Aurora, developed by Microsoft Research, is a large-scale foundation model trained on over a million hours of diverse weather and climate data. Leveraging the advantages of foundation model approaches, Aurora provides operational predictions for various atmospheric prediction problems, including those with limited training data, heterogeneous variables, and extreme events. Aurora can generate 5-day global air pollution predictions and 10-day high-resolution weather forecasts in less than a minute, outperforming state-of-the-art traditional simulation tools and the best specialized deep learning models. These results demonstrate that foundation models can transform environmental prediction.