Vehicular Edge Computing (VEC) enhances traditional edge computing by integrating computational resources within the vehicular environment, allowing vehicles to offload tasks to nearby Roadside Units (RSUs). However, as urban traffic grows, the distribution of workload across RSUs can become imbalanced, with high-congestion areas facing heavy computational demands while other regions remain underused. This paper addresses the task load balancing problem by implementing a future-aware resource orchestration strategy that employs the Seasonal AutoRegressive Integrated Moving Average with eXogenous regressors (SARIMAX) model to forecast traffic conditions. We determine the SARIMAX hyperparameters using an automated “auto-ARIMA” procedure. We assess the performance of SARIMAX predictions in both Monte Carlo and Rolling Forecasting scenarios using real vehicular traffic data from Bucharest, Romania, obtained through the HERE Traffic API. Furthermore, we present a new traffic dataset containing over 120 h of street-level traffic data from Bucharest, which is publicly available on the Kaggle platform. The simulation results demonstrate the effectiveness of the proposed strategy in reducing task processing delays through proactive workload balancing, achieving a reduction of overloaded RSUs by over 60% and a 20–25% improvement in the Gini coefficient.