You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
The Citi analyst has put TSMC a year ahead of Intel in the foundry business. Much of this is attributable to TSM being a foundry only and the ecosystem it has developed. This especially is true when it comes to the latest nodes. It will be interesting to see what actually happens to the Intel foundry business and if their management can successfully pull it off, although I have no doubt TSM will have the lead for a long time.
Only one year? According to my sources TSMC will lead the industry for the rest of the decade. Beyond that who knows but if I had to bet it would be on TSMC. I'm writing up my summary of the TSMC OIP right now.... There was some VERY interesting chatter in the halls, wow.
I don't see Intel staying in the foundry business for much longer. Not only are they behind TSMC and Samsung they are more expensive and less flexible. There is also the IDM Foundry vs Pure Play Foundry dilemma customers have to deal with.
TSMC has 10nm in HVM now. Intel 10nm is a year from now and Intel Foundry 10nm customers are probably 2 years from now which means they will be competing with multiple flavors of TSMC 7nm.
Dan, the post said TSM is one year ahead and indicated it will probably stay that way, which would eventually cause Intel to exit the foundry business or become a bit player, if at all. Bottom line, Intel doesn't have the broad and deep ecosystem TSM has.
Dan, the post said TSM is one year ahead and indicated it will probably stay that way, which would eventually cause Intel to exit the foundry business or become a bit player, if at all. Bottom line, Intel doesn't have the broad and deep ecosystem TSM has.
Intel's foundry business is still quite small after several years' effort. If Intel is smart enough, they should utilize TSMC's capability to strengthen Intel's own product offerings. It will save Intel own resource, time, and money in order to catch market opportunities. So far, I think Intel knows this very well and they are doing more or less on this direction. The question is how much Intel's own ego out there to stop them doing more collaboration with TSMC.
Let me throw in some speculations. The newly announced ARM/Intel licensing deal on the surface is for Intel's custom foundry business. But if Intel decided, it can be one of the first step for Intel to diversify their core product architectures and to venture into non x86 market. Who say "Intel Inside" must be "x86 Inside"?
When we talk about who will be first to 10nm, we should also consider the yields. TSMC will debut their 10nm with much better yields than Intel. Intel are not initially producing their mainstream CPUs on 10nm. Instead they are introducing a new family called Coffee Lake, which will be on the old 14nm node. Only, the smaller Y-series and Core M CPUs are on 10nm. The reason why is probably because 10nm yields won't be too good at the start and it's better to start with SoCs with smaller die area at smaller volume for a process with low yield. Intel 14nm Coffee Lake and 10nm Cannonlake To Coexist in 2018
