In the context created by the evolution of the EU Copernicus programme and with the increased availability of both space and in-situ Earth Observation data, the EU and ESA Member States started building their own infrastructures to effectively manage the access and the processing of the massive amounts of Copernicus Satellite data. Within the scientific community, several ideas have emerged in order to address the increasing amount of Earth Observation satellite data. Among these concepts, the Data Cube concept and the necessity of dealing with multi-sensor multi-temporal analysis-ready data allowing generation of decision-ready products became obvious.This paper introduces a cloud-native implementation of a national spatio-temporal data cube that leverages Spatio-Temporal Asset Catalogs (STAC) and cloud friendly storage and distribution technologies. This implementation is suitable for providing training data for geospatial analysis using Deep Learning techniques. Additionally, we emphasize advantages of adopting an approach that utilizes STAC catalogs and associated technologies for creating data cubes.