The output or answer that a machine learning model produces when given new data. Like a weather forecast that uses past patterns to estimate tomorrow's conditions.
An e-commerce site uses ML predictions to estimate which products a customer will likely purchase based on their browsing history.
All four clouds provide managed ML platforms that host trained models and return predictions via online (real-time) endpoints or batch jobs. They differ mainly in surrounding ecosystem, deployment options, and pricing models, but the core capability—serving model inference—is comparable.