Ocean phytoplankton biomass is predicted to decline in Earth system models, due in large part to an expansion of nutrient-deplete ocean regions. However, the representation of ecosystems in these models is simplified and based on only a few functional types. As a result, they fail to capture the high diversity known to exist within and across phytoplankton communities. Here we present an assessment of the global biogeography of the very abundant but little studied picoeukaryotic phytoplankton by analysing a global abundance dataset with a neural-network-derived quantitative niche model. Combining this niche model with previous assessments of the distribution of Prochlorococcus and Synechococcus, we find that different cell sizes among picophytoplankton lineages are clearly partitioned into latitudinal niches. In addition, picophytoplankton biomass increases along a temperature gradient in low-latitude regions. We infer that future warmer ocean conditions can lead to elevated phytoplankton biomass in regions that are already dominated by picophytoplankton. Finally, we demonstrate that elevated upper-ocean nutrient recycling and lower nutrient requirements of phytoplankton have the potential to support increasing low-latitude phytoplankton biomass with future warming. Picophytoplankton are partitioned into niches, globally, and their abundance may increase as ocean temperatures rise, suggest analyses of a global abundance dataset with a neural-network-based niche model.