Up until the latest crop of Apple (AAPL 0.15%) devices, Apple had its custom-designed A-series processors built at Samsung (NASDAQOTH: SSNLF). This year, though, Chipworks confirmed that the A8 processor -- found inside the iPhone 6 and iPhone 6 Plus -- is being built at Taiwan Semiconductor (TSM 2.40%).
Although TSMC won this round, there has been much speculation that Apple would be moving back to its longtime frenemy, Samsung, for the manufacture of the A9/A9X chips that will presumably power next year's iDevices. However, reports from Digitimes, SemiWiki, and others have suggested that this decision has yet to be made and/or a dual-sourcing arrangement is likely to be in place.
I would now like to present evidence that strongly suggests that Apple is designing its next-generation processor(s) at both Samsung and Taiwan Semiconductor.
Hidden in plain sight
A quick search on LinkedIn for "Apple 14nm" reveals the following profile of a former Apple engineer who claims to have worked on designing both "14nm" and "16nm" SRAMs, presumably to be used in next-generation A-series processors:
Source: LinkedIn.
This profile strongly suggests that Apple spent the effort developing chips for both TSMC's 16nm manufacturing technology and Samsung's 14nm technology.
What does this mean for Apple's plans next year?
Given that Apple seems to have designed its next-generation chips on both Samsung's and TSMC's processes, the question investors probably care about is: "Well, what does this actually mean?"
Given that some of the chip equipment vendors have signaled that the semiconductor foundries are facing "yield challenges" with their 14/16nm processes, it doesn't seem clear yet what direction Apple is going to go in. I continue to believe that Apple will ultimately go with the company that can offer the most aggressive pricing on a per-good-die basis.
Now, since semiconductor foundries don't generally sell wafers on a good die basis, the question really comes down to ...
Yields!
The cost of a chip to a fabless customer like Apple is essentially a function of the cost of the wafer and the number of good chips that can be had from said wafer. So, as an example, if it costs Apple $3,000 for a wafer, and it can get 600 good chips from that wafer, the effective "cost" to Apple per chip is $5.
Assuming that Apple's next-generation iPhone does use either a 14-nanometer or 16-nanometer chip, I think the question will come down to how many good chips Apple can get per wafer from each manufacturer, and what prices each manufacturer is willing to sell those wafers at. Digitimes reports that Samsung is offering "low quotes" for 14-nanometer wafers, but at the same time notes that Samsung's 14-nanometer yields are still challenging.
I believe that the company that can ultimately offer the lowest price per chip in this case will be the one to win most, if not all, of Apple's A9 business.
Samsung now has everything to gain; TSMC now vulnerable
Taiwan Semiconductor's financial results have been quite good, with the company reporting very robust year-over-year growth. This growth is likely in no small part due to TSMC getting the A8 and presumably the A8X business from Apple.
However, if TSMC ends up in a dual-sourcing arrangement next year or outright loses the Apple contract, then this could mean that the high 2014 revenue numbers create a difficult comparison point for TSMC's 2015 revenues. Samsung, on the other hand, stands to benefit significantly if it can deliver as promised with its 14-nanometer process.
We'll know in due time which of these chip manufacturing giants ultimately puts its big-boy pants on and signs on the dotted line.
