Hardware

Imagine we have an 8 bit (1 byte) architecture, so data is stored/processed in 8-bit chunks.

If our RAM holds 256 bits, we can store 32 pieces of data in that RAM (256/8).

If we change to a 16 bit architecture, that same physical RAM now only has the capacity to hold 16 values (256/16). The values can be significantly bigger, but we get less of them.

Bits don't appear out of nothing, they do take physical space, and there is a cost to creating them. We have a tradeoff of the number of values to store vs the size of each value.

For reference, per chunk (or "word") of data:
With 8 bits, we can hold 256 values.
With 64 bits, we can hold 18,446,744,100,000,000,000 values.
With 128 bits, we can hold 3,402,823,670,000,000,000,000,000,000,000,000,000,000 values.
(For X bits, it's 2^X)

Maybe one day we'll get there, but for now, 64 bits seems to be enough for at least consumer-grade computations.