Background: Childhood overweight and obesity are increasingly prevalent, particularly in low- and middle-income countries, and predispose individuals to chronic diseases. Conventional clinical assessments often overlook the multifactorial nature of obesity, leading to generalized and less effective interventions. This study evaluated the effectiveness of an AI ensemble model to predict school-age nutritional status from preschool data in Mexico. Methods: Data were obtained from the 2018 Mexican National Health and Nutrition Survey, including 8,327 healthy children under 5 years. A total of 75 clinical, demographic, socioeconomic, perinatal and behavioral variables were collected. The prediction task was framed as a four-class problem using WHO weight-for-height z-score. Missing values were imputed with a random forest method within the cross-validation folds to prevent data leakage. Three base models (multinomial logistic regression, support vector machine, and random forest) were trained, optimized and stacked into a two-hidden-layer neural network. Performance was assessed using accuracy, precision, F1-score, and AUC, reported as macro-averages across the four classes. Model interpretability was examined using LIME. Results: Class distribution at age five was imbalanced (undernutrition 1.3%, normal 83.5%, overweight 6%, obesity 9%). The ensemble achieved 95% accuracy, 75% macro-precision, 81.9% macro-F1 score, and 0.98 macro-AUC in the multiclass task. Key predictor included parental socioeconomic and educational levels, prenatal care, and child anthropometric z-scores. Conclusions: A multi-class ensemble model can accurately predict school-age nutritional status form preschool information while providing interpretable risk factors. Integrating such models into public health surveillance could enable earlier, individualized prevention and reduce the long-term burden of childhood obesity.