Apple (AAPL -1.22%) has, for quite a while now, designed its own applications processors for its iDevices. Samsung (NASDAQOTH: SSNLF), a longtime supplier and competitor to Apple, has been the sole manufacturer of these chips.
But it is widely rumored that Apple will be moving its chip business to independent semiconductor foundry Taiwan Semiconductor (TSM -3.45%) for the iPhone 6.
That said, concerns about whether TSMC will be able to capture the chip business for Apple's follow-on to the iPhone 6 surfaced last month. TSMC's management remarked on its July earnings call that it would not begin mass production on its next-generation 16-nanometer manufacturing technology until late 2015. This would be too late for the successor to the iPhone 6, which is likely to launch in the fall of 2015.
Digitimes now reports, though, that TSMC "will advance volume production on its 16nm process to the first quarter of 2015 with monthly output of 50,000 wafers in order to meet demand for Apple's A9 processors."
A question that may be on investors' minds, then, is whether TSMC can actually achieve this rumored production schedule, and if so, what doing so might mean from a financial perspective.
What appears to be going on
Samsung and Taiwan Semiconductor are moving to a fundamentally new transistor structure known as a FinFET for their respective 14/16-nanometer technologies. This transistor type, according to Intel (INTC -2.40%) -- which has been successfully mass-producing FinFET-based chips since late 2011 -- offers significant performance improvements and power reductions over more traditional "planar" transistors.
The light gray line represents the performance of a hypothetical 22-nanometer "planar" transistor. The black line represents the performance Intel's 32-nanometer transistor, and the blue one the performance of Intel's 22-nanometer FinFET transistor. Source: Intel
Unfortunately, these FinFETs seem to be proving very difficult for the foundries to bring into high volume production.
KLA-Tencor (KLAC -2.49%), a major semiconductor equipment vendor, offered investors some insight into this issue on its April earnings call.
Now, in logic and foundry, with the introduction of the new 3D gate architectures, the yield issues our customers are grappling with today are proving the most challenging that the industry have ever faced, and even the smallest variation and process margin can cause significant yield losses for these devices.[Emphasis mine]
Now, notice that KLA-Tencor's CEO cited "significant yield losses," which means that when a chip wafer is produced, a large portion of the chips that come out aren't suitable for sale.
How can TSMC just "advance" production?
The reason that good yields are so important is that yields directly affect cost. If a chip wafer costs $3,000 to make and from that one can get 300 good chips, the cost to produce those chips is $10 each. However, if due to yield losses, one can only get 150 good chips, then the cost per chip balloons to $20.
Now, according to an EETimes piece, TSMC shifted to charging customers on a per-wafer basis rather than a per-good-chip basis over the last several years.
Given this context, what could "advancing production" of 16-nanometer, particularly if yields are proving challenging, mean economically for both Apple and TSMC?
Got yield?
To provide some background, according to a 2009 wafer agreement between IDT and TSMC published on the SEC's website, TSMC and IDT agree to a "minimum" or "standard" yield. This standard yield is "based on the average Yield of the first three hundred (300) Wafers of each Product manufactured using Qualified Processes and passing all production tests."
The agreement further stipulates, "Any Wafer manufactured by TSMC that yields less than sixty five percent (65%) of the standard Yield for such Product shall, if sorted by TSMC, be scrapped, or, [...] be deemed defective and returnable for replacement."
While the agreement between TSMC and IDT is unlikely to be identical to the one between TSMC and Apple, the basic idea is probably similar.
The next question to ask, then, is if TSMC and Apple are really gunning to get 16-nanometer chips into the 2015 iPhone, who will eat the cost of the low yields, should they persist?
Will TSMC, in a bid to keep Apple's business, offer really cheap wafers to offset the relatively low yield per wafer (and keep this business away from arch-rival, Samsung)? Or will Apple, in a bid to minimize its reliance on Samsung, be willing to buy more, "full-price" wafers from TSMC and agree to a lower yield requirement?
Foolish bottom line
Only time will tell which chip manufacturer will actually wind up with Apple's business for the 2015 iPhone/iPad devices. But given that Digitimes has claimed on separate occasions that both Samsung and TSMC have "won" this business, the "behind-the-scenes" battle is extremely interesting.
