Robot cars are cars that allow for automated driving. They can drive closer together than human driven normal cars, and thereby raise road capacity. Obtaining a robot car instead of a normal car can also be expected to lower the users value of time losses (VOT), because travel time can be used for other activities than driving. With a mix of normal and robot car users, the VOT is therefore (more strongly) heterogeneous. We study the effect of robot cars on social welfare for a number of market organizations: private monopoly, perfect competition and public supply. Increasing the share of robot cars raises average capacity (especially if robot cars drive concentrated in time), but may hurt existing robot car users as the switchers lowered VOT will increase their bottleneck-congestion externality. When the capacity effect dominates, buying a robot imposes a net positive externality, but otherwise, it causes a net negative externality. Numerical analysis suggests that a net positive externality is more likely; nevertheless, for a small, but still plausible, capacity effect a net negative externality results. With a positive (negative) externality, marginal cost provision tends to lead to an undersupply of robot cars, and a public supplier needs to subsidise (tax) robot car purchase in order to maximise welfare. A monopolist supplier ignores the externality and tends to add a mark-up to its price. This almost always leads to a substantial undersupply.