Interesting semantic game. I had to stop and think for a second, but, yes, that is probably literally true (albeit just barely).
Of white Trump voters, probably a little under half are women.
Some of-color people (that semantic boundary is itself flexible) also voted for Trump, probably enough to make up for whatever man/woman difference there was among the white voters, plus enough to get over the 50% threshold.
So yes, literally, "the majority of Donald Trump's voters in 2016 and 2020 were women and people of color" is probably true.
Interesting trick.