Large-scale wave modeling is essential for science and society, typically relying on resource-intensive numerical methods to simulate wave dynamics. In this study, we introduce a rolling AI-based method for modeling global significant wave height. Our model achieves accuracy comparable to traditional numerical methods while significantly improving speed, making it operable on standard laptops. This work demonstrates AI's potential to enhance the accuracy and efficiency of global wave modeling.