This paper compares securities settlement gross and netting architectures. It studies settlement risk arising from exogenous operational delays and compares settlement failures between the two architectures as functions of the length of the settlement interval under different market conditions. While settlement failures are non monotonically related to the length of settlement cycles under both architectures, there is no clear cut ranking of which architecture delivers greater stability. We show that while, on average, netting systems seem to be more stable than gross systems, rare events may lead to contagious defaults that could affect the all system. Furthermore netting system are very sensitive to the number and initial distribution of traded shares.