In this paper, we study an irreversible investment problem under Knightian
uncertainty. In a general framework, in which Knightian uncertainty is modeled through a
set of multiple priors, we prove existence and uniqueness of the optimal investment plan,
and derive necessary and sufficient conditions for optimality. This allows us to construct the
optimal policy in terms of the solution to a stochastic backward equation under the worst-
case scenario. In a time-homogeneous setting { where risk is driven by a geometric Brownian
motion and Knightian uncertainty is realized through a so-called "\$\kappa$ --ignorance" - we are able
to provide the explicit form of the optimal irreversible investment plan.