The paper introduces CloudAIBus, a new testbed for AI-driven cloud computing environments, designed to optimize resource allocation and enhance efficiency. The authors address the challenge of overprovisioning CPU resources, which leads to financial inefficiencies, by leveraging deep learning models like DeepAR for accurate resource usage forecasting. CloudAIBus employs Amazon SageMaker for scalable and efficient training, and Google Colab for performance evaluation. The proposed approach outperforms baseline methods (LSTM and ARIMA) in terms of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Mean Squared Error (MSE). It significantly reduces the percentage of unused CPUs, demonstrating its effectiveness in reducing over-provisioning. The system aims to improve economic and environmental sustainability in cloud operations, aligning with global efforts to achieve sustainable IT practices. The main contributions include the development of CloudAIBus and the utilization of DeepAR for predictive resource allocation, addressing the gaps in dynamic scaling and environmental impact.The paper introduces CloudAIBus, a new testbed for AI-driven cloud computing environments, designed to optimize resource allocation and enhance efficiency. The authors address the challenge of overprovisioning CPU resources, which leads to financial inefficiencies, by leveraging deep learning models like DeepAR for accurate resource usage forecasting. CloudAIBus employs Amazon SageMaker for scalable and efficient training, and Google Colab for performance evaluation. The proposed approach outperforms baseline methods (LSTM and ARIMA) in terms of Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and Mean Squared Error (MSE). It significantly reduces the percentage of unused CPUs, demonstrating its effectiveness in reducing over-provisioning. The system aims to improve economic and environmental sustainability in cloud operations, aligning with global efforts to achieve sustainable IT practices. The main contributions include the development of CloudAIBus and the utilization of DeepAR for predictive resource allocation, addressing the gaps in dynamic scaling and environmental impact.