The Density Dividend Is Not Evenly Shared
3D NAND sidesteps planar scaling limits by stacking vertically, and layer counts keep climbing generation over generation. For a drive builder the gain lands in two places: cost per gigabyte, and maximum drive capacity — the portfolio's planned 122.88TB top capacity comes directly from higher per-die density. Performance, latency distribution and endurance do not improve in step, and some regress.
Longer Strings, Wider Tails
A longer string means higher path impedance from cell to bit line, which lengthens read settling and adds program-verify iterations. The media side pulls averages back with multi-plane concurrency, wider pages and stronger LDPC, at the cost of a wider tail: 99.99th-percentile read latency is far more sensitive to layer count than the mean. Enterprise firmware therefore cannot budget from average read latency on a spec sheet alone; it must reserve time in the read-retry and soft-decision paths or the tail surfaces under mixed workloads.
Bigger Drives, Longer Rebuild Windows
When capacity grows while the DWPD figure stays fixed, absolute daily writes grow proportionally. QLC is therefore placed where the write pattern suits it: J760Q and J760QD in Q7 serve object storage, backup and archive, while J960Q and J960QD in Q9 serve AI checkpointing and very large capacity arrays. Both classes of workload have low write amplification and good sequentiality, which is what makes a large-capacity endurance budget viable. The other constraint is rebuild: the larger the drive, the longer the array rebuild window after a failure, making sustained sequential read under background pressure a hard requirement.
How This Splits the Portfolio
Media density is treated as a tiered choice rather than a blanket upgrade. TLC serves CORE 7 and TITAN 7/9 where stable random performance matters; QLC concentrates in the high-capacity Q7 and Q9 roles; X3, X7 and X9 invert the trade, spending capacity on endurance and low latency. Both the 122.88TB maximum drive capacity and the 120 DWPD write endurance are portfolio-level planning boundaries at the product-definition/EVT stage, sitting respectively in the capacity tier and the endurance tier rather than being the rating of any single series. Ten series and 99 capacity SKU combinations exist precisely so density, endurance and latency can be selected independently per workload.