We investigate a long-run impact of a compound disaster in northern Taiwan by describing a recovery process from the disaster with a dynamic computable general equilibrium model. After simulating losses of capital and labor in combination with a nuclear power shutdown, we conduct policy experiments that are aimed at recovery of Taiwan’s major industries by subsidizing their output or capital use. We found that the semiconductor industry could recover but need a huge amount of subsidies while the electronic equipment sector could almost recover even without subsidies. Capital-use subsidies would cost less than output subsidies. When we use two-year longer duration for a recovery program of semiconductors, we could save the subsidy costs by 7–10%.