Here comes the home stretch.
1x … 1y … 1z
End of the alphabet.
1x or 19nm will be in full swing next year in 2012.
1y, around 15nm, is scheduled for 2013.
1z, likely the last generation of NAND, could make its appearance 2014-ish.
It’s looking ever more likely that NAND will will go 3D, leveraging the existing infrastructure. But that’s another post.
This post is going to briefly review the end of the world as we know it- the last three generations of 2D NAND.
It’s worth noting that SanDisk and Toshiba have pulled in their endgame schedule.
Implying that they have figured something out. Both for their NAND technology roadmap and the thereafter.
This should be all good news. But they still have to execute.
1x, 1y, 1z
Last year SanDisk announced that 1x was was going to be a late 2012 story.
The graphic below- SanDisk’s NAND Roadmap- is from the SanDisk Investor Day 2010.
Transitions roll right along 43nm ⇒ 32nm ⇒ 24nm, then there is a year+ gap, from late 2011 to late 2012, before 1x and then nothing thereafter.
1x has been pulled in almost a year: from the second half of 2012 to late 2011. 1y which is nowhere to be seen in the 2010 graphic, now appears mid 2013.
If this weren’t enough, in May, Toshiba noted that mass production of 1x would start even earlier- this month in July.
Up until May the SanDisk/Toshiba party line had been that Fab 5 would start with “leading-edge 20-nanometer generation, with subsequent generations to follow.” Translation: 24 nanometer.
This appears to have changed according to Morgan Stanley in their Toshiba report of 25 May:
“TOSHIBA shows confidence in its competitiveness, with ~70% of the OP growth targeted in its new business targets for F3/11-3/14 coming from electronic devices, and the expectation of huge demand for NAND ahead. In this respect, we are positive on comments that 19nm mass production will start in July, ahead of competitors, and that process development for 1Ynm (the generation after 1xnm [i.e., 19nm]) and 1Znm (the generation after 1Ynm) is on track. (This is consistent with SanDisk’s comment that floating-gate NAND can be extended to at least 1Ynm, using existing immersion lithography.) Assuming continued growth for smartphone & tablet markets, we can expect a big jump for NAND margin in Semiconductors: NAND 19nm mass production timing was updated from ‘the Sep Q’ previously to July. Mass production starting at Yokkaichi Y5 in July and shipments starting in August.”
It sure sounds like Fab 5, Yokkaichi Y5, will be starting up in July with mass production at 1x, 19nm.
It makes eminent sense that Fab 5 start with 1x, and has for a while now.
It is also encouraging that Morgan Stanley reports that the process development for 1y and 1z are “on track,” using “existing immersion lithography.”
It wasn’t that long ago that such scales were only considered feasible using extreme ultraviolet lithography, or EUV.
SanDisk/Toshiba appear to have figured out how to stretch the immersion lithography envelope all the way to the end of the alphabet. Likely through multiple patterning and various lithography techniques with a few other tricks thrown in.
These other tricks no doubt include adaptive flash management (AFM) and error correction capability (StrongECC). Both AFM and StrongECC are proprietary and patent protected.
AFM and StrongECC are purportedly behind SanDisk’s success with three bit per cell memory (X3). To date, the competition doesn’t appear to have an answer.
As NAND moves below 20nm, performance and endurance will continue to deteriorate.
NAND producers could soon face some of the same challenges in scaling MLC as they face today in delivering viable X3.
This last dance could get interesting.
In any case, if SanDisk can deliver, the payoff will be in product gross margins.
Going from 32 nanometer to 24 nanometer has given SanDisk a chip cost reduction of 30% to 40%. Going from 24nm to 19nm should be a tad less, but nonetheless significant.
Something to look forward to in the first half of 2012, when 19nm yields are likely to become mature.