Cost estimates for Computer Assisted Instruction (CAI) depend crucially upon the particular assumptions made about the components of the system to be included in the costs, the expected lifetime of the system and courseware, and the anticipated student utilization of the system/courseware. The cost estimates of three currently operational systems (PLATO IV, TICCIT, Computer Curriculum Corporation) are considered in terms of these assumptions. The assumption made in cost comparisons that the instructional effectiveness of CAI is the same as that of traditional instruction is questioned. Future developments which will affect cost estimates of CAI are discussed. (Author/VT)