It really is about physical implementation. In TTL "single input NAND" and invertor is the same physical structure and extending it to multiple inputs involves just adding additional emitters to one BJT structure. For CMOS logic, NOR is the smallest two-input "true" gate, because PMOS transistors have to be larger than NMOS and two series connected MOS transistors can be collapsed into one structure with multiple gates that is significantly smaller than two connected transistors.