you are overstating things significantly, ARM is still way behind (an order of magnitude or more) on single thread preformance. Nobody cares about preformance per watt if it means everything lags a massive amount.
As evidenced by the fact that Intel has failed to compete in mobile, this is obviously false.
We've been hearing nonstop from message board enthusiasts that Intel is going to kill ARM for years and years now. In reality there's absolutely no sign of that happening, and in fact their recent deep cuts to the Atom line suggest the opposite.
*edit: would be interesting to see an elasticsearch cluster running on a high clock speed arm cluster... I wonder how it'd run... (and yes I know it's not that simple)
http://www.theverge.com/2016/9/16/12939310/iphone-7-a10-fusi...
I can list about 150 programming languages that are "faster than C" if you apply similar restrictions.
Also it's 4 year old Xeon. If you're going to compare it to a 4 year old Xeon, compare it to the 2 core versions, which will have much higher clocks.