Marine calcifying phytoplankton known as coccolithophores are a key part of the ocean carbon cycling, contributing to both the export of organic matter via the biological pump and the establishment of the vertical seawater alkalinity gradient known as a the carbonate counter pump. As dominate producers of calcium carbonate production, the rate at which coccolithophores produce calcium carbonate, determines the magnitude of the carbonate counter pump, and is sensitive to future changes in temperature, pH, and nutrient and light availability. Our understanding of spatiotemporal distributions of calcium carbonate production is limited, as measuring calcium carbonate production at sea is labour-intensive, and currently biased both seasonally and spatially. Here we leverage novel machine learning techniques to predict calcium carbonate production on a global scale to a depth of 200m, providing resolution with depth that has not been possible for previous remote sensing based approaches. We find a global integrated annual production of 1.23 PgC/yr [0.78–1.71], with the majority (59%) occurring in the subtropical gyres, challenging the dominance of more-often studied high latitude bloom regions. Our model also shows 18% of global annual CP occurring below the euphotic zone. Our results highlight the need for additional sampling and process studies in the subtropical gyres and at low lights levels/deep in the water column in particular to further constrain our understanding of the ocean carbonate cycle.