This paper provides an ethnographical study of the ways in which infrastructure matters in the production of knowledge in the social worlds of rare diseases. We analyse the role played by a relational database in this respect, which exists at the crossroads of a large and complex network of individuals, institutions, and practices. This database forms part of a " boundary infrastructure " , in which knowledge production constitutes one output of infrastructural work, that needs to be articulated with other kinds of activities and matters of concern. We analyse how members of the network negotiate the place and forms of knowledge production in relation to these other purposes, and highlight the political nature of the distinction between knowledge and information, which frames collective action. We also show how infrastructural inversion serves to articulate knowledge production with other forms of mobilisation, thereby shaping and reconfi guring the boundary infrastructure as a whole.