Intel (INTC) is the world's largest semiconductor manufacturer and the world's largest supplier of CPUs (Central Processing Units) for personal computers with about 90% market share. Let me expand on that: Intel 90%, someone else 10%. For simplicity, we will call the "someone else" Advanced Micro devices (AMD). That 10% market share is extremely important to the pricing and, therefore, the gross margin structure of the CPU business.
For most of my 35 years in the semiconductor business and continuing for the 10 years after that AND today, Intel has always been at least one process node ahead of AMD (process node being the smallest feature [transistor] size able to be made on the chips). Node size reductions are generally a 1/3 reduction of the linear dimension, resulting in approximately a 50% reduction in die area and therefore, a doubling of the units produced on the same cost wafer, ergo half the cost. (Wow, when I don't use acronyms or industry vernacular, the sentences seem to grow like weeds.)
Think of it this way: you have a 9" by 9" square of plywood. That would be 81 sq. in. of plywood. Shrink the 9" dimension by one third to 6". Now you have a 6" by 6" square. That would be 36 sq. in. The 6" square is 81-36 = 45 sq. in., smaller than the 9" square. 45/81 = 56% smaller in sq. in. The one half area shrink isn't perfectly true, but it is close enough to grasp the concepts that follow.
Using the above explanation, historically Intel has ALWAYS had chip sizes, for equivalent functionality, of one half when compared to AMD. Therefore, the Intel cost on an approximately equivalent chip is one half that of AMD. So, Intel has the luxury of letting AMD set the market price for a given product. As an example, let's assume that AMD can manufacture a device for $50 per chip. We will assume that AMD wants at least 30% gross margin, so $50/(1.00-.30)=$71.42 price to the customer.
Remember that Intel's cost is not $50 but $25 because of that little one node advantage in manufacturing. So, Intel is perfectly justified in selling that equivalent device for $71.42, plus or minus a few bucks. We'll use a $70 Intel price for simplicity. The customer should be happy and Intel will make a margin of $70-$25/$70=64.2% gross margin. That is exactly how it works in the real world. Intel margins are actually significantly higher than the example because Intel's smaller chip produces better yields, and Intel's generally better manufacturing control also produces better yields.
The conclusion that one must come to after digesting the above is that Intel has an interest in AMD limping along, but never totally leaving the business. Ironic, right?
Let's carry this on to the mobile business.
The Apple (AAPL) A6 chip, the last I heard, was about 97 sq. mm. Let's use 100 sq. mm to make the arithmetic simpler. Samsung (GM:SSNLF) makes the A6 chip for Apple (Tim Cook must be grinding his teeth down to nubs about that) on its 32nm process. That would produce about 570 A6 chips per wafer. Since both companies are in about the same place in regard to depreciation, wafer manufacturing cost for both Samsung and Intel are about the same at about $5000/wafer. Therefore, Samsung's cost on an A6 chip would be about $8.77. iSuppli thinks Apple pays about $17.50 for A6 chips. The Samsung margin would be $17.50-$8.77/$17.50 = 50% gross margin.
If that A6 chip were simply moved to Intel on a 22nm process, the chip size would decline to about 60sq. mm. That would produce 950 A6 chips per Intel 22nm wafer. The Intel cost would be 5000/950=$5.26 per A6 chip. If Intel sold the probably higher performance, lower power, Intel version of the A6 chip for the same $17.50, the Intel gross margin would be $17.50-$5.26/$17.50= 70%. Again, it is actually worse than that because the smaller Intel chip would have higher yields, and the inherently higher yield Intel system would further improve yields.
This should give a reasonable conceptual understanding of the significance of this thing called, "a one node manufacturing advantage." I have simplified the above example and totally ignored actual yield numbers (which, if included, would give even more advantage to Intel).
The other mobile chip makers, all fabless manufacturers, funnel their fabrication business into, primarily, TSMC (TSM). TSMC is also one node+ behind Intel, and very likely to stay there (or worse) for at least the next two years.
On May 6, 2013 Intel held a webcast discussing the details of Silvermont that will be seen in upcoming SoC products such as the Bay Trail quad core device aimed at tablet computers and the Merrifield dual core device that will be aimed at smartphones. During the webcast, one of the Intel presenting engineers referred to the new ARM (ARMH) big-LITTLE architecture as an abomination that was a software nightmare and unnecessarily wasted power and silicon (COST). When Intel refers to your latest technology (new baby) as an abomination, it has to shock everyone in the ARM organization.
SoCs built with the Intel Silvermont will have much higher performance and much lower power than any existing or future ARM based mobile chip offered by any fabless manufacturer or Samsung, and they will have much lower manufacturing cost. These superior Intel devices will be able to be priced at the same price, or less, as competing devices coming from -- let's be honest about this -- the two manufacturing competitors, Samsung and TSMC, and the Intel gross margins will be much higher.
Nvidia, Qualcomm, and even ARM have utterly no control over the outcome of the mobile chip business because they don't actually manufacture their SoCs. The extent of their control is to decide, perhaps, to leave the mobile business like Texas Instruments (TXN) and Freescale Semiconductor (FSL) did last year.
The right play might be a paired trade of long INTC and short ARMH.
Additional disclosure: I will buy puts on ARMH when/if it hits $50/share.