Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/index.php?threads/will-cost-effective-foundries-be-the-next-battleground.16506/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021370
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Will Cost Effective Foundries be the Next Battleground?

Arthur Hanson

Well-known member
With Samsung and TSM competing on the high end with expensive complex fabs, the demand for low-end and moderate-end chips is not being met. It looks like China maybe set to win the race in producing and lowering the cost of low and mid-range chips that are also critical to a functioning economy. One question I have as an investor is what companies outside China are best set to compete in the mid and low-end chip market in design and production. Does China have the will and resources to control the non-cutting edge chip market in the future through total lower costs in design and production? Any thoughts or comments appreciated.

 
TSMC and Samsung provide lots of trailing edge capacity, they have some pricing advantage over smaller players like UMC. If UMC wants to build a 16nm fab in a few years they will have to build one and do the R&D for this new node. Ti has a similar issue, as it will have to spend a large sum of money to convert the Leti fab to it's process. TSMC and Samsung have the benefit that as their leading edge nodes become trailing edge they can just lower the price on the node. By that time the equipment has payed for itself, the R&D money has been made back, and the equipment has depreciated. For these reasons the leading edge fabs can just expand leading edge capacity while lowering the prices on previous gens to attract new customers.

As for non standard nodes that might make a big splash, I would keep your eyes on GF 12/22 FDX, intel 16, and samsung 17nm. These nodes have impressive power draw, leakage, and pricing for their performance.
 
The demand for legacy node service is bound to grow for simple reason that 180nm-200mm entry level service is booked for a few years into the future.

130-180 game on 300mm will be a very interesting thing to watch
 
Last edited:
Ti has a similar issue, as it will have to spend a large sum of money to convert the Leti fab to it's process. TSMC and Samsung have the benefit that as their leading edge nodes become trailing edge they can just lower the price on the node. By that time the equipment has payed for itself, the R&D money has been made back, and the equipment has depreciated. For these reasons the leading edge fabs can just expand leading edge capacity while lowering the prices on previous gens to attract new customers.

Um.... TI bought the Lehi fab from Micron for $900 million for the fab and $600 million for the equipment. And offered existing Lehi fab employees the opportunity to continue under TI ownership. TI has stated that it plans to use this fab for "65-nm and 45-nm production for TI’s analog and embedded processing products and be able to go beyond those nodes as required." Considering that Micron estimated costs of building the plant in 1995 at $2.5 billion, and had been upgrading equipment during its IM Flash run from 2006 - 2021 as technology node sizes came down, it's probably a steal for TI. Yeah, presumably they'll have to spend money to get it the way they want, but at a fraction of the cost of a new plant. (RFAB2 estimated at $3.1 billion.)
 
Um.... TI bought the Lehi fab from Micron for $900 million for the fab and $600 million for the equipment. And offered existing Lehi fab employees the opportunity to continue under TI ownership. TI has stated that it plans to use this fab for "65-nm and 45-nm production for TI’s analog and embedded processing products and be able to go beyond those nodes as required." Considering that Micron estimated costs of building the plant in 1995 at $2.5 billion, and had been upgrading equipment during its IM Flash run from 2006 - 2021 as technology node sizes came down, it's probably a steal for TI. Yeah, presumably they'll have to spend money to get it the way they want, but at a fraction of the cost of a new plant. (RFAB2 estimated at $3.1 billion.)
Oh yeah it is a great deal for TI. My point was that even with this favorable example they still need to spend a large amount of capital to switch over to analog as well as the price of buying the fab, whereas a samsung or TSMC don't. All they do is lower the prices, and because everything has already been payed for and they can make crazy margins on fabs that basically don't cost them anything to operate (relatively speaking). Of course the semi shortages forced their hand, and for the first time in a long while TSMC (as well as many others) had to add capacity to their legacy process nodes. Now that these shortages are easing, I presume the old model of not retooling older fabs and lowering the prices of the nodes they produce will continue to put immense pressure on smaller foundries like UMC GF and Onsemi. After all the majority of TSMC's money does come from their old planar nodes.
 
Back
Top