Intel Corp (INTC)

[VeeCee]

Can TSMC Close The Gap With Intel?

By Shuli Ren

This blog can’t have enough of Taiwan Semiconductor Manufacturing (2330.Taiwan/TSM), the world’s largest contract chip manufacturer.

Much of the debate lately is how much of the next-generation iPhone A9 chip business Apple (AAPL) will divert to TSMC’s chief competitor, Samsung Electronics (005930KS/SSNLF). On the technology front, Samsung is exerting very strong competitive pressure on TSMC in the 14/16nm node.

TSMC has been working hard on the next step, the 10nm node. Its management has said that 10nm would begin to make money late 2016. Leapfrogging Samsung is obviously one reason why TSMC is pushing for 10nm. But the company has also stated that its 10nm will close the technology gap with Intel3, “suggesting the ambition to get a slice of the lucrative PC market,” noted Bernstein Research‘s Mark Li and David Dai.

So how is TSMC doing on the 10nm node?

Not very well. According to the Bernstein analysts, TSMC was surpassed by Intel (INTC) on transistor density at 14/16nm node because TSMC’s 16nm is not a full shrink from 20nm, i.e., not a true technological development. To transition to 10nm, TSMC will have to reduce its transistor area by about 70%, “the most aggressive shrink in the history.” Good luck!

As such, “we forecast TSMC’s transition to 10nm will take 1.5 years at the earliest, and that it won’t be able to completely close the gap with Intel,” noted Bernstein.

Here are some more details:

Despite the elevated R&D intensity, we forecast 10nm won’t generate significant revenue for TSMC until mid-2017 and the transition from 16nm to 10nm will take ~1.5 years at the earliest.

Meanwhile, the recent slides from INTC indicate INTC is planning to have an aggressive shrink in 10nm too, while maintaining the 2-year cadence. We hence don’t think TSMC can catch up with INTC entirely, although narrowing the gap is possible.

In light of the history in 14nm, Samsung may offer two versions of 10nm as well. The first version (i.e. 10LPE) will be earlier, followed by a later version, 10LPP. 10LPE likely will have lower specifications but may be ready slightly earlier than TSMC’s 10nm. 10LPP, on the other hand, will arrive later but will deliver more compelling features.
http://blogs.barrons.com/asiastocks/2015/03/16/can-tsmc-close-the-gap-with-intel/?mod=BOLBlog