– Rumor has it Apple and Intel are discussing a fab relationship by which Apple’s next ARM-based processors could be built on Intel’s world-leading 22nm and coming 14nm processes.
– For Apple, whose processor design advantage has eroded with each generation, using 14nm in 2013 could leapfrog rivals, simultaneously improving performance and power efficiency by nearly 30%.
– For Intel, it would mean selling much of the capacity of its most valuable process to a 3rd party, filling its factories but an implicit admission of its failure to sell its own mobile architecture.
– While the business needs match, the benefits seem greater for Apple than for Intel, which may still harbor fantasies of its mobile success – expect Intel to demand a substantial premium.
For the first time in a while, there is some encouraging news on Apple. A week ago, Intel’s head of its new custom foundry business, Sunit Rikhi, told Reuters that his group was “ready to take on a new, unidentified potential mobile customer”. Considering that Intel has upped its capital spending plans for 2013 by nearly 20% to $13B amidst a global PC slowdown, many observers are giddy with speculation of an Apple-Intel tie up. Apple’s proprietary ARM-based processor designs vie with Qualcomm’s Snapdragon line and Samsung’s Exynos for bragging rights in the mobile world, while Intel’s TriGate chip manufacturing process has hit 22nm on its way to 14nm later this year, at least 12 months ahead of the Asian foundries that make almost all of the world’s mobile semiconductors. Apple processors implemented on Intel’s TriGate manufacturing would be like chocolate and peanut butter.
Global Foundries performed simulation tests to estimate the performance and power efficiency implications of using a 14nm process for an ARM-based processor vs. the 28nm that is the current standard. It estimated a 61% improvement in performance at constant power levels, and a 62% reduction in power use at constant performance levels. Assuming that a processor is designed to take advantage of both benefits in equal measure, performance and power efficiency could be simultaneously improved by nearly 30%. Even with TSM and Samsung moving to 22nm soon, Apple could gain a serious edge vs. its mobile device rivals in the processing speed and battery life of its products, at least until the Asian foundries closed the process gap.
Of course, this advantage wouldn’t come cheaply. Intel has been trying to use its process leadership as its wedge to get into the mobile device market with its own processor architecture. Opening its manufacturing lines to Apple would be akin to an acknowledgement that the strategy had failed – if you can’t beat’em, join’em. The only way to make that crow palatable to Intel is with fat margins on the price of its birthright. Having a superior, but likely significantly more expensive, core processor would certainly help Apple re-establish itself as best-in-class at the top of the mobile device market, albeit with yet another deleterious effect on the relative cost of goods sold.
Intel has been beating its head against the wall trying to penetrate the mobile processor market. Its line of SoCs for the smartphone and tablet markets based on its Atom core have not found an enthusiastic market, and its own roadmap for delivering 22nm and 14nm solutions appears to have slipped, with the 22nm “Silvermont” chips now forecast to show in commercial devices in 2014 and the 14nm “Airmont”, a year later. Intel’s solutions suffer from two major problems. First, Atom is based on a Complex Instruction Set (CISC) architecture that is inherently bigger and more power hungry than the Reduced Instruction Set (RISC) approach of ARM and its licensees. For general purpose computers, the CISC architecture has significant flexibility benefits, but mobile devices have evolved to a much narrower set of needs, making CISC an expensive overkill. Even with the performance and power efficiency benefits of Intel’s superior process, rivals close the gap with their more streamlined designs. Second, the world’s mobile device makers have standardized on ARM’s widely licensed RISC processor architecture for generations, with substantial investment in software and system design based on it. Using Intel’s chips is not as simple as plug and play, and OEMs are not enthusiastic about investing to change. Still, the rhetoric from Intel has not softened, making a deal with Apple to manufacture its ARM-based “A” line of processors in its fabs a bitter pill, particularly if, as is likely, Apple’s part broadly outperforms Intel’s own offerings.
Still if Intel can get past its wounded pride, there is the making of a deal here that could benefit both parties. Given the performance of Apple and Intel stock over the past 6 months, that should be welcome news to investors.
For our full research notes, please visit our published research site.