Risk management has evolved as one of the key success factors for enterprises especially in the financial services industry. It is highly demanding in terms of business requirements and technical resources, making it an almost ideal application for distributed computing concepts like e.g. grid computing. In this paper we focus on a specific problem—the estimation of covariance matrices that provide a powerful tool for decisions on investments. In this context we analyze different network topologies that the corresponding calculations can be performed on. We derive complexity classes for a distributed calculation scenario on these topologies. As a general result we find an upper and lower bound for the complexity of a distributed calculation in an arbitrary network structure. These results not only provide a different view on grid resource allocation but also make a contribution towards better understanding the business value of grid computing.