Amyloid fibrils aggregation is a key pathological feature of many severe degenerative disorders including Alzheimer’s disease and clinical dementia. Moreover, amyloids have been classified as intriguing molecules due to their exceptional strength, sturdiness and elasticity. However, physical models that explain the structural basis of these properties remain largely elusive, preventing the description of the link between their hierarchical structure and physical properties. Here we present an atomistic-based multiscale analysis based on computational materiomics, utilized to predict the structure of the two known polymorphous Alzheimer’s Aβ(1–40) amyloid fibers. We report an analysis of the energies, structural changes and H-bonding for varying amyloid fibril lengths, elucidating their size dependent properties. We also propose an explanation for the different stability of the two morphologies. A structural model of amyloid fibers with lengths of hundreds of nanometers at atomistic resolution is obtained. It predicts the formation of twisted amyloid microfibers in close agreement with experimental results. The approach used here provides a link between the fibril geometry, the chemical interactions and the most stable configuration, and resolves the issue of missing atomistic structures for long amyloid fibers.