This study sets out the task of optimizing cloud performance utilizing AI predictive analytics to aid resource allocation and reduce performance bottlenecks in cloud computing. The motive is to obtain practical cloud computing where predictive models can impact the potential for resource allocation and costs. The study obtained publicly available data from Google Cloud BigQuery. It utilized machine learning algorithms to predict cloud resource utilization and degrade performance through time-series prediction and regression models. The main findings reveal that AI-based models, especially the ARIMA time-series forecasting methodology, perform better than traditional linear regression, with a lower (MAE) mean absolute error of 0.054 on validation data versus 0.062 on linear regression. These predictive models allow resource management to be proactive, reducing performance bottlenecks and decreasing costs. The findings suggest an example of the viability and effectiveness of AI predictive analytics in cloud performance optimization, with a process that can eventually scale to more favourable approaches to cloud infrastructure management. This study reaffirms that if organizations move towards AI-driven predictive analytics to assist in running cloud systems, they will reap significant improvements in cloud system reliability, efficiency, and cost optimization within their current practices.