How long will flash memory actually last in my product, and how do I estimate it?

The estimate, in three steps

The arithmetic is simple; the inputs are where buyers go wrong.

• Bytes written per day: records per hour × size per record × 24. Use peak, not average.
• Total bytes over life: daily write volume × expected service years.
• Compare against endurance: total bytes written versus what the memory grade and wear levelling can sustain, with a safety factor.

Worked example (illustrative)

A logger writes 1MB per minute continuously. That is 60MB an hour, 1.44GB a day, about 525GB a year. Over a five-year service life the total is roughly 2.6TB written.

Whether that is comfortable or marginal depends entirely on the memory grade and how the controller distributes the writes — which is why the question "how long will it last" cannot be answered without knowing the workload. The same device writing 1MB a day instead of a minute would be a completely different problem.

The point of the number is not to predict a date. It is to make the workload explicit so the grade can be chosen against it.

The two factors that decide more than the arithmetic

**Wear levelling.** Flash cells have a finite number of erase cycles. A controller that spreads writes across the whole memory uses each cell far less often than one that keeps rewriting the same area. Without wear levelling, a small region dies early and takes the product with it, regardless of how much total endurance the memory theoretically had.

**Power loss during a write.** Losing power mid-write can corrupt not just the file being written but the file system structures that map the drive. A device that is unplugged, browns out, or has a failing battery will eventually lose power at the worst moment. Handling this is a controller and file system design question, not a memory quality question — and it is often invisible in a datasheet.

When data retention matters more than endurance

A drive written once and stored is a different problem. Flash retains data without power for years, but retention is affected by temperature and by how much the cells have been worn — a heavily written cell holds data for less time than a fresh one.

For archival use the question is not "how many writes" but "how long without power, at what temperature", which is a different specification to ask for. Products that sit on a shelf for years before first use fall into this category.