“Microsoft has approximately $10 billion in inference-related revenue, primarily from deploying a small number of models including GPT-4, 4.0, 4.0 Mini, and O1.”