New low-activation martensitic steels for creep resistance between 600 and 700 degreesC were designed using thermodynamic, kinetic and neural network modelling tools. Suitable compositions for a matrix stabilised by vanadium nitride (VN) particles were firstly suggested on the basis of phase stability calculations using the thermodynamic software MTDATA. A neural network method was then used to predict the creep rupture stress of the possible compositions. It was predicted that a creep rupture stress close to 100 MPa for 100 000 It at 650 degreesC could be achievable. Finally, the precipitation and growth kinetics of VN were calculated using an existing kinetic model. These calculations suggested that a fine (nanometre-scale) and homogeneous distribution of particles could be obtained using a high nucleation site density. This could, in principle, be achieved using thermomechanical treatments and should be even better than alloys produced using the classical normalization and tempering route. (C) 2004 Elsevier B.V. All rights reserved.