I don't think he's claiming that it's causally a function of service time (although this can be the case if demand is elastic). Rather, utilisation is calculated from service time and arrival rate, and they're just stating that this equation can be rearranged to calculate arrival rate if you know the other two.