Anyone, care to comment ?
In June we reported that TSMC (Semiconductor Manufacturing Company) and its Integrated Circuit Design service partner Global UniChip had secured a three-year agreement with Apple to supply foundry services for the next A-series chips built using 20nm, 16nm and 10nm process nodes. The foundry was to complete installing a batch of new 20nm fab equipment, which is capable of processing 50,000 wafers, in the first quarter of 2014. Today a report claims that TSMC is aggressively ahead of schedule.
According to a new DigiTimes report, TSMC has stepped up its purchases of manufacturing equipment for its 20nm process which is slated to enter volume production in the first quarter of 2014, according to sources at fab-tool suppliers. Today, Apple's new A7 uses a 28nm processor while Apple's A8 is scheduled to shift to a 20nm processor in 2014.
The report further noted that TSMC's investment in 20nm manufacturing equipment for the fourth quarter of 2013 is expected to outpace that allocated for the third quarter. In addition, shipments of equipment for TSMC's 16nm HKMG process have kicked off. TSMC expects to initiate volume production of 20nm chips in early 2014 followed by volume production of 16nm FinFETs in about one year which is in line with their contract with Apple
It is very shocking that TSMC has caught up to Intel so fast. Intel was boasting of multi-year process advantages only a few months ago. Intel is moving way too slow, period. New CEO is supposed to be a manufacturing guy. But, he's executing as poorly as Otellini.
It is evident that apple will replace the x86 chips in the macs (air + pros) with their own . giving them more margins in their products. this means two things: 1) ARMH long 2) short intel . Which i happily did on both.
Apple is never ever going to switch to intel chips in the phones, and is not going to stay with intel x86 chips.
My estimate is that ARMH will hit 120 by next April, intc will be round17.
Sure, I'd be happy to comment:
This is just another fluff announcement from TSMC's marketing department.
The fact is that they have had increasing ramp problems for several nodes now. The idea that they will suddenly eliminate those problems is silly. The 28nm ramp was problematic and the 20nm one will be more so. The increase in foundry (ARM) ramp times is well-documented.
One must differentiate between production and volume production. Volume production at 20nm is a long ways off.
A couple of other key issues:
1.) Even when ARM 20nm arrives, it is important to note that it has damaged economics because it is planar. This means it won't produce the same performance gains and the same cost advantages as Intel's 22nm has. That's very bad news for ARM's ability to compete as Intel continues to step things up.
2.) If Apple does use TSMC for 20nm, there will be a huge clusterfunk at the top as Qualcomm, Nvidia and AMD compete for what is going to be very scarce capacity at a very high cost. Great for TSMC but a disaster for the ARM shops and ARM in general. Limited improvements, a high cost, and scarce capacity. What's to get excited about here? Nothing from an ARM perspective. It's a trainwreck about to happen even if it arrives on schedule which it won't.
3.) We are now being told that TSMC is moving CapEx from FinFET back to the 20nm effort. This is incredibly bad news for ARM investors. It means that the ARM FinFET future is in serious doubt. The bottom line is: No FinFET, no future. It's that simple.