“OpenAI's compute allocation strategy involves explicitly budgeting a large portion of its resources to the most scalable methods believed to drive general model intelligence.”