One of the most costly aspects of running an AI company is what is referred to as “compute,” which typically describes the infrastructure and processing power required to build, train, test and then make available its offerings to the public, such as a chatbot.