A physics solver calculates results from governing equations, numerical methods and a defined engineering model. An AI surrogate learns relationships from previous simulation or physical-test results and predicts outputs for new inputs. Surrogates are generally faster, but their reliability is limited to the design space represented in their training data.