We study departures from network neutrality through implementing a quality of service tiering regime in which an Internet service provider charges for prioritization on a nondiscriminatory basis. We find that quality of service tiering may be more efficient in the short run because it better allocates the existing network capacity and in the long run because it provides higher investment incentives due to the increased demand for priority services by the entry of new congestion sensitive content providers. Which network regime is the most efficient depends on the distribution of congestion sensitivity among content providers, but a guideline is that the regime that provides higher incentives for infrastructure investments is more efficient in the long run.