Novel technologies require the support of larger technological innovation systems (TIS). A key feature of innovation systems are system resources - collective structures such as common standards, support programs, shared expectations or testing facilities all actors can use. System resources emerge either uncoordinated or as a result of strategic action by \'system builders\'. In this paper we explore the conditions of system building. Taking a strategy perspective, we analyze how system building depends on resource constellations at a certain point in time. Drawing from research in the field of stationary fuel cells in Germany, we identify three generic modes, of system building: a) the \"single mode\", in which a system builder uses its own organizational resources to create a system resource, b) the \"partner mode\", in which a system builder joins forces with partners in order to co-create system resources, and c) the \"intermediary mode\", in which a system builder collaborates with other actors to set up an intermediary organization, which then works towards the creation of system resources. We show that the modes were chosen depending on i) what resources were initially available and ii) how they were distributed in the innovation system. Our paper contributes to a more differentiated understanding of system building in the TIS literature and beyond.