This code and associated data uses a physics informed machine learning approach to facilitate large-scale global simulations for global agrivoltaic food-energy productivity. High fidelity physics-based models are used to create a synthetic data set for the training of artificial neural networks. Based on this model, we predict where the agrivoltaics is most suitable globally, how to choose module land coverage and crop across climate zones, and agrivoltaic potential in the context of global warming.