Assuming that Target Smart has correct demographic information, which I most certainly question, comparison to 2020 is really misleading because of clear change in early voting patterns due to the pandemic.
A comparison to 2022 is in my mind more interesting as it shows a greater voter share for Asian/Pacific Islander, Black, and Hispanic voters and a lower White share (66.6% in 2024 versus 70.1% in 2022 for early vote 6 days out). The early vote share for whites in 2022 decreased over the last 6 days to 68.5%. And on election date that percentage shrank further to 66.9%.
Another interesting data point, in 2020 absentee ballots favored Democrats (which was a change from what had previously occurred). According to TargetEarly, only 19.5% (332,000 of 1,699,000) of the those white absentee voters from 2020 have voted early so far. On the other hand, 46% (377,000 of 813,000) of the white election day voters from 2020 have already voted. Assuming that election day vote favored Trump, this may be additional evidence that Trump is cannibalizing election day voters in the early vote. It would also suggest that Harris has a lot more votes still available. The question there is whether this shows reduced enthusiasm (I doubt that) or a shift back to election day voting (more likely in my opinion). Oh, and 5.6% (248,000) of the white early vote so far was unregistered in 2020. And 4.5% (197,000) were registered in 2020 but didn't vote in 2020 (or apparently 2022).