We are currently not putting a lot of effort in supporting 32bit ARM and RISC-V. For ARM, the 32bit version is old, and as a result it has a lot of variations and different generations (hardware floating point or not? ARMv6 or ARMv7?). This means every device is different and it is an endless source of problems for us. The 64bit version sets a more sane base to start from.
In the case of RISC-V, we simply don’t expect any 32bit “desktop” machine to be available.
In the case of Sparc, the architecture is 64bit since 1998, there is really no point in supporting older system (and that’s assuming there is a point in supporting Sparc at all).
The PowerPC port is currently 32bit only, and generating not a lot of interest and excitement. The m68k port is also 32bit only since m68k never had a 64bit version.
So, long story short: we have enough problems supporting new architectures, and it seems not wise to focus on older versions of CPU architectures. 32bit is making our work harder and is not “natural” at all since on all modern platforms, you have to deal with some kind of PAE: 64bit physical addresses, 32bit virtual addresses, so that you can handle more than 4GB of RAM (and on ARM and RISC-V, this would even be “more than 4GB of address space”, which there includes a lot more than RAM, you will probably be limited to 2 or 3GB of RAM).