This paper develops two-period analytical and numerical models to study the question: given a stock of greenhouse gases that poses a risk of future damages of unknown magnitude, and the possibility of learning about damages, how do sunk abatement capital and a nondegradable stock of greenhouse gases affect optimal first-period investment? We show that both affect investment, the former negatively and the latter positively. Additionally, endogenous risk--the risk of damages dependent on the stock of gases-- results in an increase in optimal investment for any level of capital "sunkness" or greenhouse gas degradability. Quantitatively, though, the effect of sunk capital is much stronger than the effect of greenhouse gas irreversibility or that of endogenous risk.