I know Intel made great strides with their 22 nm silvermont, however keep in mind ARM will have it's licensees reach that in the near future and when they do, the power consumption on ARM will go down just as much percentage wise as it did on Intel. Keep in mind also that 5nm is the limit as to how small they can go, so if both of them got to 5nm apples to apples ARM is a lot more efficient. Also ARM has an advantage in that chip makers can customize chips for their needs. Companies that use Intel have to rely on Intel to make their chips and they don't get to customize it like an ARM licensee can. I think there is a very good reason AMD decided to start using ARM's architecture, when it comes to power consumption it's the best.
It is not just ARMH vs INTC; it's a group of companies vs INTC. INTC builds on a business model of heavy gross margin to pay for its huge overhead. The ARMH group is much efficient. INTC has to make sure every nodes are in best condition which is almost mission impossible. For the ARMH group, they can pick the better one among TSM, Global foundry, UMC, ... on wafer fab and ASX, SPIL, ... on chip packaging. Also, these companies have far less incentive as INTC to be top heavy on management actually slow down development.
Chip buyer looks at price-performance-power. ARM is cheaper because of their distributed business model -- they create the IP, their licensee add value to that and foundry fabricates. Starting with ARM's core is a lot cheaper than R&D for your own core. This reduces cost of market entry and many players are there (4 big ones in mobile chips). Intel is fully integrated. They can build one chip highly efficiently. But they don't really know which product it will go into so they have to tune to a broad target. The ARM camp has 4 different chips, each one tuned to a different targeted mobile device. This makes their solution better optimized and they're already ahead on price.
Buyers don't care what underlying technology you're using to optimize your chip's fabrication.
Unlike in the past, the foundries are well funded due to the enormous new market for mobile devices, most of which are ARM based. This makes it likely that they will become a lot more competitive on the fabrication front. Intel's first 22nm low power chips are due out end of this year whereas Apple's A7 based on TSMC's 20nm planar process is due out mid next year. In other words, as of now, they're more or less level on that front and expected to maintain relative position going into the future. Intel rumoured fabrication advantage over ARM's foundry based fabrication is therefore a myth. As you correctly highlight, ARM has significant business model advantage over the vertically integrated Intel.
"Intel's first 22nm low power chips are due out end of this year whereas Apple's A7 based on TSMC's 20nm planar process is due out mid next year."
Intel's first 22nm FULL FINFET SoCs will be shipping in August 2013. Apple's 20nm PLANER (non finfet) will be out in mid 2014. In mid 2014, Intel will begin shipping 14nm FULL FINFET for the fall 2014 devices.
In 2015, Intel's 10nm SoCs will begin production and TSMC will TRY (AND FAIL?) to produce 16nm HALF FINFET processors.
INTEL'S FABICATION LEAD IS MEASURED IN YEARS and TSMC is about $20 billion in cap ex from getting to where Intel is today.
ARM still needs to pump out 64bit, the current 40bit bridge concept on it's way to 64 is baffling to me. Too early to say that its more effective to modify existing ARM designs than to rely on Intel because intel hasn't released the details about how their foundry business is going to operate...just leaves a bunch of ?????'s. ARM has the market now so it will definitely be a challenge to Intel.....Inel will have the best chip's but that doe mean that ARM loses the war.....it is definitely a break ARMs most heavly defended front lines.