Intel Corp (INTC)

[wbmw]

It's not AMD any more. ATIC thinks that by throwing money at it and driving aggressive schedules and working their employees to the bone, they can accelerate progress.

It's not like there's no precedent for that. The space race is a good example of accelerated progress - but at least in that case there was national pride and everyone's hearts were in it, even though they knew it would be hard.

I think in this case people are generally skeptical. GloFlo employees may be very intelligent and very hard workers, but Intel has had such a large technological head start, and has also retained some of the brightest minds in the industry, and they have a very well tuned process that makes them successful every single generation.

Global Foundries has done absolutely nothing by themselves. It's all purchased existing technology, and they hope to outrun the very partners who helped to get them this far. They also have to overcome 3 process generations over the course of 3 years.

Not even AMD trusts that they'll have 20nm ready in 2014, or else they wouldn't have pushed out their roadmaps and asserted the need for a longer 28nm cycle. Really, if they thought they had an opportunity to have 14nm structures, they would jump at being a lead partner. But even they don't think it's real. Read more here:

http://www.xbitlabs.com/news/other/display/20121213170238_AMD_We_Would_Like_to_Use_Process_Technologies_for_Longer_Period_of_Time.html

GloFlo thinks they can do the following:

- 28nm: shipping (in limited quantities) today
- 20nm: risk starts soon, shipping late 2013
- 14nm-on-20 w/FinFET: risk starts 2013, shipping 2014
- Real-14: skipped
- 10nm w/? (FD-SOI?): risk starts 2015, shipping 2016

I'd say let's see when vendors actually start shipping 20nm, because I think it will have crap yields throughout 2014. And no sooner will they ship it and claim victory, but they think they can start offering FinFETs. That has HKMG fiasco written all over it. And the fact that it isn't a real 14nm process speaks volumes of their executability - but then they claim to have a real-10nm process right afterward.

It's a nice fantasy, but we've seen claims like these before.