Share
3,488 Posts.
lightbulb Created with Sketch. 415
clock Created with Sketch.
27/01/21
22:41
Share
Originally posted by jsbuser:
↑
Reality check. We only have a $2B addressable market on first product release (NOR FLASH). We might address 40% of that market on release since without the selector we are at 1Mb size. Without stacking or multibit or die shrink you cant go further (NAND FLASH and other NVM uses). The Leti selector is the key to the broader NVM market, that is when significant uplift is likely to occur since it will lead to quick scaling. I dont understate Fab announcements etc since that means we are going to market but the selector increases that potential market dramatically. We cant dominate until we can scale, simple as that. Now at the risk of being the perpetual squib to the more exhuberant here and in Weebit management (ie Dadi): 1. they've indicated multibit is possible 2. the layering of each cell is small (2-4 layers vs 8-16 for flash) so in theory more stackable than NAND FLASH, just need to see quantum effects and leakage across layers 3. they've already done tests to see if they can go to the next level 28nm. This also bodes well for 2. since as you push things together you have to adjust things to stop 2. Note most SPI NOR is bigger than 40nm. 4. we know Leti have published widely on selectors, but that is with materials like Gallium Arsenide, not sure with standard CMOS etc. They also have techniques for post-fitting selectors (after cells formed in the process) so that may make the job easier. I am relatively confident on this but would like more validation. It would be great if the selector was massively scalable right away. 5. be good if our first fab was capable of getting shrink happening out of the box. When I invested in WBT, I've been mentally checking off this kind of stuff from day 1. We have the team to give this the best shot possible. I cant see a reason why we cant enter the market. However, it is sustaining a rise through the broader NVM market with a much better product that is more important to overall value. Now, there is a company around that is focused on delivering the best reram cell they can get. That's not Weebit. I believe they will get a better cell than ours. All credit to them. But that isnt what its about for me . Making a better product than the incumbent leader (Flash) capable of being easier to manufacture consistently is Weebit's goal and then getting across as much of the market as fast as possible is the key. Not once have Weebit management set poor expectations on what the process steps to production are. If we look at what WBT has been doing all along, its the other things beyond the cell. It is about building a scalable manufacturable memory. They've got the memory in the design tool already, they have tested the ability to shrink, they had a clear optimisation phase as is typical in the industry, they have been assessing their yields and wafer and cell consistency for manufacture, they've been tackling the scale question with Leti, and they've also made sure the cell is workable and performant and will last. All this is occurring concurrently. This is where the experience of the team in taking a real semiconductor product to a broad market before pays off in spades. You have to do all these things to be competitive and also to have the best chance of succeeding and you have to work always with that end goal in mind. Its what differentiates product engineering from R&D. Of course, we are talking theoretical market here. Someone actually has to want to think the incumbent tech is inadequate for their needs and I am not completely sure that view is universally held. Only real world testing will bear that out.
Expand
js, great stuff, couple of points, I was sure the layers were more, 7? but very happy to be wrong here, the market % for now is correct, the selector I thought we had that locked down solid hence that move, retention has been tested and is waaay up there meaning less dud cells per wafer, with xtc new fab am certain shrink will indeed come right away, I keep banging on about this but unless we can scale fast and stack like mad over the next 2 years we may mis out on that full potential market, am sure that wont happen , need all of this to make cost viability thtas the big risk left for me. another is Intel, with Pat onboard and the latest figs out on the market re Optane I was shocked to see Optane made up 30% of intels money last quarter and growing, so we really need to move or are pat and dadi doing a joint thing?. would you like to won Intels stock now that Pat Gelsinger is the man?.