Many infrastructure networks, such as power, water, and natural gas systems, have similar properties governing flows. However, these systems have distinctly different sizes and topological structures. This paper seeks to understand how these different features can emerge from relatively simple design principles. Specifically, we work to understand the conditions under which it is optimal to build small decentralized network infrastructures, such as a microgrid, rather than centralized ones, such as a large high-voltage power system. While our method is simple it is useful in explaining why sometimes, but not always, it is economical to build large, interconnected networks and in other cases it is preferable to use smaller, distributed systems. The results indicate that there is not a single set of infrastructure cost conditions that cause a transition from centralized networks being optimal, to decentralized architectures. Instead, as capital costs increase network sizes decrease gradually, according to a power-law. And, as the value of reliability increases, network sizes increase abruptly-there is a threshold at which large, highly interconnected networks are preferable to decentralized ones.