I haven’t vetted their specific numbers; but the approach (regarding actual households per square mile) is consistent with mine comparing StarLink with other WISP bands. 

The upshot is that you can consider the shared medium of RF in terms of spectral density (compensating for modulation differences.) Back of the envelope and using anecdotal analysis of Tarana that we see in the field, I think it’s in the correct order of magnitude. 

The comments about Starlink using spatial reuse are legitimate — nothing is preventing it from happening eventually — but their current FCC licensing does not provide for this. Theoretically, we have infinite bandwidth (see Tony Kapela’s work on “FU-MIMO”), but in practice, we have constraints.

Starlink’s strength, as others have pointed out, is ubiquity. I suspect that traditional WISP technologies are still justifiable somewhere above five customers per square mile (reasonably high density, given that Alaska is < 1 !), but where that line is precisely is the difference between a viable business plan and a money sink.