Abstract: Sepsis is the most common and dangerous condition in intensive care unit (ICU). Sepsis is caused by various pathogens and leads to systemic inflammation and immune disorders. Severe sepsis will impair multiple organs' function and result in death. Septic patients are highly variable, and their requirements for energy, proteins, and other nutrients are quite different. In addition, the pathophysiological conditions of septic patients are changing very fast, individualized nutritional plans need to be formulated and timely adjusted according to the evolution and progre-ssion of the patient's condition. In recent two decades, international clinical organizations have released many clinical guidelines. However, due to the major large-scale randomized control trials that these guidelines relied on being highly inconsistent, the differences in recommendations between these guidelines are huge, which are especially exhibited in recommendations on the energy and protein targets and the timing of initiation of enteral nutrition. The authors propose that addressing the aforementioned challenges necessitates the integration of artificial intelligence technology. Guided by the principle of data-driven and mechanism-informed integration, a digital twin model should be established to conduct in-depth investigations into problems characterized by spatiotemporal dynamic processes. This digital twin model can dynamically and precisely simulate the effects of different energy-protein intakes for various metabolic and pathophysiological statuses, thereby encompassing monitoring, analysis, prediction, and prescription for the entire nutritional treatment process.Once the digital twin model is deployed in clinical practice, it will be continuously iterated and updated through real-world data and artificial intelligence. This enables the achievement of the goal of "personalized strategy, precise nutrition, and optimized treatment", ultimately reducing patient mortality and improving medical quality.