A non-repairable system is subjected to randomly occurring shocks whose counting process is a renewal process. If the time between two successive shocks is less than a random time called the threshold time, the system fails at the occurrence of the latter shock. A spare with random lead time is ordered when the system is put in operation. Adopting a T-policy and using an appropriate long-run average cost per unit time which reflects the cost of storing a spare and the shocks as well as the cost of downtime an optimal replacement policy is obtained. As an illustration, exponential inter-shock arrival times with constant threshold times G is considered. Some existing models are shown to be particular cases of the developed model.