Yeah he's talking about making it _economical in rural Africa_ and complementing unstable demand with bitcoin earnings to make the ends meet. I think it's a neat idea.
As for the useful computations you propose that ostensibly pay as well as cryptocurrency mining... which ones? If you're referring to something like cloud computing, keep in mind that the bandwidth requirements for cryptocurrency mining are basically negligible - which is an important factor to consider even in rural America, let alone rural Africa.