This paper discusses the application of Physics-Informed Machine Learning (PIML) in metal additive manufacturing (AM). PIML integrates physics knowledge into machine learning (ML) models to improve their reliability, interpretability, and predictive accuracy. The paper classifies PIML into three categories: Physics-Informed Domain Knowledge, Simulation-Based Input Data, and Physics-Guided Model Training. The Physics-Informed Neural Network (PINN) is a notable example of Physics-Guided Model Training, which can produce more explainable and reliable results even with noisy data. The paper also discusses the limitations of PINN and potential solutions. Additive manufacturing has advanced significantly, enabling the production of complex structures with less material waste. Metal AM processes are crucial in aerospace, dentistry, healthcare, and other fields. Common metal AM processes include Directed Energy Deposition (DED), Powder Bed Fusion (PBF), and Binder Jetting. Modeling metal AM processes is essential for understanding the process-structure-property relationship and ensuring product quality. Physics-based models simulate and predict AM process behavior using physical principles, but they may not fully capture process variations due to simplifications or high computational costs. Data-driven methods rely on experimental observations and can handle high-dimensional data, but they lack physical interpretability. PIML combines the strengths of both approaches, integrating physical principles into ML models to address the limitations of traditional methods. PIML models are suitable for modeling advanced metal AM processes, as they can respond to sudden changes during online monitoring and process control. PIML has been successfully applied in metal AM, demonstrating high accuracy in real-time prediction, particularly in high-dimensional temperature fields and melt pool fluid dynamics.This paper discusses the application of Physics-Informed Machine Learning (PIML) in metal additive manufacturing (AM). PIML integrates physics knowledge into machine learning (ML) models to improve their reliability, interpretability, and predictive accuracy. The paper classifies PIML into three categories: Physics-Informed Domain Knowledge, Simulation-Based Input Data, and Physics-Guided Model Training. The Physics-Informed Neural Network (PINN) is a notable example of Physics-Guided Model Training, which can produce more explainable and reliable results even with noisy data. The paper also discusses the limitations of PINN and potential solutions. Additive manufacturing has advanced significantly, enabling the production of complex structures with less material waste. Metal AM processes are crucial in aerospace, dentistry, healthcare, and other fields. Common metal AM processes include Directed Energy Deposition (DED), Powder Bed Fusion (PBF), and Binder Jetting. Modeling metal AM processes is essential for understanding the process-structure-property relationship and ensuring product quality. Physics-based models simulate and predict AM process behavior using physical principles, but they may not fully capture process variations due to simplifications or high computational costs. Data-driven methods rely on experimental observations and can handle high-dimensional data, but they lack physical interpretability. PIML combines the strengths of both approaches, integrating physical principles into ML models to address the limitations of traditional methods. PIML models are suitable for modeling advanced metal AM processes, as they can respond to sudden changes during online monitoring and process control. PIML has been successfully applied in metal AM, demonstrating high accuracy in real-time prediction, particularly in high-dimensional temperature fields and melt pool fluid dynamics.