Saudi Arabia’s rapid solar expansion under Vision 2030 increases the need for reliable long-term monthly forecasts of global horizontal irradiance (GHI) and ambient temperature to support siting, procurement, and capacity planning. In this study, we apply and compare a suite of traditional and AI-driven forecasting approaches, Holt-Winters, SARIMA, Random Forest, Support Vector Regression, Prophet, and Bayesian Linear Regression, using monthly data from Dhahran (training: 2018–2022, test: 2023). The integration of machine learning (Random Forest, SVR) and probabilistic AI models (Prophet, BLR) enables nonlinear pattern recognition and uncertainty-aware forecasting critical for sustainable energy planning. After standardized preprocessing and back testing, Random Forest achieves the lowest GHI error (MAPE 4.34%), while Prophet and BLR yield the most accurate temperature forecasts (MAPE 1.61% and 2.67%) with interpretable trends and predictive intervals. AI-based models demonstrate superior adaptability and transparency, underscoring their value for data-driven solar energy planning under Vision 2030.