<p>Urban climates typically exhibit high local variability and diverge markedly from mesoclimate conditions. Yet, urban microclimate data needed to assess the impacts of urban heat on biodiversity are lacking. We present a scalable modeling approach to generate high-resolution projections of urban bioclimatic conditions using Zurich, Switzerland, as a case study. We demonstrate that globally available land use and remote sensing data can effectively predict spatial patterns in urban bioclimatic conditions. Our findings show that mean annual temperatures can deviate by up to 2 °C from mesoclimate conditions, and that uncorrected mesoclimate data substantially underestimate risks to 182 species of amphibians, birds, butterflies, dragonflies, grasshoppers and trees. By the end of the century, up to half of these species are expected to exceed their climatic tolerance across all habitat patches. Our approach is transferable to cities globally using existing data, marking a major step forward for urban microclimate and biodiversity research.</p>