In the context of a green energy transition, capital adjustment costs render effective substitution between clean and dirty energy sources finite and endogenous, despite infinite long-run substitutability. Ramsey optimal paths robustly frontload clean investment before exhaustion of a given carbon budget, but also generally imply some capital stranding. Along the path of emissions reduction, new investment is quantitatively more important than reduced output or labor redeployment. An ambitious climate goal in our benchmark calibration implies modest levels of stranded capital at 1.5% of GDP, but this rises to more than 7% if implementation is delayed by a decade.