We study the problem of staffing sizing for an on-demand ride-hailing platform over a d-dimensional connected bounded service region. The decisions include a one-time capacity-investment decision on the number of drivers to employ at the beginning of an infinite planning horizon, and the real-time driver-customer-matching and driver-routing decisions during the horizon. The objective isto minimize the long-run average cost incurred per unit time, where the cost includes the wages to the drivers and the waiting costs of the customers. We show that the optimal staffing level can be written as the sum of a nominal load (the minimum number of drivers needed to prevent the system from exploding) and a safety level, and characterize how the staffing rule responds to several key factors, including the dimensionality d, the probability distribution(s) of customers’ origins and destinations, and the car-pooling capacity q (the maximum number of customers allowed in a vehicle). Specifically, we show (i) when q = 1, the safety level is proportional to the demandarrival rate to the power of d/d+1 , a result that holds broadly across distributional assumptions and mirrors earlier findings under stylized spatial models. (ii) When q ≥ 2, the scaling becomes 2d/d+1 ,which, to our knowledge, was not discovered in the past. Importantly, we derive these results by analyzing exact vehicle routing without relying on stylized assumptions and by developing provably near-optimal policies for general spatial matching settings beyond those in prior work.