I've been watching and writing about Intel (INTC) for most of the last year.
Intel is the world's largest producer of semiconductors, by dollar volume. In terms of wafer production Intel follows Samsung (GM:SSNLF), Micron (MU)/Elpida (OTC:ELPDF), and Taiwan Semiconductor (TSM), in that order. Intel is by far the leader in fabrication technology, so its wafers generate revenue of several times the wafers from the other names mentioned.
The enigma comes with the manic pace of capacity addition by Intel. Given the number of facilities added or under construction, the reduced feature size, and the sheer size of investment, it appears that Intel will end 2013 with 2-3 times the capacity required to serve a flattish PC market.
Where does Intel expect the sales to come from that would be necessary to fill this capacity?
First we look at the Google (GOOG)/Intel relationship. Seventeen months ago, Intel entered into a relationship with Google, the goal of which was to fine tune the Android operating system for X86 processors. To demonstrate the value of matching the OS and the physical chip, one only need to compare the Apple (AAPL) iPhone/iPad with other ARM-based (ARMH) smartphones and tablets. There is no comparison, in my opinion and experience. A Nexus 7 with the Nvidia (NVDA) Tegra3 can't come close to matching even a second generation iPad with the A5 Apple chip. Apple has matched iOS with the "A" chips and it shows in the relative performance.
After 17 months it would be reasonable to expect some results from the Intel/Google relationship. Google will announce the X-Phone sometime soon. Rumors are for a May introduction with a July ship date. If that is true, the X-Phone will probably use the Intel 22nm "Baytrail" chip, which is a completely new Atom architecture with a completely new graphics engine.
I have difficulty accepting that Google/Intel will let this week's MWC (Mobile World Conference) go by without an X-Phone announcement or discussion. If they were to talk about an X-Phone this week it would probably be built around the Intel Clovertrail+ chip, which is a dual core Medfield with much improved graphics. The only other product to use this processor is the Lenovo K900 which is not shipping yet. The "+" part of the Clovertrail+ can only be speculated as to more function or significantly lower power.
The X-Phone, whenever it comes, simply has to be the highest performance and lowest power smartphone, by a wide margin, on the market. If the recent Google comments about Apple and Samsung getting in the way of Google's unique approach to monetizing Android, one could imagine an X-Phone with a version of Android not available to the rest of the world.
The market for cellphone application processors, including 4G LTE where required, is quickly on its way to a billion units worldwide. Tablet application could add another 300 million units to that. A ballpark dollar figure for that business is about $40 billion. That is amazingly close to the low end of the "extra" capacity coming online at Intel. The haunting question is: "Does Intel, with detailed knowledge of the performance of the Baytrail platform, believe that the entire mobile market will have no choice but to use the Intel product in any and all mobile products?"
Ok, we'll set that aside for the moment.
Apple will need approximately 300 million "A" chips. At $30 each, this is a $9 billion piece of business. Speculation abounds on who becomes the post-Samsung foundry for this business. The latest speculation is that TSMC gets the business. I find that laughable. TSMC is falling further behind Intel with every day that passes and has demonstrated nothing better than a 28nm planar process. TSMC has had a terrible time ramping both 40 and 28nm, to the distinct detriment of their customers. The unavoidable fact is that Apple needs to cut a deal with Intel or fall victim to the Google/Intel thrust into mobile computing.
As one who has personally designed integrated circuits in both 32nm and 22nm at Intel (I am now retired), I can tell you - with certainty - that you are wrong on several counts.
1. 32 nm is not cutting edge because the chips consume much more power than 22 nm chips and they are larger, and they are inherently more expensive in volume because of that.
2. Many startups and small companies ARE eager to use the Intel fab, but they can't get access because too many LARGE customers are already signed up and the foundry can't deal with all the small customers and their support needs.
3. There is some truth to the difficulty of moving from other processes to 22nm, but that difficulty is exacerbated by item 2, above. I believe that Intel will fix this issue.
If this is to be believed, Intel must be well down the road on a foundry model business. Paul Otellini recently mentioned that we will not know whether foundry customers have changed suppliers until the change is already done, so Apple could already be in hand and we wouldn't know it. In the case of Apple, it is altogether reasonable to expect it to run Samsung and a new foundry in parallel to ensure that the new foundry could do what Apple needs done. With the cash that Apple has, it would be no problem to throw $2-3 billion worth of Samsung "A" chips in the dumpster if it would make the transition seamless.
We've heard the rumors of an Intel foundry deal with Cisco (CSCO). Other customers could be Xilinx (XLNX) and Altera (ALTR). There is not more than a few billion in revenue combined with these three potential customers. Of course, if Apple were already in process at Intel, it would go a long way to explaining the confidence that Intel has in filling the new capacity.
Intel isn't bringing on this new massive capacity for no reason; the company is much, much smarter than that. From a spending standpoint, Intel has always been known to be, umm, "frugal". Intel might be rolling the dice on filling this capacity but only if it has a plan A, and a plan B. Plan A could be total domination of the mobile chip business based on superior performance and lowest power. Failing that, Intel could roll into the foundry business with a long list of grateful customers; grateful for the state of the art fab, and grateful that they would not have to compete with Intel. We could get some clarity on the mobile intentions of Intel as early as the MWC Mobile World Congress this week. Intel is having a Monday conference with customers to take a look-see at the Intel mobile product line-up.
These giant moves always take longer than I think, so I am buying Jan 2014 30 calls and Jan 2015 30 calls. With dominance of the mobile business in hand, an Intel stock price of under $30 will be a vague memory.