Soil organic carbon (SOC) is critical for sustaining agricultural productivity, enhancing resilience to climate change, and supporting ecosystem functions, particularly in fragile regions facing increasing aridity like Patagonia. Knowledge of SOC is often represented by decades old, coarse-scale maps or sparse data, limiting its utility for land managers and policymakers. This study leverages a novel SOC database (1,724 samples) integrated with remote sensing and spatial variables in a machine learning model to produce high-resolution (30 m) SOC data that captures decision-relevant scales of variability across diverse land covers and uses. Results revealed that Random Forest modelling performed best in the NW Patagonian mountainous region. Feature selection procedures identified soil depth, spectral indices, and climatic factors such as evapotranspiration and aridity as important co-variates. We found significant heterogeneity in SOC distribution, ranging from the greatest SOC concentration in Nothofagus pumilio forests (132.4 ± 19.2 t ha-1 at 0–30 cm depth), to the lowest in the grasslands of the Monte ecoregion (27.6 ± 8.0 t ha-1). Due to landmass size, the grasslands of the Steppe ecoregion have the most carbon (276.5 million tons), followed by Nothofagus pumilio forests (103.7 million tons). These SOC (t ha-1) estimates agree with other studies, showing little difference for forests (10 %) and grasslands (14 %). The resulting maps of this study provide a critical baseline for evaluating SOC distribution, informing land management strategies, and guiding future climate resilience efforts in Patagonia and other similarly vulnerable regions across the globe.