The increase in wealth-to-income ratios in the second half of XXth century has recently received much attention. We decompose the trend in physical capital and housing, further decomposed into structures and land. In four out of five major countries analyzed, the positive trend in capital-income ratio arises from housing and specifically from its land component. We therefore revisit the question of wealth inequality and taxation in adopting a Georgist perspective (from Henry George, 1879) subsequently endorsed by prominent economists. We introduce land and housing structures in Judd’s optimal taxation framework. We show that an optimal taxation implies a property tax on land and no tax on capital. When the range of property taxes is politically constrained, taxing the product of housing rents is not optimal, even with additional taxes on "imputed rents". Rent taxes are however less distortive than a capital tax. The distortion depends on the share of housing structures and how they react to the tax on rents. However, a tax on rents complemented by a subsidy on structures investments in rental housing units does almost as well as a land tax. As a side result, we find that Judd’s result of no second best capital taxation extends to a larger range of parameters at the steady-state.