Quite a few people think that NAND flash is about to fall off the edge of the earth, and that some new technology is just about to take its place. Here's Jim Handy's take on that thought.
Quite a few people think that NAND flash is about to fall off the edge of the earth and that some new technology is just about to take its place. I was surprised to participate in an IBM/Texas Memory Systems teleconference about a new flash storage system and hear a number of storage system experts try to steer the conversation to IBM’s plans for next-generation memory technologies. It seemed these folks thought that NAND flash would soon be replaced by some alternative memory type.
Let me share what I presented to the HotChips conference last July. Flash will be mainstream for about another ten years. The business is just now entering what SanDisk calls the “1Ynm” generation, and after this comes the “1Znm” generation. After that, SanDisk thinks that planar NAND will run out of steam, and that we will go to 3D. Other leading NAND makers have expressed similar expectations, but given the way this industry works, it wouldn’t be at all surprising for them to find a way to push out another planar generation after 1Znm. If you give each generation a two-year lifespan that takes you out four to six years.
After that we will have 3D NAND. I wrote an extensive series on this technology on my blog The Memory Guy. Yes, I know that Samsung announced mass production of its V-NAND last August, but the scarcity of the product indicates that it isn’t ready to take market share away from planar NAND. My guess is that 3D will not really ship in volume until 2017. Today, most companies expect 3D to run for three generations before it, too, runs out of steam. At two years per generation that would add another six years onto the four to six years in the paragraph above. That gives us ten to twelve years before the market will need to switch to an alternative to force prices down.
It’s not that there’s anything wrong with these alternative technologies. It’s just that insufficient effort has gone into them to allow them to become cheaper than NAND, and until NAND reaches its scaling limit, there’s no motivation to change that fact. Whatever is cheapest always ships in the highest volumes, and NAND flash and DRAM are significantly cheaper than any other memory technology.
Next week, I’ll look at SSDs.
Jim Handy is a semiconductor industry analyst and the director at Objective Analysis.