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!
Hi all - I’ve been wondering about this fir a while: we all know that those cutting edge processors need advanced manufacturing nodes while power and other analog chips tend to require mature bodes at 28nm or older. but how does one, especially those semiconductor industry analysts, know which products will use which manufacturing nodes?
They don't. Analysts aren't privy to confidential data, so they guess.
The designer of the product makes the decision early in their design process. They need to know their speed/power requirements, die area, foundry capacity, pricing, market window, etc.
With TSMC it is hard to tell. For example, Apple get's a custom version but revenue is recognized in buckets. For example TSMC N7, N7+, and N6 are all one bucket. Same with N5 and N4. Customer node designations is covered under NDAs so it is usually not publicly discussed, especially by TSMC. The media however speculates and that speculation turns into "fact" after a while. The semiconductor insiders who work in the trenches know of course as do the EDA and IP vendors, but again, there are NDAs.