The problem of how to design multiproduct batch plants for the case of uncertain product demands described by any continuous/discrete probability distributional form, is considered in this paper. Based on a stochastic programming formulation, featuring an objective function comprising investment costs for equipment sizing, expected revenues from product sales and a penalty term accounting for partial feasibility and unfilled orders, it is first shown that tile relaxation of the feasibility requirement enables the reformulation of the problem as a single, yet large-scale and nonconvex, nonlinear optimization model, The exploitation of the special structure of the multiproduct batch plant design model, however, results in orders of magnitude reduction of the number of relaxed dual problems which are required for the computationally efficient application of the global optimization algorithm (GOP) of Floudas and Visweswaran (1990).