A SATA II SSD manufactured roughly sixteen years ago has accumulated 1 petabyte of writes, reaching 25 times the drive's rated TBW (Total Bytes Written) limit while remaining operational, according to a report published on 20 June 2026. The drive continues to function — a result that cuts against a persistent misconception in the field.

That misconception is worth naming directly: TBW is a warranty boundary, not a cliff edge. Kingston's endurance documentation is explicit on this point — TBW quantifies the total data a drive can absorb over its serviceable life and sets the threshold beyond which warranty coverage no longer applies. It does not describe a hard failure point. NAND flash degrades probabilistically as program-erase cycles accumulate; manufacturers rate conservatively to bound their warranty liability, not to predict the exact moment a cell array becomes unreliable.

The practical gap between rated and actual endurance has always been substantial, and the NAND generation matters here. Early consumer SSDs — particularly those built on single-level cell (SLC) or the denser multi-level cell (MLC) NAND that dominated the SATA II era — carry significantly higher per-cell endurance than the triple-level cell (TLC) and quad-level cell (QLC) geometries that now dominate consumer and mainstream enterprise lineups. A drive from that period storing one bit or two bits per cell tolerates far more program-erase cycles before the oxide layer in each floating-gate transistor degrades to the point of data retention loss. So while 25× over-rating is striking, it is not physically inexplicable given the NAND vintage involved.

The SATA II interface itself is a historical marker. SATA II (3 Gb/s) was the dominant desktop and notebook storage interface through the mid-to-late 2000s, predating the widespread adoption of SATA III (6 Gb/s) and long before NVMe over PCIe became the performance baseline for consumer storage. A drive still running on that interface after sixteen years has outlasted the interface's own relevance in new hardware — most modern motherboards retain a SATA port for backward compatibility, but SATA II as a primary interface is firmly legacy territory.

For practitioners making retention or disposal decisions on aging storage infrastructure, this result has a practical implication: TBW exhaustion alone is not sufficient grounds to retire a drive. Drive health telemetry — particularly S.M.A.R.T. attributes covering reallocated sector count, uncorrectable error rate, and wear leveling count — gives a more granular picture of actual cell degradation than TBW remaining does. A drive that has sailed past its rated write limit but shows clean S.M.A.R.T. data and stable read error rates is behaving differently from one that hit its TBW limit with a rising reallocated sector count.

None of this overturns the utility of TBW as a planning metric. For procurement, capacity sizing, and warranty management — especially in write-intensive workloads like database logging, video surveillance, or caching tiers — TBW per day per drive workload (DWPD) remains a sound basis for specifying hardware. The point is narrower: TBW expiration is not synonymous with drive failure, and treating it as such can cause premature retirement of hardware that still has usable life.

Looking at what this means for longer-term storage assumptions: the consumer SSD market has conditioned users to expect five to ten years of service life under normal conditions, a figure that holds up well in practice. A result at 16 years and 25× rated endurance sits well outside the median — but it is a useful data point for understanding how conservatively the ratings are drawn, particularly on older, less-dense NAND. As TLC and QLC drives age into their own multi-year tenure, the endurance curves will almost certainly look different. Those geometries carry fewer P/E cycles by design, and the statistical floor is lower.

The drive in question is, at this point, a curiosity as much as a benchmark. But curiosities can reframe assumptions — and the assumption that a TBW-exceeded SSD is a failed SSD has needed recalibrating for some time.