A simple model of software costs for the square kilometre array

Software costs for radio telescopes have nearly always been underestimated. Since the Square Kilometre Array is often called a software telescope, repeating the usual error would be particularly egregious. We estimate software costs by scaling from the reasonably well-known costs for the Atacama Large Millimeter Array. The resulting model has sharp dependency on the complexity of the SKA, suggesting the obvious - that software costs can most easily be limited by constraining the scientific and operational requirements. A bottom-up costing will not be possible until SKA is much more clearly defined. For the moment, we recommend that 20% of the SKA budget be allocated to software development.