The process of actually running an AI model to get answers—as opposed to training, which is the process of building it. Both require significant computing power.
A country doesn't just need compute to train AI — it also needs inference capacity to let millions of people use it every day.
