Yeah, very true. The algorithm to solve the cube is simple. The accomplishment here is the mechanical part.
At 60 fps, unless you're streaming subframe data you're looking at around 160ms ms per frame, and that's not counting any time for image processing.
(Presumably they could cheat by looking at the cube beforehand, but it sounds like a realtime thing)
More accurately - probably not. It depends how the actuators work and whether they include optical monitoring of the twist.
If so, the solution won't need to monitor the colours, but it will need to make sure the actuators spin the sides accurately.
Otherwise - cube explosion.