The rapid advancement of information technologies has significantly reshaped industrial operations and daily life, leading to a growing demand for responsive and scalable digital services. Among the technologies addressing this growing need, cloud computing has emerged as a foundational infrastructure for delivering on-demand computing resources over the internet. However, its increasing adoption presents complex challenges such as managing dynamic workloads and minimizing virtual machine (VM) usage costs. Therefore, cloud service providers aim to optimize performance and reduce the operational costs of VMs by integrating intelligent scheduling algorithms. In response to this need, the present study explores the use of algorithms, particularly focusing on machine learning driven approaches, to enhance the sustainability and efficiency of cloud systems. Specifically, the study investigates the effectiveness of reinforcement learning through Q-learning for optimizing task scheduling against the traditional Round Robin (RR) scheduling algorithm. The primary objective is to evaluate their performance in minimizing VM usage costs within dynamic and continuously evolving cloud environments. Experimental results indicate that in reducing costs, Q-learning outperforms RR with a 33.14% improvement, demonstrating its superior adaptability and cost efficiency under varying conditions. These insights highlight the potential of reinforcement learning to enable intelligent and cost-aware scheduling strategies in modern cloud computing systems.