All existing models to forecast the corrosion performance of reinforced concrete structures exposed to chloride environments are based on one common theoretical concept, namely, a chloride threshold, as a sharply defined trigger for corrosion, followed by a period of active corrosion. We critically review the resulting treatment of corrosion initiation and propagation as two distinct, successive stages. We conclude that this concept presents a major barrier for developing reliable corrosion forecast models, and that a new approach is needed. In reality, steel corrosion in concrete is a continuous process, that is, rarely separable into uncoupled, sequential phases. We propose that the focus be placed on the quantification of the time- and space-variant corrosion rate from the moment steel is placed in concrete until it reaches the end of the service life. To achieve this, a multi-scale and multi-disciplinary approach is required to combine the scientific and practical contributions from materials science, corrosion science, cement/concrete research, and structural engineering.