“When facing GPU constraints, OpenAI's standard policy is to prioritize allocating compute resources to research over supporting its commercial products.”