Intel Corp (INTC)

[wbmw]

what was really stopping Intel bringing out 32nm Medfield and CloverTrail in 2010 with their first 32nm product

You can't rewrite history, Mas. Intel outlined their mobile roadmap from Menlow, to Moorestown, to Medfield years ago at IDF. It was a 3 year plan to reduce board size and standby power with each generation, while maintaining flat performance. iPhone was launched in 2007, and this was a reasonable plan to catch up.



Keep in mind that the original iPhone had a 620MHz ARM-11 equivalent processor - something that the Atom core would run circles around many times over. So I don't blame Intel for believing that they could build a leadership 2011 product, based on simply maintaining the performance of an Atom core, and that it would be very unlikely a phone chip could grow in performance by >10x in 3 years. Such a performance growth rate was far beyond any CAGR in recent computer history.

And yet, the phone market has moved at incredible velocity, such that Intel announced their strategy last year of moving the mobile processor line to a yearly cadence with a plan to move it to N generation process.



This was always meant to be a very aggressive roadmap, since it had to have accelerated a number of internal schedules. But since Intel has a history of setting very high goals and beating them, this was the direction PO gave his company. Of course, it seems like some delays have caused a push out of maybe a quarter or two, with Clover Trail to fill the middle, but Intel has a solid history of execution, so I am not expecting any more material delays.

As an investor, I continue to believe that Intel will come back with winning products. But the Romans could not build Rome in a day, and neither can Intel turn around their processor roadmap to the whim of impatient investors. Medfield couldn't be built in 2010 to be lead on 32nm no more than Silvermont can be built in 2012 to be lead on 22nm. These are execution realities, and if investors can't live with the realities of the semiconductor design cycle, then they should probably invest in a stock that can tilt on a daily basis - like Facebook.

As it now happens Intel's mobile designs will not catch up with its process lead until 2016 on 10nm.

As for when Intel can deliver the "process lead", just look at what the industry is having to settle for in process cadence. After 20nm, they get a 20nm-part-2 process that attempts to build 14nm or 16nm FinFETs on top of it. And the date for when we're likely to see SOC's built on this wonderful next-gen process won't be until 2015, a year into Intel's 14nm process. And do you think they will jump from 20nm to 10nm in one fell swoop? No, probably not. It might be a 16nm process with 12nm features, falling farther an farther behind Intel.

So the way I see it, Intel won't have to wait for 10nm to be ahead. They'll be ahead at 14nm because no one else will have a real 14nm process. Intel will be so far ahead, that they can probably afford to wait a year into the 14nm ramp before releasing their phone chip, and it will still be well ahead of any *real* 14nm chip from the foundries.

Once Intel's roadmap pipeline is through a full cycle, the ARM competition will be so far in the dust that everyone will pretend they always knew it would be true. I can't wait for the 20/20 hindsight geniuses to come out of the woodwork and claim that they predicted success all along. It would be bullshit of course. Almost everyone is down on Intel right now, and in the past, that's always been the best time to buy.