Holy Grail Predictions:
1) There will be no L'Intel PCRAM "samples" in H1... H2... or any other H's;
2) Neither PCRAM nor MRAM are anywhere near ready for profitable prime-time;
3) Odds of OUMazing PCRAM or MRAM ever arriving at said prime-time are not improving fast enough;
4) The real winner of The Next Best Holy Grail Award will not be decided until the industry reaches consensus on both the necessity and identity of a new and improved process for logic devices beyond the current CMOS IP;
5) CMOS replacement for logic processes will not get serious for another 5 years at the very least and, if it ever dies at all, it may well last to 2020. The technology landscape in existence by the time either logic or memory is ready to move off CMOS will probably be wholly unrecognizable;
6) None of the current Advanced NV-Memory technologies can seamlessly slip into the unknown post CMOS environment and they will only make sense as current products if they can compete on cost/performance in the twilight years of high cost CMOS process development;
7) There will never be a sufficient CMOS market for high density PCRAM with 10^8 cycle endurance to justify its R&D expense in a world which has already leveraged its grandchildren with the massive R&D/development costs of advanced CMOS FLASH... and which could easily be dropping its CMOS foundations by 2020;
8) The financial markets will have no reason to arrive at these same conclusions before CMOS is nearly dead... unless and until FRAM development produces a massive and entirely new market "niche" in which some "Killer Application" aids in pushing RMTR revenues to "shocking"... and competition drawing... levels;
9) Neither L'Intel, MOT, IBM, ST, Samsung, nor anyone else will be able to afford to pay for bleeding edge NOR/NAND development and FRAM, PCRAM, MRAM or any other _RAM at the same time. The industry will see what Wall Street won't. No one is going to fire up a $4B PCRAM or MRAM fab on spec; and
10) The industry has to either (1) have a high degree of certainty that FLASH will die within 5 years; or (2) the industry will need equivalent certainty that a new market exists for PCRAM/MRAM which the old FLASH IP, and its already obligated costs, cannot serve. Without one of those two "certainties" only a truly desperate and/or ignorant manager would accept the investment/return odds on production of an entirely new semiconductor IP.
So I predict the PCRAM/MRAM industry supporters are going to hold their water for now. They will watch the CMOS costs rise and they will watch FRAM and its "niche." If RMTR proves you can grow a worthwhile CMOS market for non FLASH or EEPROM sockets, some of them may well flip the switch on competing PCRAM/MRAM production. But at this point FLASH is no more likely to die in 5 years, than CMOS itself.
And at that point the primary question is no longer which NV-memory IP is CMOS friendly. The question will then be what is the next logic process going to look like, and what NV-memory IP can co-exist in its fabs.
We know that PZT based FRAM can do CMOS profitably enough (on a non-GAAP basis:) and we also know that PZT as a material can play a variety of roles in an digital-analogue-organo-opto-electric-nanotubed world. If you have to place a bet, it's as safe a place as any to do it.
0|0 |
