In the realm of cloud computing, optimizing electricity consumption and accurately predicting costs are pivotal for Ecological and cost-effective operations. This paper introduces a pioneering machine learning methodology designed to predict electricity usage and estimate associated expenses in cloud computing environments. Leveraging historical consumption data and diverse parameters, the proposed technique employs a sophisticated algorithmic model, such as ensemble learning or deep neural networks, to forecast electricity consumption patterns. Additionally, it incorporates factors like workload fluctuations, resource allocation strategies, and pricing models to enhance the precision of cost predictions. In this research, we suggest an Extreme Gradient Boosting (XGBoost) model to predict electricity consumption, offload or transfer storage, rates, reducing the costs associated with energy use in data hubs. Based on an actual world dataset provided by the Independent Energy Systems Provider (IESO) inside the efficacy of this in the Canadian province of Ontario approach is evaluated to alleviate data storage in data centers and successfully cut down on energy use. Seventy percent of data goes toward testing, while the remaining thirty percent data is used for training. The results emphasize the potential of this technique to change the management of electricity consumption and cost projections in cloud computing, offering a significant step towards a more efficient and Ecological development of cloud ecosystem.