Grid computing has become a driving force of scientific computing. By combining the resources of distributed computing centers under a unified management and offering their computing power (nearly) as seamless as electrical power, grid computing is seen as an important step to solve large scale-out problems. Storage grids and storage clusters try to transfer this idea into the storage domain: By combining the capacity, computing and communication power of storage appliances under a single middleware, storage grids promise to scale in capacity and bandwidth while keeping the administration overhead low. Futhermore, they claim to be more cost effective than standard storage architectures, because they can be built up from standard components. Inside this paper, we analyze the costs related to the least expensive storage grid architecture, which is build solely from standard server components and we show that it is possible to build large-scale architectures with enterprise functionality from standard hardware.