In the United States, defense R&D share of GDP has decreased significantly since 1960. To analyze the implications on growth and welfare, we develop an R&D-based growth model that features the commonly discussed crowding-out and spillover effects of defense R&D on civilian R&D. The model also captures the effects of defense technology on (a) national security resembling consumption-type public goods and (b) aggregate productivity via the spin-off effect resembling productive public goods. In this framework, economic growth is driven by market-based civilian R&D as in standard R&D-based growth models and government-financed public goods (i.e., defense R&D) as in Barro (1990). We find that defense R&D has an inverted-U effect on growth, and the growth-maximizing level of defense R&D is increasing in the spillover and spin-off effects. As for the welfare-maximizing level of defense R&D, it is increasing in the security-enhancing effect of defense technology, and there exists a critical degree of this security-enhancing effect below (above) which the welfare-maximizing level is below (above) the growth-maximizing level.