For the kind of http-pushing boxes that are called cloud servers, architecture matters very little. As far as ISA goes, ARM has no advantage over x86 and x86 has no advantage over ARM.
What matters is cost, system topology, IO, memory latency/throughput, and being low power. The code they will be running is all branches and pointer chasing, and they will spend most of the time waiting for IO.
ARM has the advantage of very low-cost power-sipping systems, but the disadvantage of generally anemic memory subsystems. Also, since everyone can design their own systems, they can ship things like chips that have 4 separate quad-core systems, each with a memory channel, ethernet, and sharing the rest of the system architecture.
x86 advantages are better manufacturing process (for Intel chips), better memory subsystem, with the disadvantage that Intel can't make them too good or it will cannibalize other, higher-margin segments.
I understand that "cloud server" is a growing market, however I question why people think it's so interesting. The whole market is driven by TCO, they don't want high performance. If a core can run a typical PHP script in 100ms, it's good enough, and beyond that only throughput/$ matters. Since everything is "good enough" by a margin of 10x or more, really the only way to compete is cost. Intel will release chips to compete in the market, but if they price them low enough to be actually competitive, it will cost them more money than they can earn by cannibalizing sales in segments where x86 can command a premium. So, what will happen is that all the different ARM vendors, and possibly AMD, will release chips to sell that will take all the market share, but since the only way to compete is price, and there are so many vendors, none of them will make any money.