It's a tough call... you have to realize that the choices made by the two CPU companies were not made by flipping a coin. There are reasons behind it.
Intel had many reasons to finally create a new, non-x86 architecture, just about all of them good, including a cleaner (some would debate this. ) architecture, improved performance, and less byzantine CPU designs. But one of the downsides (depends on your point of view) is that it will be much tougher for another CPU company to make IA64 compatible CPUs, as Intel owns a great many of the patents required to implement IA64 in a fast design. Not saying that it would be impossible, it would just likely take a whole different approach to the ISA.
AMD also had reasons for staying with x86. The first is the inverse of the above -- they might not even be able to MAKE an IA64-compatible processor. (Of course they could take the VIA-like approach and say ah heck off until they get dragged into court, but I don't think that would be wise).
The second reason is simple: x86 is bloody popular.
People have been predicting the death of x86 almost since it first was released, but like some out of control automaton it just keep plugging away. Along the way it has made up the performance gap and surpassed almost all of its competitors, those who said it could never deliver in performance like a RISC design could (Notice the top single integer CPUs in SPEC2000 today are the AthlonXp1800 and the P4 2ghz).
So with all of that momentum, is it likely that x86 will die away? My opinion is: not likely. x86's popularity has nothing to do with technical features but just the fact that it owns 98% of the market.
When IA64 comes out, the OS will support it, and high end apps will be compiled to gain the floating point superiority of that design, but the vast majority of the apps, for at least 5 years afterwards IMO, will continue to be written for x86 and thus run through the hardware emulation on ia64.
The reason for this is simple: Someone writing an application or game wants to reach the broadest consumer base possible. It's sort of a no brainer if targeting x86 means you get the entire audience using x86 CPUs, PLUS those using ia64 (through emulation). If you just target IA64, you loose a lot of that legacy market -- and in the near time frame, that's pretty much everything.
During that timeframe it is likely that AMD will thus have the highest-performing x86 processor, since there is a somewhat harsh penalty for running in emulation.
After 5 years from ia64's widespread release, my guess is it will gain enough popularity that AMD will have to look at making a compatible CPU. But we'll see...