Self-organization of nanoparticles is an efficient strategy for producing nanostructures with complex, hierarchical architectures. The past decade has witnessed great progress in nanoparticle self-assembly, yet the quantitative prediction of the architecture of nanoparticle ensembles and of the kinetics of their formation remains a challenge. We report on the marked similarity between the self-assembly of metal nanoparticles and reaction-controlled step-growth polymerization. The nanoparticles act as multifunctional monomer units, which form reversible, noncovalent bonds at specific bond angles and organize themselves into a colloidal polymer. We show that the kinetics and statistics of step-growth polymerization enable a quantitative prediction of the architecture of linear, branched, and cyclic self-assembled nanostructures; their aggregation numbers and size distribution; and the formation of structural isomers. T he focus of nanoscience is gradually shift-ing from the synthesis of individual nano-particles (NPs) to the organization of larger nanostructures. Ensembles of NPs show optical, electronic, and magnetic properties that are determined by collective interactions of indi-vidual NPs (1). To fully understand and exploit these cooperative properties, it is important to achieve control of the structural characteristics of NP ensembles. Self-assembly has emerged as a promising, cost-efficient methodology for gen-erating different types of nanostructures (2–10). In particular, one-dimensional (1D) NP arrays have potential applications in optoelectronics (11–15) and sensing (16, 17). Currently, the lack of mod-els describing the kinetics and statistics of the self-assembly of 1D arrays does not allow the quantitative prediction of their structural features (for instance, the length of NP chains; the degree of branching; or the coexistence of rings, linear chains, and branched structures). Phase diagrams provide useful information on the equilibrium