Folks are always coy about what their product WILL NOT DO, so i hope this si correct, but IMO the the seismic m.2 pcie3 ssd market, seems a winner take all~ situation presently.
They are designed as a 4 lane device/interface - i.e., the device connection has a max bandwidth of 4GBps. Present best ssdS (samsung methinks) can max out on reads at about 2600 - 3200MBps, depending - so already this 4GB limit is being approached by nand.
The device can run using only 2 lanes, but suffers disproportionately it seems, and only rates abour 1GBps. It seems so, but is unclear.
So if a user must use a 4 scarce lanes for a device anyway, its presently a false economy to get cheaper slower ssdS. May as well have the best, and extract maximum value from the allocated lanes.
Further, the killer app for them is in modern server farms, as a vast ram substitute for some situations, which uses a fraction of the power when idling (NV).
Raid 0 four pci nvme ssdS on a ~standard 16 lane prosumer controller card like Dells or highpointS, for a theoretical max of 16GBps, and you have respectable ram like speed for a server, using some clever and discerning memory management software.
There is excessive noise about Intel and NAND and Intel likes it as it makes some investors dream.
However, if you look at revenue , in Q2 last year they had the NVM segment at 554 million while in Q1 (latest reported quarter) it was 866 million. Seems substantial but a large chunk of that is the NAND pricing recovery.
In comparison Micron in equivalent quarters (their fiscal quarter is 1 month off) went from about 900 million to about 1.4 billion for the NV segment, quite similar growth in %..
Intel selling wafers, how much can they afford to sell given the long term margins for NAND. Maybe they can reach 30-35% gross margins with current pricing but the industry is cyclical and pricing won't hold. Ofc Intel itself could push it into oversupply if they ramp too hard. Intel's overall margins are at some 63% with strong downside risks so they won't want too much revenue at 15-20% margins.
I wonder if Intel going at it on its own is not about 3D NAND requiring more clean room space and Intel having too many fabs. Maybe for the JV to meet their needs, they would have needed to expand it to more facilities so why do that when you have more clean room space than you need.
Intel does not talk about bit growth and that's suspicious and likely aimed at letting some investors dream big. They do have SSDs with 3rd party controllers in consumer, like the 600p- the worst in its category (PCIe 3.0 x4).
All in all it doesn't seem likely for Intel to sell much if any wafers, unless their SSDs underperform and they have excess captive NAND supply.
Yangtze aims to sample 64 layers in 2019, if they can break even with that and ramp in 2020 will depend on the state of the market at the time.
I don't dispute the opportunity for selling 3D NAND standalone. The iPhone 7 had the Toshiba 48L BICS. With YMTC and Intel entering standalone 3D NAND separately from the other 4 (Samsung, SKHynix, Toshiba/WD, Micron), that's a lot of players.
@resistion: So my understanding is that Intel will not sell standalone 3D NAND,
I'd be startled if that were the case. Semiconductor fabs are enormously expensive to build and operate. You make money by running 24/7, turning out product customers want to buy. 3D NAND is a premium product Intel can charge a higher price for. They are targetting next-gen SSDs as the primary use case. But if you have a use case, want to place a large order, and will pay what Intel wants, I'm sure they'll be happy to sell to you.
The issue will be demand and capacity. If the product takes off as Intel hopes, you may have to wait in line if demand exceeds production, but I doubt you'll be turned away.
The interesting question is what you might do with 3D NAND flash that isn't an SSD.
I don't think so. With actual Dalian Fab fully on line, they are plenty of capacity to supply customers with standalone chips. Still the real news is that, looking from the space, Intel is doubling the Fab with a new module, likely online at the end of 2018. This is congruent with some Intel statemants about a large capex dedicated to NAND in 2017 and 2018.