I do not believe that Intel (NASDAQ:INTC) will be building Apple's (NASDAQ:AAPL) "A" series processors anytime soon, and on the flip side, I have absolutely no reason to believe that Apple will suddenly drop its in-house CPU and SoC development efforts and switch to Intel's in-house Atom processors anytime soon. Apple will probably keep having chips built at Samsung (OTC:SSNLF) or, if it really wants out of the Samsung relationship, build them at TSMC (NYSE:TSM).
Apple's Design Points Differs from Intel's
There seems to be this notion that Apple needs to have the absolute best processor hardware in its phones to effectively compete - I submit to you that this is probably not true. Apple's major strength in its system-on-chip designs has actually been to include just enough CPU power and then go all-out on graphics hardware. This, I believe, is one of the key reasons that while the rest of the Android market is busy somewhat mindlessly lapping up "quad core" solutions from the likes of Nvidia (NASDAQ:NVDA) and Qualcomm (NASDAQ:QCOM) that sacrifice graphics performance in order to stuff in "quad core" CPUs into a smartphone, Apple has consistently provided the best gaming/multimedia experiences on its own devices.
Intel's first smartphone chip - Medfield - ironically enough had rather lack-luster on-die graphics, and I believe that this - not CPU performance which was actually quite good - was one of the major reasons that this chip was not widely adopted in the industry (that, and Qualcomm's stranglehold on integrated LTE). "Clover Trail+", which is the beefed up version of "Medfield", actually has a quite good graphics implementation that should be right about on par with its Android peers. However, in both tablets and phones, Apple's much more aggressive implementation of Imagination Tech's latest PowerVR graphics cores still leaves the rest of the industry behind.
So, that's the first barrier for Intel; while I expect designs from them to have leading power consumption and CPU/memory performance, I remain skeptical that Intel is as of yet willing to commit to using its expanded transistor/power budget to really taking home the graphics crown.
Intel Will Not Have Excess Capacity
I think that the hype surrounding Intel's "excess capacity" has gone too far, and this is one of the key tools that Intel bears - particularly on the sell side - have used to great effect to spook investors. As unit volumes in the PC space grew for years, Intel built capacity to match that new demand. In 2009, Intel took excess capacity charges when demand fell off thanks to the economic recession (and it managed to take entire fabs and move them over to next generation process technology), and then it took them during Q4 2012 and Q1 2013 as the PC industry suffered a "reset".
But this is not a long-term thing. Intel is on the cusp of one of the largest product refreshes in history, and at the same time the company is expanding its product offerings into a variety of new areas that it hadn't played in before. In particular, these are some of the things that will help keep the factories running full
- A major ramp up in "Atom" products for tablets
- Moving Infineon Wireless' modems to Intel's own factories (currently all Infineon modems are built at TSMC)
- A ramp of 22nm, LTE-capable cellular apps processor + modem platforms in late 2013 with consumer availability in Q1 2014
- The continued growth of the new "Xeon Phi" product line for HPC, which involves selling very large die parts
- A refresh of the Xeon line, in everything from 2-socket workstations, to a much needed refresh in the big-iron 8-way Xeon-EX line
- Attacking the micro-server space with both Atom and low wattage Xeon E3 parts
- Aiming for communications and networking with "Rangeley"
- Expansion into low end storage with the Atom S12x9 and its successors
- "Haswell" for convertibles and notebooks
- In-vehicle infotainment with an as-of-yet unnamed Atom-based product
Intel's current leading edge capacity will be more than filled by Intel's own products this year, and as Intel gains market segment share in these new areas - and as it identifies new areas - the capacity going forward will continue to see healthy utilization.
What about Intel Custom Foundry?
If Intel's factories will see organic utilization, then why is there any need for this at all? Simple - it's a way to make more money. Now, note that if Intel were to enable a customer like Apple to have access to Intel's latest process technology, then that would have broad political and financial implications. First off, Intel would be helping to make its own "Atom" parts that it is trying to sell to its competitors have less of a "bite". Apple's design teams seem very competent, and it is not unreasonable to think that normalized for process technology that Apple's in-house CPU team would be able to put up a fight against Intel's "Atom" team (the "Core" team within Intel - which is a monster that uses custom tools, lays out all of the key parts of the CPU by hand, and is generally unstoppable - is a different story).
Process technology in and of itself doesn't yield superior designs (although some metrics do improve with a straight shrink), but it enables designers to do in a given performance and $ envelope what others cannot feasibly do from an economic standpoint. This is why process technology leadership is so key, and this is why Intel can't enable any of its competitors.
But will Intel enable/work with chip companies that don't compete with Intel? Sure. You're seeing it now with Altera (NASDAQ:ALTR) and Microsemi (NASDAQ:MSCC), since the process tech will give that company a tangible edge against its competitors, and the margins should be mostly okay, although probably that the low end of the 55% - 65% target for the company. This will be a steady income stream that serves to foster partnerships in the semiconductor industry - a major win/win. But enabling competitors in Intel's server, PC, or mobile markets is a big no-no.
TSMC Isn't All That Behind…For A Non-Merchant Chip Customer
Yes, Intel has a pretty solid lead on TSMC, and for the merchant vendors trying to sell the best possible low power/high performance chip to device vendors, Intel will have a huge advantage. But Apple isn't trying to win design sockets from, say, Qualcomm - it's just building its own chip so that it can have the control that it wants over the design points of the hardware. Would being a process generation ahead be nicer? Sure, things do get better. But at the end of the day, Apple's iOS is very well optimized for the single platform that it needs to run on, and the performance of even Apple's multiple generation old iPhone is superior than many newer Android phones due simply to more tightly optimized software. Being on 20nm planar while Intel-based Android phones are on 22nm FinFET or even 14nm FinFET does pose a performance/watt problem, but at the end of the day, Apple isn't selling chips, it's selling a brand, a software platform, and yes, hardware.
I don't think Apple will stop its in-house development efforts on the chip side, and I don't think there is a need for Apple to be on the bleeding edge of chip tech. Apple's iPhone and iPad have never had the "best" CPU technology, but Apple does spend its transistor budget on world-beating graphics horsepower. Could things get better if Apple had Intel's latest process technology? Sure. But will Apple pay Intel's foundry premium, and would Intel take chip business that in a very real way competes with its own Atom? It's not likely.
From Intel's perspective, its job is to give every two-bit Android phone maker a better chip/platform than what Apple has in a bid to entice Apple to switch over. So even if Apple really wouldn't ever switch over, Intel has to keep trying, and in the process of doing so it aims to take massive share from other industry players.