Antimicrobial peptides (AMPs) are an ancient, powerful, and ubiquitous component of the innate immune defence in all domains of life and play a key role in controlling the human microbiome. Many AMPs are known to selectively target and form pores in microbial membranes, killing a wide variety of pathogens at low micromolar concentrations. Fundamental questions remain, however, regarding the molecular mechanisms of membrane targeting, pore formation and function, and the extent to which poration contributes towards antimicrobial activity. Here we report an experimentally guided and validated unbiased long-timescale simulation methodology that yields the complete mechanism of spontaneous pore assembly in the membrane for the amphiphilic pore-forming AMP maculatin at atomic resolution. Rather than a single well defined pore, maculatin was found to form an ensemble of well defined temporarily functional channels that continuously form and dissociate in the membrane. All channels are formed from highly symmetric low-oligomeric assemblies of up to 8 membrane-spanning peptides that mimic integral membrane protein channels in structure. Each channel has a different architecture, functional lifetime, and ion conductivity and overall membrane permeabilization is dominated by higher order oligomers, which form ion channels that also conduct water at high rates. The highest order oligomers were also found to efficiently conduct small dyes. Channel architectures as well as their relative weighting in the ensemble are sensitive to minor mutations as well as variation of lipid tail length. All pores are formed spontaneously by the consecutive addition of individual helices to a transmembrane helix or helix bundle, in contrast to current poration models postulated for AMPs. The diversity of channel architectures formed by a single sequence is remarkable and could explain why no sequence-function relationships in AMP sequences have been discovered to date. Structural ensembles formed by AMP may be the key to preventing bacterial resistance.