Electronics Weekly Electronics Design & Components Tech News
“Now it makes complete sense to dedicate a whole fab to just one customer and hold that – to hold fabs in fact to just one customer, ” Morris Chang, CEO of TSMC, told his Q2 results meeting, ” there are customers that are getting bigger and bigger. So it makes sense that we dedicate a whole fab or even more than a whole fab to just one customer.”
The most likely customers for a dedicated fab are Apple and Qualcomm. Both have hinted at looking for alternative business models for securing supplies, but have kept very quit about their plans.
Apple must be thinking: ‘Our biggest rival is Samsung. Samsung has leading edge fab. We don’t, and problems with 28nm have shown that we can’t guarantee finding a source of leading edge fab.’
Apple stuck with 40nm when it felt it couldn’t rely on getting enough product at 28nm.
Apple also knows it could get massive aid from Uncle Sam if it decided to build its own fab in the USA. The US administration is super-keen to lead the way back to on-shore US manufacturing.
Which may be why Morris Chang, whose interests lie very much in keeping the fab industry off-shore for US companies, is proposing the one company dedicated fab model.
As Intel has demonstrated, it is easier to stay ahead of the game in manufacturing when you only build one product in a fab. Intel only makes x86 processors. TSMC has to make multiple products for hundreds of customers. So dedicated fabs for one company at TSMC could turn out to be running more advanced processes than the fabs for multiple customers.
Which won’t please the multiple customers.
Samsung does have leading edge fab as well as has its foot in almost every market. Apple ought to work overtime with the 28nm, so they wont become the tortoise in this race.
Oh. Was it dear old Fred who came up with that one, Ian? Well these top chaps always seem capable of rationalizing their blunders ex post facto.
I wonder what Fred Shlapak thinks of the asset-light strategy now?
Thanks, Dr D, and Yes with Apple’s profts running at nearly $9billion a quarter it could fund a 450mm fab out of a single quarter’s profit. 0r a 100k wpm Gigafab.
No David, in this case you do happen to be right. Processors up to the A5x ( for iPad3 ) were on 40 nm w/o HKMG. Staying with this process for a quad GPU ( A5x ) to drive a large screen increased heat generation which impacted system design & performance. Samsung kept their 32 nm HKMG process for themselves as long as they could, till they debuted in their Galaxy S3 ( the quad core Exynos SoC in which uses their latest 32 nm HKMG procvess to great advantage : smaller & faster die @ 2 GHz but power consumption 30 % lower than the dual core S2 etc etc. ). But now Apple too is getting access to Samsung’s 32 nm HKMG and have used it for the processor in Apple TV. Expect to see the A6 ( for iPhone 5 ) on the same process soon. But having gotten burned once, it would be very reasonable for Apple to make alternative plans. For all you know TSMC may actually be building a one customer Fab for them even as we speak !
Coming full circle > the whole idea of fabless was that the fab company (TSMC) would be able to optimize utilization by aggregating the demand from multiple users. IF they are going to dedicate to A customer, the logic falls on its face > we have just added a ‘middleman’ to the process. Costs will go up.
I’m fairly sure the latest iPhone4S now use the 32nm version of the A5 processor. Later iPad2s also used the 32nm A5 but I believe the iPad3 currently ships with the 40nm version of the A5. No idea why this is – to use them up maybe ?
Of its two latest big products, iPhone 4S was definitely 40nm and I thought iPad 3 was too, though i may well be wrong.
Apple has been on 32nm for quite some time. The difference between this and 28nm isn’t that huge.