Buldyrev et al. (Nature 2010) showed that coupled interdependent networks suffer catastrophically worse failures than isolated ones. Network A (power grid) depends on network B (communication), and vice versa — each node requires support from its counterpart.
When a fraction p of A-nodes fail, their B-counterparts also fail (coupling q), triggering further A-failures in a cascade. An isolated network shows a continuous (2nd-order) percolation transition; coupled networks show a discontinuous (1st-order) collapse — there is an abrupt jump in the giant component size, with no warning.
Left=Network A (red), Right=Network B (blue). Red/blue nodes = alive; grey = failed. The inter-links (dependency edges) are shown as dashed lines.