mql5

Intel, TSMC, Samsung, Apple: Who's On First?

• Seems unlikely that TSMC will supply a 20nm A8 chip to Apple.
• Samsung might have won the business after all.
• Could Intel be the stealth supplier of the A8 chip?

The common, and I DO mean common, wisdom is that TSMC (TSM) will be building the Apple (AAPL) A8 chip to be used in the iPhone 6. It's right here in black and white. According to the article, Samsung can't yield 20 nm. Hmmm, that seems to be a recurring theme, with Global Foundries thinking of skipping 20nm altogether.

Then, less than a week later, Samsung is back in the fray, apparently with a 20nm process. So is it TSMC or Samsung? Or, is it someone else.

The only other potential supplier for this bleeding edge part would be Intel (INTC) on a 14nm TriGate process, but there are virtually no rumors about Intel and Apple, which I find odd.

So, in an article out of Taiwan yesterday, we learn that TSMC will have 10% of their revenue coming from 20nm planar by the third quarter and 20% by the fourth quarter. We also learned that the total TSMC capacity, stated in 8 inch (200mm) wafer equivalents is 15.67 million wafers per year. Divided by 2.25, we get the 300mm wafer capacity of about 7 million wafers or about 580,000 wafer starts per month. That number is up from (taking a shortcut here) 520,000 300mm wafer starts per month in 2012. That is total of 60,000 new wafer starts of capacity added in 2013. Nowhere near what I would have guessed.
The 20nm process being run at TSMC is so edgy that it would be safe to assume the company is charging 4X for a 20nm wafer over the average selling price of all TSMC wafers. Now we're getting somewhere. The 10% of revenue for 20nm becomes 2.5% of wafer capacity or about 14500 wafers, doubling to 29,000 wafers per month in the fourth quarter.

Everyone in the world claims to be getting at least a few of these wafers, but if Apple got them all it wouldn't come close to filling their demand for A8 chips. In an early ramp, TSMC would be fortunate indeed to get 200 good 100 square mm chips from these 20nm wafers. That would be 2.9 million A8 chips per month in the third quarter, way below Apple demand. And, of course, that would be in the third quarter; the same quarter the iPhone 6 is expected to launch. The real world would require massive shipment of these parts at least a quarter before the iPhone 6 release, like right about now; you know, like the second quarter that Intel just told us was going to be about a billion dollars better than thought.

The bottom line is that it is extremely unlikely that TSMC will be supplying 20nm A8 chips to Apple. It is far more likely that Samsung snatched victory from the jaws of defeat on the A8 business AGAIN. Maybe the A8 isn't 20nm, but that is not what the rumor spreaders are saying.

The Intel angle here is that there is an infinitesimally small chance that Intel could be the stealth supplier of the A8 chip. The more real world is that Samsung supplies the chip again and leaves Intel alive for the A9 chip in 2015. That would be encouraging since, if Apple change suppliers for the A chip, they are not likely to change suppliers again in my lifetime and Intel would never get the business.

That's a sort of tortured way around coming to the conclusion that Intel/Apple is ALIVE…ALIVE, I tell you! Well, OK, I'll settle down now.
I think I just talked myself into buying some more Intel.