In this paper we report the results of a randomized experiment testing impacts of subsidies for modern agricultural inputs in rural Mozambique. One-time provision of a voucher for fertilizer and improved seeds leads to substantial increases in fertilizer use, which persist through two subsequent agricultural seasons. Voucher receipt also leads to large, persistent increases in household agricultural production and market sales, per capita consumption, assets, durable good ownership, and housing improvements. Consistent with learning models of the adoption decision, we find positive treatment effects on farmers' estimated returns to the input package. We also document positive cross-household treatment spillovers: one's own fertilizer use rises in the number of social network members receiving vouchers. Our findings are consistent with theoretical models predicting persistence of impacts of temporary technology adoption subsidies, in particular due to learning effects.