Can subsidies to renewable energy efficiently internalise CO2 costs in electricity production? Under current policy design it only matters that the replaced energy is dirty, but not how dirty it is. We use a modified peak-load pricing model, including variable renewable generators and the external costs of carbon, to examine whether a unit subsidy to variable renewables successfully restores first best equilibrium. In our model, electricity is generated using any combination of three technology types: two dispatchable, thermal, and CO2 emitting technologies, differing in their emission intensity, and a non-dispatchable renewable technology. Using this model, we show that available wind capacity is never idle, and derive equations determining optimal installed capacities for all technologies. We then demonstrate how a subsidy that does not discriminate between dirty energies fails to restore first best: it either replaces an insufficient amount of dirty energy, or does not replace the most carbon intensive energy source.