NVIDIA has not only pioneered high-performance AI GPUs, spearheading the global AI trend, but has also embraced AI technology extensively within its own operations to streamline the GPU chip design process. Bill Dally, the company's chief scientist, disclosed that AI has been seamlessly woven into various stages of chip design, encompassing design exploration, standard cell library development, bug resolution, and verification. Take the migration of a standard cell library as an illustration. Traditionally, this endeavor would necessitate a team of eight individuals dedicating 10 months of effort. However, with the deployment of an AI tool on a single GPU, this task can now be accomplished overnight, with the design quality eclipsing that achieved through manual labor. Furthermore, AI has proven instrumental in overcoming long-standing hurdles, such as the placement of carry-lookahead chains. It proposes layout solutions that would be challenging for human engineers to conceive, leading to a performance enhancement of 20-30%. NVIDIA also leverages two large language models, Chip Memo and Bug Nemo, to empower junior engineers to swiftly address issues and boost their productivity.