<p>Despite the popularity and availability of a wide range of cloud service providers, cloud computing still faces a major challenge: the accurate prediction of high-energy consumption, which arises from massive workloads, dynamic resource allocation, and the need for cooling systems. This issue not only increases operational costs for providers but also poses a serious environmental threat. To address this challenge, we propose a hybrid framework that integrates optimized meta-heuristic-based clustering with deep transfer learning to improve the accuracy of energy consumption prediction in cloud environments. The proposed method has been evaluated on multiple real-world datasets, and its performance was measured using Mean Squared Error, Root-Mean-Squared Error, Mean Absolute Percentage Error, and Mean Absolute Error. The results demonstrate that, compared to recent state-of-the-art methods such as the Bidirectional Gated Recurrent Unit, our approach significantly reduces prediction errors—achieving improvements of more than 93% across all metrics—thereby highlighting its effectiveness for accurate and scalable energy consumption prediction in cloud computing.</p>