I analyze the process steps for a GPU New Product Introduction (NPI), i.e. the process for getting a new chip design from tapeout to mass production, focusing on existing technology nodes. The speed of NPI is important, as it determines the strength of the compute-compute feedback loop, which in turn influences the speed of AI development. I gives a qualitative overview of the NPI process, with detailed descriptions of individual process steps, present time estimates for individual NPI steps, and list potential levers for acceleration. I find that a further acceleration of the NPI process would likely be possible if design firms are willing to spend more resources on the process than currently economically viable.