A key ingredient for GF success will be how they can address linearity of revenue and steady growth given that they (just like TSMC) do not have a built in internal demand like Samsung does. I also believe that it is not just about constantly shrinking technology nodes but it is also about enriching and enabling IP into mundane nodes that have niche applications with good margins like MEMS, high voltage power management, LED, video sensors, TSV even solar panels. This is specially tricky to pull off (by anyone who is reshaping and leapfrogging and merging) since you still have to deliver on the leading edge sub-100nm manufacturability and competitiveness for bread and butter revenue. A slowdown could also put further pressures on Capex needed to deliver. Mark Lapedus just posted this article re: TSMC slowdown..http://www.eetimes.com/electronics-news/4208893/Analyst--TSMC-faces-slowdown-semiconductor
I think to really answer GloFo vs TSMC question is to know the yields at 28nm. Any data available? GloFo is getting a lot of press, which is understandable as this is something new in a foundry arena. But TSMC is a leader to beat. If I were them I would be more concern with Samsung though...Kris
History shows that defeating a dominant source of profits with an "I have more money to loose in the bank strategy" has never worked unless you go openly into price dumping (WTO Banned!).
TSMC is not going to be threatened at all by GloFo, ultimately this array of money loosing Fabs with incompatible processes, incoherent strategy and global scale complexity will come crashing into the ground amidst a feast of taxpayers paid jobs rescue packages.
TSMC will be only challenged if Intel decides to enter the Foundry business, which may happen if Intel moves to acquire ARM. This move would allow the creation of a profitable colossus that can challenge TSMC at its core through a cluster of profitable Fabs for hire in NM, AZ and UT.
Chances of this happening?....gradually increasing.
Of course, we all agree with comments that GF has a deep-pocketed owner in ATIC. Moreover, Samsung is also deep-pocketed and has a record of aggressive strategy and investments and – it is also a member of Common Platform based on IBM technology.
Without wishing to dominate this fascinating and well-timed discussion – thanks to Mark and EETimes, I would like to restate that I firmly believe that the key ingredient here is IBM technology rather than money. I am also confident that IBM had discussion with ATIC and UAEs long before GF was formed.
The renown Research Division is IBM’s innovation machine and it is worth studying by other companies and even by entire countries. We have identified its six unique capabilities – each worth an extensive discussion in itself:
- Built-in culture of automation
- Focus on productivity
- Innovation-for-manufacturing culture
- Order-of-magnitude improvement screening criteria
- Raising levels of abstraction mentality
- Stability of the research collective-IQ
IBM’s corporate culture is to build tools for automation of processes and to assist these tools in execution of target processes with additional tools. A major implication of such an approach to automation is an ever-increasing level of productivity; the cumulative effect on productivity is very high. The results of this focus are also self-evident in IBM’s Research—an organization that consistently outperforms all their global counterparts
GF has the financial backing of a deep-pocketed owner (ATIC). I bet they will be as patient as it can get, so GF should take full advantage of this fact by adopting an agressive strategy in leading edge processes.
Goafrit - the article really discusses 28 and 22 nodes - and beyond. The battle for 40 node and older is won and old coffee.
In 28nm and finer we will have a different foundry industry and - different players. And behind GF and Samsung are wordclass and unique research capabilities of IBM.
The battle is shifting to HK/MG and in 14nm beyond silicon. TSMC is a superb and formidable manufacturing company - it is not known for its strngth in basic research in material sciences --IBM and Intel are.
In summary, the success of both GF and Samsung in Advanced Nodes is largely dependent on IBM IP and consulting services ("Common Platform"). Having analyzed IBM research and MO -- IBM is ahead of the industry at least five to seven years; IBM has moved beyond silicon alone at least 5-6 years ago. Please see:
Blog Make a Frequency Plan Tom Burke 17 comments When designing a printed circuit board, you should develop a frequency plan, something that can be easily overlooked. A frequency plan should be one of your first steps ...