The robustness of a network is depending on the type of attack we are considering. In this paper we focus on the spread of viruses on networks. It is common practice to use the epidemic threshold as a measure for robustness. Because the epidemic threshold is inversely proportional to the largest eigenvalue of the adjacency matrix, it seems easy to compare the robustness of two networks. We will show in this paper that the comparison of the robustness with respect to virus spread for two networks actually depends on the value of the effective spreading rate τ. For this reason we propose a new metric, the viral conductance, which takes into account the complete range of values τ can obtain. In this paper we determine the viral conductance of regular graphs, complete bipartite graphs and a number of realistic networks. © IFIP International Federation for Information Processing 2009.