[fool.com] - Broadcom Corporation Voices FinFET Cost Concerns

[ikachu]

Thought this article was pretty interesting.

http://www.fool.com/investing/general/2014/05/29/broadcom-voices-finfet-cost-concerns.aspx

 

[Homeles]

Intel claims FinFETs add only 2-3% to the cost.

The main cost adder is definitely going to be the higher patterning costs... not the FinFETs, even if they don't manage Intel's "2-3%."

 

[witeken]

This confirms ARM's slide:

I don't it's FinFET, but just rising wafer costs and probably (much?) worse yields without an increase in density.

 

[Wall Street]

Don't forget that the per transistor cost is based on supply and demand. Usually, TSMC has a line out the door to use the newest, smallest and highest performance process. So naturally, the people who are willing to pay the most will be willing to pay up for the few wafers that are available.
Even if yields are great and wafers are really cheap to make, if TSMC has demand that is five times the number of wafer than they can supply, they would be idiots to not just give those wafers to the highest 20% of bidders. Of course they need to protect their relationships, but the supply demand imbalance of any new process does a lot to drive the pricing. There is a long history of cost sensitive buyers waiting to transition to a new process.

 

[Idontcare]

Its a number of things.

Supply and demand on TSMC's part is definitely going to drive wafer contract prices in their favor.

At the same time, one of the key care-abouts captured by a node shrink development team is to engineer for lower cost of manufacturing.

But, since the beginning of the industry itself there has always been the opportunity for management and engineer alike to "bungle" their part of the development of a node and find themselves with a node shrink that costs way more than it really should when it goes to production.

Intel had Finfets in R&D for a long time before moving them to production, and for a good economic reason.

Now that everyone else (except ST ) wants to jump on the Finfet bandwagon, they are naturally going to find the learning curve to be rather steep and fraught with all manner of cost-adders if they don't appropriately and responsibly manage the way they build cost into (or out of) their new finfet-enabled nodes during the development of the node.

TSMC may be taking advantage of the supply/demand situation created with their domination of the 28nm market, or they may have bungled things in development and built themselves an Spruce Goose of a finfet node.

Either way, blaming the technology (finfet) for what are essentially business decisions and business management (whether it be either case or a mix of both) is kind of a lousy excuse.