A new seismic design philosophy is illuminated, taking advantage of soil "failure" to protect the superstructure. Instead of over-designing the foundation to ensure that the loading stemming from the structural inertia can be "safely" transmitted onto the soil (as with conventional capacity design), and then reinforce the superstructure to avoid collapse, why not do exactly the opposite by intentionally under-designing the foundation to act as a "safety valve"? The need for this "reversal" stems from the uncertainty in predicting the actual earthquake motion, and the necessity of developing new more rational and economically efficient earthquake protection solutions. A simple but realistic bridge structure is used as an example to illustrate the effectiveness of the new approach. Two alternatives are compared: one complying with conventional capacity design, with over-designed foundation so that plastic "hinging" develops in the superstructure; the other following the new design philosophy, with under-designed foundation, "inviting" the plastic "hinge" into the soil. Static "pushover" analyses reveal that the ductility capacity of the new design concept is an order of magnitude larger than of the conventional design: the advantage of "utilising" progressive soil failure. The seismic performance of the two alternatives is investigated through nonlinear dynamic time history analyses, using an ensemble of 29 real accelerograms. It is shown that the performance of both alternatives is totally acceptable for moderate intensity earthquakes, not exceeding the design limits. For large intensity earthquakes, exceeding the design limits, the performance of the new design scheme is proven advantageous, not only avoiding collapse but hardly suffering any inelastic structural deformation. It may however experience increased residual settlement and rotation: a price to pay that must be properly assessed in design.