Why could AI become dramatically cheaper to run? Atlassian’s CEO argues that the sheer amount of money and talent going into chips, data centres and model efficiency could drive inference costs down sharply over time.