This chapter analyzes several issues present in the field of Generative artificial intelligence regarding computing optimization，including the complexity of the technology stack，the poor generality of optimization methods，and the difficulty in identifying performance bottlenecks.It proposes solutions such as independent optimization of large model training and inference，cloud-end collaborative optimization，Generative artificial intelligence application system optimization，and infrastructure and algorithm collaborative optimization.By combining software and hardware optimization techniques，the chapter approaches problem analysis and location from the hardware layer to the application layer in a business and scenario-specific manner.It leverages traditional performance optimization experience and，in conjunction with the characteristics of strong computing，high throughput，and multi-modality of the large model industry ecosystem in the Generative artificial intelligence era，conducts overall computing optimization.Computing optimization emphasizes not only the ultimate optimization of individual module but also the resource collaboration between different modules，ultimately achieving end-to-end cost reduction and efficiency enhancement.