A model for cascading failures in complex networks is presented, focusing on how small initial shocks can trigger large-scale failures in infrastructure networks. The model is based on the dynamic redistribution of flow in the network. It shows that the failure of a single node, particularly one with a high load, can lead to a cascade of failures that collapse the entire system. This is especially relevant for real-world networks like the Internet and electrical power grids, which have highly heterogeneous load distributions. The model considers a network as a weighted graph, where nodes have capacities to handle traffic. Initially, the network operates in a stationary state with loads below capacity. When a node fails, the load is redistributed, potentially overloading other nodes. If these nodes cannot handle the increased load, a cascade of failures occurs, leading to a significant drop in network performance. The study applies the model to artificial networks, the Internet, and the electrical power grid of the United States. It shows that the efficiency of the network decreases with lower tolerance parameters (α), and that the collapse of the system is more likely when the failed node has a high initial load. For example, in the Internet, a single node failure can trigger a cascade that reduces network efficiency by up to 30%. The results indicate that cascading failures are more likely in scale-free networks, where a few nodes have very high loads. These networks are more vulnerable to failures of high-load nodes than to random failures. The model also explains real-world events like the 1986 Internet congestion collapse and the 1996 and 2003 power outages, where small initial shocks led to large-scale failures. The study emphasizes the importance of considering cascading failures in the design of complex networks. It shows that even though most networks are stable to small shocks, the failure of a single high-load node can trigger a cascade that collapses the entire system. This highlights the need for robust network designs that account for such vulnerabilities.A model for cascading failures in complex networks is presented, focusing on how small initial shocks can trigger large-scale failures in infrastructure networks. The model is based on the dynamic redistribution of flow in the network. It shows that the failure of a single node, particularly one with a high load, can lead to a cascade of failures that collapse the entire system. This is especially relevant for real-world networks like the Internet and electrical power grids, which have highly heterogeneous load distributions. The model considers a network as a weighted graph, where nodes have capacities to handle traffic. Initially, the network operates in a stationary state with loads below capacity. When a node fails, the load is redistributed, potentially overloading other nodes. If these nodes cannot handle the increased load, a cascade of failures occurs, leading to a significant drop in network performance. The study applies the model to artificial networks, the Internet, and the electrical power grid of the United States. It shows that the efficiency of the network decreases with lower tolerance parameters (α), and that the collapse of the system is more likely when the failed node has a high initial load. For example, in the Internet, a single node failure can trigger a cascade that reduces network efficiency by up to 30%. The results indicate that cascading failures are more likely in scale-free networks, where a few nodes have very high loads. These networks are more vulnerable to failures of high-load nodes than to random failures. The model also explains real-world events like the 1986 Internet congestion collapse and the 1996 and 2003 power outages, where small initial shocks led to large-scale failures. The study emphasizes the importance of considering cascading failures in the design of complex networks. It shows that even though most networks are stable to small shocks, the failure of a single high-load node can trigger a cascade that collapses the entire system. This highlights the need for robust network designs that account for such vulnerabilities.