64-bit address space is sort of the point of diminishing returns for memory. You are not going to have more than 18 quintillion bytes of memory! Going to larger and larger bus widths should happen, since bus bandwidth, not CPU speed, it pretty much the limiting factor for most massively parallel computing problems. GPUs are currently using 256-bit busses.
What determines the memory of a system is not the Bit-ness. Is the address bus size. The Bit-ness of a processor is determined by the Register size.
If only bit-ness determines the size of memory, 8-Bit computers would have had only 256 BYTES (not Kbytes, BYTES total) of memory. But, while the registers were 8-bit only, address buses were 16Bytes.
The OG 8086 was 16 Bit (as in 16 Bit registers) but did not have 64Kbytes of memory. No. It had a whopping 1024Kbyets. You know how? 20 Address Lines.
The 80286 was also a 16 bit processor, but it had 16MBytes of RAM. How you ask? 24 Address lines.
The mighty PentiumPro was a 32 Bit processor, but it could address more than 4Gbytes of RAM. How? 36 Address lines instead of only 32...
The OG 64 bit AMD processors did no support the full address space, only having 128GB of RAM... I guess now you understand why...
Going from 64Bit processors to 128Bit processors allows a few things:
1.) Bigger registers are "moar betta"
2.) One can expand the number of registers
3.) One can do clean-up of the architecture, not implementing inefficient instructions in 128 Bit mode.
4.) One can implement more efficient instructions and ways of doing things that would otherwise be disruptive in the lower bit-ness modes.
So yes, companies like Intel, AMD and Microsoft are already preparing for the 128 Bit future. Are you?
