Altera has announced that its next generation 64-bit ARM parts will be built by Intel. The announcement caused shockwaves and rippled through the industry for no apparent reason. Many observers wrongly viewed it as a deal under which Intel is yielding to competitive pressure and opening its fabs to ARM outfits. This is simply not the case.
The Altera deal includes the fabrication of several designs based on ARM’s new 64-bit instruction set and the quad-core chips in question will apparently be based on the Cortex A53 cores. They are not consumer parts, they are aimed at the embedded market.
For some reason overexcited hacks and analysts quickly started saying the deal would open the floodgates to ARM chips built in Intel’s fabs, using the upcoming 14nm FinFET process. This is simply not the case. There is absolutely no reason for Intel to allow competitors like Qualcomm, Samsung or even Nvidia to use its process for consumer-grade parts.
In fact, the only way Intel’s own x86 SoCs can be competitive in the short run is if Intel leverages its superior manufacturing process. As x86 is inherently less efficient than ARM, Intel simply needs to cram more transistors to get the same results, but x86 parts also have a lot of advantages, namely Windows support.
While it is true that fabs don’t come cheap and Intel could use the cash, that doesn’t mean it will put a gun to its head for a bit of fab revenue. It will not and we will not see third-party application processors manufactured alongside Intel’s own Airmont parts. Anyone claiming otherwise is spectacularly wrong.