The steel supply chain generally comprises a steel producer, manufacturing semi–finished products, and a steel user, manufacturing formed components. A review is presented of the deformation models needed by the steel industry, considering the current and expected future requirements of the end users. A trend that is expected to accelerate is their increased use of automation. This imposes a tighter dimensional tolerance, which requires, among other things, models that predict rolling loads and physically based models being introduced, offering distinctive benefits that will be discussed. A narrow distribution of mechanical properties is needed, requiring techniques that link deformation parameters such as strain, time and temperature, through microstructure models to property prediction, and this should include toughness as well as yield strength. Being able to predict toughness at all points in a rolled product would significantly reduce waste and costs associated with destructive testing. The need for lightweight designs is leading to wider application of high–strength steels. Because of the requirements of lower cost and good weldability, the high strengths are being achieved through leaner chemistries, using controlled thermomechanical processing, where the transformation from austenite needs to be incorporated into the models. This is a major difference between steel and aluminium. Laser fabrication techniques are increasingly being used; these require high degrees of flatness in the steel, including after it has been cut, where non–uniform residual stress distribution can lead to distortion. Models that predict residual stresses are needed, and, since the customer might buy the steel in a coil and uncoil it himself, the residual stress model and data will need to be shared. Customers will become more than recipients of steel; they will share in the knowledge embedded in the steel, and the ability to do this, along the varying deformation processes used by the steel industry, is a challenge for those developing the modelling techniques.