A lot of the rotation resistance is dynamic bending forces as the shaft is continuously re-bent during rotation.
The simplest way to test AWES scaling performance is to build a larger prototype of the same type and measure the loss of power-to-weight performance. The projected scaling loss between those two data points is roughly linear for the next up-scaling loss. Many teams up-scaling efforts reveal this. Makani’s M5 5MW concept died quietly as they learned the M600 was already too big for its premature architectural down-select.
KiteLab Group and kPower have long used many such “quick and dirty” empirical methods at small scales for insight into AWES principles. Compare such practice against narrow numeric calculations based on poor heuristics prevailing in other circles (GIGO effect).