Kioxia Puts 1000-Layer NAND On The Roadmap And It's A Big Deal For Storage

Render of stacks layers on a BiCS flash chip.
At times it has felt like meaningful capacity increases in the realm of solid state drives (SSDs) have come at a snail's pace, and it's especially disheartening when you look at how big mechanical hard disk drives (HDDs) have become. Well, take solace in that fact that memory chip makers are hard at working stacking layers to enable more capacious SSDs. Kioxia is one of them, and it thinks reaching 1,000 layers is achievable in the not-too-distant future.

It won't happen next year or even the year after, but according to Kioxia, 3D NAND development is on track to hit 1,000 layers by 2027. That represents a massive increase over what's employed in today's cutting-edge SSDs. For example, Samsung's flagship 990 Pro is built around 176-layer triple-level cell (TLC) NAND flash memory chips.

To be fair, that's now an older model—we posted our Samsung 990 Pro review back in October 2022. Memory chip makers have made headway in stacked density increases since then, with SK hynix last year announcing the mass production of 238-layer 3D NAND. It also demonstrated 321-layer 3D NAND last summer.

We're also starting to see higher stacks in consumer SSDs as the PCIe 5.0 party gets underway. One example is Adata's Legend 970 that we reviewed earlier this year—it uses 232-layer TLC NAND flash memory made by Micron.

Kioxia infographic showing 2D flash memory and 3D flash memory.

Still, on the surface it would seem that achieving 1,000-layer NAND is further out than just three years, but that is what Kioxia claimed at the international Solid-State Circuits Conference (ISSCC), according to Akira Fukuda at PC Watch. Not without technical challenges, there's more to it than just stacking more layers on top of one another. Challenges such as channel resistance and signal noise become issues.

To get around this, Kioxia and other chip makers would need to shift from TLC (3 bits per cell) over to quad-level cell (QLC with 4 bits per cell) or maybe even penta-level cell (PLC with 5 bits per cell).

However Kioxia plans to get there, the end result would be a massive increase in SSD storage capacity. There would also be a cost increase, but perhaps not as much as one might assume. According to Fukida, manufacturing a 20TB SSD (25.6TB if you include spare space) would cost around $120 to $150 to make, and sell for $250 to $350. That latter is roughly in line with what 20TB HDDs go for currently.

We have a hard time envisioning a 20TB SSD hitting that price point in the near future, but we'll have to wait and see. It's also worth pointing out that even if Kioxia and others hit 1,000 layers in 2027, shipping products would undoubtedly come later, and potentially much later. At that point, the stated pricing looks more feasible.
Tags:  Storage, NAND, kioxia