I’d go with a phase 1 kit if buying the prebuilt one would add significant cost.
Personally I can’t see the point in phases 2 or 3. The original goal was a computer where you could understand every part of it. Phase 1 fits that scope much better. Also, it seems to me that planning subsequent phases just makes the project bigger. I’d suggest concentrating on phase 1 and shipping it before making further plans.
Being realistic, very few people will buy this if they’re not interested in retro computer architectures or programming already. It’s not going to have a software ecosystem to compete with today’s consoles. I’d take that clarity of purpose and super optimise for it.
I’m more conflicted about the X8, but would probably lean towards not releasing it for the same reason. The difference in accessing VRAM is my main concern — that would make X8 software incompatible with the X16.