Cloud computing has become one of the most essential technologies used by companies today. Many organizations rely on cloud services in their daily activities, and the scale of these systems keeps increasing. As data centers grow, their energy use rises too, and practical issues such as resource efficiency, task scheduling, and energy management become harder to handle. Recent work published between 2024 and 2025 shows that artificial intelligence is increasingly used to deal with these challenges. Researchers have explored several techniques to improve the use of cloud resources, including AI-driven optimization, metaheuristic algorithms, reinforcement learning, deep learning, machine learning, and hybrid approaches. These techniques appear in cloud data centers as well as in edge, fog, and mixed architectures. Although many of the proposed models achieve good results, the research remains scattered across different problems, methods, and experimental settings. This paper reviews more than twenty recent studies and organizes them into a multilayer taxonomy covering four dimensions: problem type, AI paradigm, target objectives, and deployment architecture. The review also examines how the selected works consider sustainability, especially energy efficiency, and identifies research gaps that point toward new opportunities for advancing more efficient and responsible cloud systems.