Micron has apparently decided that ten is a number for other people. The company's next generation of NAND flash will be called G11, skipping G10 entirely, and it stacks more than 400 layers of memory cells while packing twice the density of what came before, according to a report from StorageReview carried by Google News. The same report puts a 512-terabyte ceiling on the solid-state drives the new chips could enable and attaches a second-half 2027 date to the plan.
Let me unpack the jargon before it piles up, because I have personally watched a friend's eyes glaze over at the word NAND. It rhymes with hand, and it is the kind of memory inside your phone, your laptop's SSD, and the storage drives in every data center feeding an AI model. The name comes from the logic gate the cells are built around, which is a fun fact for exactly one type of party.
The layer count is the part engineers care about most. Early flash chips laid memory cells flat, like houses on a single street. When the street ran out of room, the industry started building up instead of out, stacking cells vertically the way a parking garage stacks cars. More floors means more cars on the same patch of land. Micron's G11, per StorageReview, goes past 400 floors.
Density is the payoff. If a chip holds twice as many bits in the same silicon area, the cost per bit has room to fall, and the drives built from those chips can grow enormous without growing physically. A 512-terabyte SSD would hold roughly half a million gigabytes in a single device. That matters for the AI training clusters this magazine spends most of its time on, where the appetite for fast, dense storage sits right behind the appetite for GPUs and power.
The skipped number is the detail I keep coming back to. StorageReview's headline frames the jump as G11 rather than G10, which suggests Micron is signaling a bigger step than a routine refresh. Why the company chose to leap a generation in name, and whether that reflects a change in architecture or simply in marketing, is unknown at this point.
The business stakes here are straightforward even if the physics is not. NAND is a brutally competitive commodity market, and Micron sells into it against the likes of Samsung, SK hynix, Kioxia, and SanDisk. Each generation is a bet that a company can hit a new layer count on schedule, at acceptable manufacturing yields, before the rivals do. Winning that race tends to mean better margins for a while. Missing it tends to mean selling last year's chips at this year's prices.
What the second-half 2027 date refers to is the open question. It could mark when the chips sample to customers, when they reach volume production, or when the first 512-terabyte drives appear, and the report as carried by Google News does not settle that for me. Nor is there any confirmation yet from Micron itself on pricing, customers, or how the G11 chips will perform on read and write speed, which is the number that decides whether a dense drive is also a fast one.
So file this one under promising and provisional. A 400-plus layer NAND with double the density is the kind of roadmap item that quietly determines how much data an AI data center can keep close at hand two years from now. The gap between a roadmap headline and a drive on a rack is where the real story will play out, and I will be watching for Micron's own words on it.
Leave a Reply