Rink,
The Tosh and Samsung chips that you mentioned the data densities for, they are single die solutions, right? Samsung is currently at 73nm for existing products and with 8Gb as max density for single die so I'm presuming the 16Gb is <73nm single die and announced for the near future.
Those Toshiba and Samsung chips are not current products, but announcement for future products on 50 and 55nm process technologies.
When Saifun didn't believe that 4b/cell was possible for NAND (M-Systems' claim) it did say that if it would turn out to be possible they would develop 8b/cell NROM. I don't know if this is related but guess it is.
From looking at the slides that went along with the analyst conference, it appears that the Double Density they were talking about was not about 8 bits per cell, but about a way to pack the cells more closely together.
As far as I know, Spansion has not announced that they would use Double Density technology, but it may only be a question of time before Spansion makes such an announcement.
Seifun's goal is to sign up a NAND Tier-1 partner for their NROM, but they did not seem to show much confidence that it is imminent. Infineon was apparently their ticket, so it was a big setback. Seifun seems to think that the Tier-1 NAND players are at an inflection point, and believes the only way forward is to use Seifun technology. Anyway, Seifun didn't give any hint that they are working with some other NAND player, so maybe the way they think they can "acquire" new licensees is through litigation.
From the point of view of Spansion, it would be better it Saifun does not sign new players, who could be competition for Spansion, and also, maybe Spansion could buy Saifun, if the price is not too high.
Anyway, what seemed exciting to me, from the call was the market that is apparently growing very fast (in high 40s percentage growth rate), which is the market for combination of code and data, and that appears to be to be a huge market. Spansion's goal is to go after this market, and avoid for now, the cutthroat commodity part of the NAND market.
So Spansion is taking some share in the stagnant NOR market, which makes Spansion grow faster than NOR market, but the big opportunity is the Code + Data segment.
The advantages Spansion says they have in this Code + Data market (which is mostly mostly multichip packaging) is tha they offer combination of ORNAND, NOR and RAM (SRAM, etc) where each component is doing job that it is suited the best for, and only as much (as little) of it as needed. The problem of the NAND - RAM combo apparently is the cost, size and power consumption of the RAM portion. With NAND, much more of the memory needs to be shadowed by RAM, which means bigger RAM chip is needed. The RAM chips are apparently fairly expensive for low voltage operation, and use too much power in standard voltage operation.
Anyway, the way it looks (this is my comment) is that pretty much all of the wireless phone handset market is going to be moving to this model (of code + data) eventually, which is an opportunity as well as risk for Spansion. But with the amount of optimism Spansion is expressing about the wireless phone market, it seems to be mainly opportunity.
The capacity constraints suck, especially at this particular time of the opportunity. The good news is 75% of 90nm wafer starts in Fab 25, but I wonder how much of the 90nm stuff will contribute to Q4. The cycle time for flash must be much shorter that almost a quarter (13 weeks) for CPU, but still, there may not be enough time to sell these in Q4. It will probably be only the initial wafer starts (before Spansion turned the dial hard to 90nm) that will contribute. The bulk of the 90nm transition contribution will be in Q1.
Another big positive, in my mind is that Spansion already has some 65nm wafers (200 and 300mm) it is evaluating, and that 65nm production will be moving to Texas fab. I was a little worried here, if Fab 25 would need some hugely expensive conversion (like Dresden Fab 30). It appears that Spansion will be able to continue productin / processing of 200mm wafers down to at least 65 nm, which extends life of that Fab, and delays some expensive conversion much further down the line.
With Fab 25 at 65nm, Spansion will be able to ramp 300mm SP1, move it to 45nm, making SP1 the leading fab. Than, optionally, with SP1 fully ramped, Spansion could do Fab 25 conversion. This may be roughly 2009 (my guess).
Joe |
