The implementation of a finite element program used for nonlinear industrial applications on parallel computer architectures with local memory such as the Supernode network of transputers is described in this paper. First, two strategies for the solution of the linearized system of equations are outlined and tested. Then coupled problems, in particular themomechanically coupled forming processes, are considered. Moreover, a new method for the automatic spatial decomposition of the discretized finite element domain is presented and compared with other approaches. Finally, strategies for adaptive mesh refinement are discussed.