a USB drive?
There is no reliable single figure. A good new drive stored cool may hold data for years; a worn cheap drive in a warm place may not last one. That uncertainty is exactly why it is unsuitable for archives.
Does reading files refresh them?
Reading lets the controller notice degraded cells and rewrite them, on drives whose controllers do this. It is a weak mitigation rather than a guarantee.
Are SD cards the same?
Same underlying technology and same concerns, with the added counterfeit problem — see fake SD cards.
What should I use for long-term storage?
Multiple copies on different media, with at least one off-site, and periodic verification. The medium matters less than having more than one copy.
Where to go next
For detecting counterfeit capacity, read fake SD cards. For a durable backup approach, NAS backup strategy, and for phone storage decisions, phone storage tiers.